In the acronym RESPECT for active listening, the letter P stands for
(b) Pay attention to what the speaker is saying P = Pay attention to what the speaker is saying (look at the speaker, avoid distractions).
Minimum Learning Support Material · CBSE 2026-27
Notes · Numericals with working · 889 solved questions · Flashcards · Quiz · Practice Paper Maker
CBSE Sample Question Paper 2026-27: 21 questions, 50 marks, 2 hours
| Question | What comes | Attempt | Marks |
|---|---|---|---|
| Q1 | 6 objective questions on Employability Skills | any 4 | 1 x 4 = 4 |
| Q2 to Q5 | 6 objective questions each on the AI units (MCQ, True/False, Assertion-Reason, match, small numericals) | any 5 in each | 4 x 5 = 20 |
| Q6 to Q10 | Employability Skills, 20 to 30 words each | any 3 of 5 | 2 x 3 = 6 |
| Q11 to Q16 | one question from each AI unit, 20 to 30 words each | any 4 of 6 | 2 x 4 = 8 |
| Q17 to Q21 | Units 2, 5, 6, 7 and 8, 50 to 80 words each | any 3 of 5 | 4 x 3 = 12 |
The paper has 21 questions and is of 2 hours. You answer 15 questions: all of Q1 to Q5 (with the choice inside each) and 10 out of Q6 to Q21.
Unit 1 (Python Programming-II) and Unit 4 (AI with Orange Data Mining Tool) are assessed in the practical only.
One common blueprint, as per the CBSE Sample Question Paper 2026-27. All the practice papers of this kit follow it. Each cell shows the number of questions given from that unit.
| Unit | 1mark | 2marks | 4marks | Totalquestions |
|---|---|---|---|---|
| Part ACommunication Skills-IV | 1 | 1 | - | 2 |
| Part ASelf-Management Skills-IV | 2 | 1 | - | 3 |
| Part AICT Skills-IV | 1 | 1 | - | 2 |
| Part AEntrepreneurial Skills-IV | 1 | 1 | - | 2 |
| Part AGreen Skills-IV | 1 | 1 | - | 2 |
| Part BUnit 2: Data Science Methodology | 5 | 1 | 1 | 7 |
| Part BUnit 3: Making Machines See | 5 | 1 | - | 6 |
| Part BUnit 5: Big Data and Data Analytics | 3 | 1 | 1 | 5 |
| Part BUnit 6: Understanding Neural Networks | 5 | 1 | 1 | 7 |
| Part BUnit 7: Generative AI | 3 | 1 | 1 | 5 |
| Part BUnit 8: Data Storytelling | 3 | 1 | 1 | 5 |
| Questions given | 30 | 11 | 5 | 46 |
| To be answered | 24 | 7 | 3 | 34 |
| Marks | 24 | 14 | 12 | 50 |
Where the 50 marks come from, easiest first
10 marks · five units · detailed chapters
Effective communication at work is not only speaking well - it is also listening well, writing correctly and choosing the right words. This chapter covers active listening, parts of speech, sentence types, active and passive voice and capitalisation.
Active listening is listening with full attention, understanding the message, remembering it and responding thoughtfully. A good listener makes the speaker feel valued and avoids mistakes caused by misunderstanding.
| Stage | What happens | Example: a doctor's advice |
|---|---|---|
| Receiving | paying attention to the sounds and signals | you hear "take one tablet twice a day" |
| Understanding | interpreting the words | you understand: morning and night |
| Remembering | keeping it in memory | you note it in your phone |
| Evaluating | thinking about it critically | "Should I take it before or after food?" |
| Responding | verbal or non-verbal reply | you ask the doctor and nod |
phone away, face the speaker
look at the speaker
nod, smile, "I see"
focus on the words and feelings
understand the speaker's point of view
ask questions, paraphrase
adjust to their pace and tone
| Barrier | Example |
|---|---|
| Distractions | phone notifications, noise |
| Prejudice | "This person never says anything useful" |
| Interrupting | completing the speaker's sentences |
| Daydreaming | thinking about something else |
| Preparing your reply | not hearing the full message |
| Information overload | too much information at once |
Manager: "Send the revised budget to finance by Thursday noon and copy me." Employee: "So, I'll e-mail the revised budget to finance before 12 on Thursday with you in CC. Right?" Paraphrasing shows listening and avoids errors.
| Part | Function | Example |
|---|---|---|
| Noun | naming word | Anil works in Pune. |
| Pronoun | replaces a noun | He likes it. |
| Verb | action / state | She presented the report. |
| Adjective | describes a noun | a detailed report |
| Adverb | describes a verb, adjective, adverb | spoke confidently |
| Preposition | shows relation | the file on the desk |
| Conjunction | joins | fast but careful |
| Interjection | emotion | Hurray! We won. |
| Type | Example |
|---|---|
| Proper | India, Ravi, Monday |
| Common | city, boy, day |
| Collective | team, class, herd |
| Abstract | honesty, courage, joy |
| Material | gold, wood, cotton |
| Type | Function | Example |
|---|---|---|
| Statement (declarative) | gives information | The exam begins at ten. |
| Question (interrogative) | asks something | When does the exam begin? |
| Emotion (exclamatory) | shows strong feeling | What a wonderful result! |
| Order (imperative) | command or request | Please open your books. |
A sentence has a subject (who / what) and a predicate (what the subject does). Many sentences also have an object (who / what receives the action): *The team (subject) won (verb) the trophy (object).*
| Tense | Active | Passive |
|---|---|---|
| Simple present | She writes reports. | Reports are written by her. |
| Simple past | The chef cooked the meal. | The meal was cooked by the chef. |
| Future | They will open the shop. | The shop will be opened by them. |
| Present continuous | He is painting the wall. | The wall is being painted by him. |
Use active voice for clear, direct business writing. Use passive voice when the doer is unknown or less important: "The road was repaired last week."
March, Monday
"I am ready."
Priya, Delhi, the Ganga, Infosys
Dr. Sharma, Prime Minister
first word of every sentence
| Mark | Use | Example |
|---|---|---|
| Full stop (.) | end of a statement | The bus left. |
| Comma (,) | pause; lists | pens, pencils, files |
| Question mark (?) | question | Are you ready? |
| Exclamation (!) | strong feeling | Great job! |
| Apostrophe (') | possession / contraction | Riya's bag; don't |
| Quotation marks (" ") | exact words | He said, "Wait." |
Receiving, understanding, remembering, evaluating and responding.
(i) The plan was approved by the manager. (ii) The classroom is being cleaned by the students.
At Level IV, self-management focuses on motivation and positive attitude, result orientation, self-awareness and personality. These help a young person succeed in higher studies and the workplace.
Motivation is the internal or external force that starts, directs and keeps up our behaviour towards a goal.
People are motivated first by basic needs; when those are met, higher needs motivate them. A worker who has a secure job (safety) may next be motivated by recognition (esteem).
A positive attitude is a way of thinking that focuses on solutions and possibilities. It improves health, relationships and performance.
note what is going well
"I can learn this"
spend time with positive people
exercise, sleep, good food
ask "What can I do now?"
kindness improves mood
Result orientation means focusing on achieving the desired outcome, not only on doing activities. A result-oriented person sets clear goals, plans, uses the right tools, tracks progress and finishes on time.
Goal: score 90 in AI. Plan: one chapter a week, 20 questions daily from the kit, a weekly sample paper. Tools: notes, question bank, timer. Track: marks in each sample paper. Improve: re-study weak chapters. Focus stays on the result - 90 marks - not just on "hours studied".
Self-awareness is the clear understanding of our own personality, strengths, weaknesses, thoughts, beliefs, emotions and motivations.
| Aspect | Question | Benefit |
|---|---|---|
| Strengths | What am I good at? | choose suitable careers |
| Weaknesses | What do I need to improve? | plan training |
| Emotions | What triggers my anger or stress? | better self-control |
| Values | What is important to me? | right decisions |
Ways: reflection and journaling, feedback from others, personality tests, meditation, trying new experiences.
Personality is the set of qualities - thoughts, feelings and behaviour patterns - that make a person unique. Psychologists describe it with the Big Five (OCEAN) factors:
curious, creative, likes new ideas (low: prefers routine)
organised, careful, dependable (low: careless)
outgoing, energetic, talkative (low: reserved)
kind, helpful, cooperative (low: competitive)
anxious, easily upset (low: calm and stable)
| Factor | Workplace example |
|---|---|
| High openness | designer who tries new styles |
| High conscientiousness | accountant who never misses deadlines |
| High extraversion | salesperson who enjoys meeting people |
| High agreeableness | nurse who cares for patients |
| Low neuroticism | air-traffic controller who stays calm under pressure |
A personality disorder is a long-lasting pattern of thinking and behaviour that is very different from what is expected and causes difficulties in daily life and relationships. They are diagnosed only by qualified professionals; the CBSE book groups them into three clusters:
| Cluster | Description | Disorders |
|---|---|---|
| A - odd, suspicious | unusual thinking | paranoid, schizoid, schizotypal |
| B - emotional, impulsive | dramatic, unstable behaviour | antisocial, borderline, histrionic, narcissistic |
| C - anxious, fearful | worry and fear | avoidant, dependent, obsessive-compulsive |
This is for awareness only. We must never label classmates or others; if someone is struggling, encourage them to talk to a counsellor or doctor.
Intrinsic motivation comes from within, such as interest and enjoyment of the task. Extrinsic motivation comes from outside, such as rewards, marks or praise.
Openness - curiosity and creativity; Conscientiousness - being organised and responsible; Extraversion - being outgoing and energetic; Agreeableness - being kind and cooperative; Neuroticism - tendency to feel anxious or upset (low neuroticism means emotional stability).
A spreadsheet is application software that stores data in rows and columns and performs calculations, analysis and charts. Examples: LibreOffice Calc (.ods), Microsoft Excel (.xlsx), Google Sheets.
household and business accounts
totals, averages, ranks of students
stock in a shop
pay and deductions
sales trends, survey results
| Term | Meaning |
|---|---|
| Workbook | the whole file |
| Worksheet | one sheet (tab) in the workbook |
| Row | horizontal line, numbered 1, 2, 3 ... |
| Column | vertical line, lettered A, B, C ... |
| Cell | meeting point of a row and column |
| Cell address | column letter + row number, e.g. B5 |
| Range | a group of cells, e.g. A1:A10 |
| Active cell | the selected cell (thick border) |
Types of data: text (labels), numbers, dates, formulas.
A formula always starts with =: =B2+C2, =A1*10%. Operators: + - * / ^ (power).
Marks in B2 = 78, C2 = 85, D2 = 90. In E2: =B2+C2+D2 gives 253; in F2: =E2/3 gives 84.33. Copying E2 down with the fill handle repeats the formula for every student using relative references.
| Type | Example | On copying |
|---|---|---|
| Relative | A1 | changes with position |
| Absolute | $A$1 | never changes |
| Mixed | $A1 or A$1 | only one part changes |
| Function | Purpose | If A1 = 20, A2 = 30, A3 = 50 |
|---|---|---|
| =SUM(A1:A3) | total | 100 |
| =AVERAGE(A1:A3) | mean | 33.33 |
| =MAX(A1:A3) | largest | 50 |
| =MIN(A1:A3) | smallest | 20 |
| =COUNT(A1:A3) | how many numbers | 3 |
| =IF(A1>=33,"Pass","Fail") | condition | Pass |
| =ROUND(33.333,1) | rounding | 33.3 |
Formatting: number format (currency, percentage, date), bold headings, borders, fill colour, conditional formatting (e.g. marks below 33 in red), merge and centre for titles, wrap text, freeze panes to keep headings visible.
compare items: marks of students
trend over time: monthly sales
parts of a whole: expense share
relation between two variables
Steps: select data > Insert > Chart > choose type > add title, axis labels and legend.
Presentation software creates slides with text, pictures, charts, audio, video and animation for talks and lessons. Examples: LibreOffice Impress (.odp), Microsoft PowerPoint (.pptx), Google Slides.
| Feature | Use |
|---|---|
| Slide layout | title slide, title and content, two content, blank |
| Slide master | same design on all slides |
| Transition | effect when moving to the next slide |
| Animation | effect on an object within a slide |
| Views | normal, outline, notes, slide sorter, slide show |
| Speaker notes | notes for the presenter |
| F5 / Shift + F5 | start from the first / current slide |
one idea per slide, about 6 lines with 6 words each, large readable fonts, contrasting colours, pictures instead of long text.
Sorting arranges all rows in ascending or descending order. Filtering shows only the rows that meet a condition and hides the others temporarily.
(i) 75 (ii) 25 (iii) 40 (iv) 3.
Entrepreneurship is the process of creating a new venture by identifying an opportunity, arranging resources and taking risks, in order to earn profit and add value to society. The person is the entrepreneur; the venture is the enterprise.
sells services: salon, courier, coaching
buys and sells: grocery, wholesale
manufactures goods: textile mill, food processing
farming and allied: dairy, poultry, horticulture
technology-based: software, app, electronics
solves social problems: affordable healthcare
| Competency | Meaning | Example |
|---|---|---|
| Decisiveness | takes timely, firm decisions | chooses a supplier after comparing quotes |
| Initiative | acts without being told | starts home delivery before competitors |
| Interpersonal skills | builds good relations | keeps customers and staff happy |
| Perseverance | keeps going after setbacks | relaunches after the first product fails |
| Organisational skills | plans and uses resources well | manages stock, cash and staff schedules |
| Barrier | Example | How to overcome |
|---|---|---|
| Environmental | no good roads or electricity in the area | choose a suitable location, solar power |
| No / faulty business plan | no study of costs or customers | prepare a proper plan with market research |
| Personal | fear of failure, low confidence | training, mentors, start small |
| Financial | no money to start | savings, Mudra loan, investors |
During water shortages in a town, Kiran sets up a small RO water plant and supplies 20-litre cans to homes and offices. A problem for many became her business opportunity.
A business plan is a written document that describes the business idea, the market, operations and finances. It guides the entrepreneur and convinces banks and investors.
| Part | Answers |
|---|---|
| Executive summary | What is the business in short? |
| Market analysis | Who are the customers and competitors? |
| Marketing plan | How will we reach customers? (4 Ps: Product, Price, Place, Promotion) |
| Operations | Where and how will we produce and deliver? |
| Financial plan | How much money is needed? Expected sales, costs and profit? |
A candle maker has fixed costs (rent, machine) of Rs 12,000 a month. Each candle sells for Rs 50 and costs Rs 30 to make. Break-even = 12,000 / (50 - 30) = 600 candles. Selling more than 600 candles a month gives profit.
| Scheme | Support |
|---|---|
| Startup India | tax benefits, easy registration, funding support |
| PM Mudra Yojana | loans: Shishu (up to Rs 50,000), Kishor, Tarun |
| Stand-Up India | loans for SC/ST and women entrepreneurs |
| Atal Innovation Mission | incubation centres, tinkering labs |
| Udyam registration | easy MSME registration |
Harish Hande saw that millions of rural homes had no reliable electricity. In 1995 he started SELCO to provide solar lighting to poor families, working with banks so that customers could pay in small instalments. SELCO has lit lakhs of homes, reduced kerosene use and created local jobs for technicians. It shows initiative, perseverance and the social type of entrepreneurship.
knows the problem to solve
continued for years before success
worked with banks and communities
new payment and service models
Service (salon), trading (grocery store), industrial (food-processing unit), agricultural (dairy farm), technical (software company), social (low-cost healthcare). (Any four.)
Environmental barriers - lack of raw material, labour, roads; no or faulty business plan; personal barriers - fear of failure and self-doubt; financial barriers - lack of capital. Others: social pressure, legal formalities.
Green jobs are decent jobs that help protect or restore the environment - by saving energy and raw materials, reducing greenhouse-gas emissions, minimising waste and pollution, and protecting ecosystems (International Labour Organization).
organic farming, soil health, drip irrigation
electric vehicles, CNG, metro, cycling
rainwater harvesting, watershed, recycling
solar, wind, small hydro, biogas
nature tourism that protects local ecology
energy-efficient, GRIHA-rated buildings
segregation, composting, recycling
simple, local, low-cost tools (solar dryers)
| Field | Green job | What they do |
|---|---|---|
| Agriculture | organic farming advisor | guides farmers to use natural inputs |
| Transport | EV technician | maintains electric vehicles and chargers |
| Water | rainwater harvesting technician | builds and maintains RWH systems |
| Energy | solar PV installer | installs rooftop and grid solar panels |
| Tourism | eco-tour guide | leads nature tours responsibly |
| Construction | green building consultant | designs buildings that save energy and water |
| Waste | waste-management supervisor | runs segregation and recycling |
Greenhouse gases (GHGs) trap heat in the Earth's atmosphere. The main ones are carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), water vapour and fluorinated gases. Oxygen and nitrogen are not greenhouse gases.
| Gas | Main sources |
|---|---|
| Carbon dioxide | burning coal, oil, gas; deforestation |
| Methane | paddy fields, cattle, landfills |
| Nitrous oxide | chemical fertilisers |
| Fluorinated gases | refrigerators, air conditioners |
India launched the NAPCC in 2008 with eight national missions:
| No. | Mission | Focus |
|---|---|---|
| 1 | National Solar Mission | solar energy |
| 2 | National Mission for Enhanced Energy Efficiency | saving energy in industry and homes |
| 3 | National Mission on Sustainable Habitat | green cities, waste, public transport |
| 4 | National Water Mission | water conservation and efficiency |
| 5 | National Mission for Sustaining the Himalayan Ecosystem | protecting the Himalayas |
| 6 | National Mission for a Green India | afforestation |
| 7 | National Mission for Sustainable Agriculture | climate-resilient farming |
| 8 | National Mission on Strategic Knowledge for Climate Change | research |
In Rajasthan, women from self-help groups were trained as "Solar Mamas" to assemble and repair solar lanterns. Villages got clean light instead of kerosene, the women earned income, and smoke and CO2 emissions fell - economic, social and environmental benefits together.
| Skill type | Examples |
|---|---|
| Technical | installing solar panels, operating recycling machines, energy auditing |
| Knowledge | environmental laws, climate science basics |
| Soft skills | problem solving, teamwork, awareness |
| Values | responsibility, care for nature |
The Skill Council for Green Jobs (SCGJ) under Skill India trains youth for such roles.
| Programme | Green jobs created |
|---|---|
| PM-KUSUM | solar pumps for farmers - installers, technicians |
| Rooftop solar (PM Surya Ghar) | solar installers and maintenance staff |
| Swachh Bharat Mission | waste collectors, compost-plant operators |
| Electric mobility | EV assemblers, battery recyclers, charging-station staff |
| Namami Gange | sewage-treatment plant operators |
Green jobs are jobs that protect or restore the environment by saving energy and resources and reducing pollution. Sectors: renewable energy, organic agriculture, green transport, waste management.
They limit greenhouse-gas emissions; reduce waste and pollution; protect and restore ecosystems; use energy and raw materials efficiently; and help communities adapt to climate change, while also creating employment.
Ten marks: four one-mark questions (out of six) and three two-mark questions (out of five). One question comes from each unit, and two one-markers from Self-Management.
| Topic | Points |
|---|---|
| Motivation | Intrinsic: for your own joy. Extrinsic: for reward or praise. |
| Result orientation | set clear goals, make an action plan, use the right tools, work in the given time |
| Big Five personality factors | Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism |
| Personality disorders | Cluster A (suspicious): paranoid, schizoid, schizotypal. Cluster B (emotional, impulsive): antisocial, borderline, histrionic, narcissistic. Cluster C (anxious): avoidant, dependent, obsessive-compulsive. |
| Topic | Points |
|---|---|
| Spreadsheet | rows, columns, cells (B5); formula begins with =; functions SUM, AVERAGE, MAX, MIN, COUNT |
| Sort and filter | sort arranges the rows; filter shows only the rows that meet a condition |
| Protection | a password stops others from opening or changing the sheet |
| Presentation | slides with text, pictures, charts, animation; F5 starts the slide show |
| Programs | LibreOffice Calc, MS Excel, Google Sheets; LibreOffice Impress, MS PowerPoint |
Service, business (trading), industrial, agricultural, technical, social
Decisiveness, initiative, interpersonal skills, perseverance, organisational skills
Takes risk, believes in self, open to ideas, sees a problem as an opportunity
| Topic | Points |
|---|---|
| Green jobs | jobs that protect or restore the environment |
| Fields | agriculture (organic farming), transport (electric vehicles), water conservation, solar and wind energy, eco-tourism, green buildings, solid waste management, appropriate technology |
| Benefits | limit greenhouse gases, reduce waste and pollution, protect ecosystems, help in adapting to climate change |
| NAPCC 2008 | National Action Plan on Climate Change, with eight national missions |
| Greenhouse gases | carbon dioxide, methane, nitrous oxide, water vapour (oxygen is not one) |
The Big Five (OCEAN) and the three clusters of personality disorders.
High: curious, creative, likes new ideas
Low: prefers routine and familiar things
High: organised, careful, finishes work in time
Low: careless, postpones work
High: outgoing, talkative, energetic
Low: quiet, prefers to be alone (introvert)
High: kind, helpful, cooperative
Low: competitive, less trusting
High: anxious, gets upset easily
Low: calm, emotionally stable
Disorders: paranoid (distrusts others), schizoid (stays alone), schizotypal (odd beliefs)
Disorders: antisocial (ignores rules), borderline (unstable moods), histrionic (seeks attention), narcissistic (feels superior)
Disorders: avoidant (fears criticism), dependent (cannot decide alone), obsessive-compulsive (wants perfect order)
Four types of sentences
| Type | What it does | Example |
|---|---|---|
| Statement (declarative) | gives information | The exam begins at ten. |
| Question (interrogative) | asks something | When does the exam begin? |
| Emotion (exclamatory) | shows a strong feeling | What a wonderful result! |
| Order (imperative) | gives a command or request | Please open your books. |
Spreadsheet functions
| Formula | Result if A1 = 20, A2 = 30, A3 = 50 |
|---|---|
=SUM(A1:A3) | 100 |
=AVERAGE(A1:A3) | 33.33 |
=MAX(A1:A3) | 50 |
=MIN(A1:A3) | 20 |
=COUNT(A1:A3) | 3 |
Tap the five stages in the right order, from first to last.
Read the personality disorder and tap its cluster.
8 marks · five one-markers, one short and one long question
Example project: a school wants to reduce the number of students who fail in Class XI.
Asks: What is the problem that you are trying to solve?
Work: Meet the people concerned and write a clear goal.
In the example: Goal: find the students at risk of failing by October, so that they can be helped.
Asks: How can you use data to answer the question?
Work: Choose descriptive, diagnostic, predictive or prescriptive analytics.
In the example: Predictive: classify each student as "at risk" or "safe".
Asks: What data do you need?
Work: List the data, its format and its sources.
In the example: Attendance, unit-test marks, homework record, of the last three years.
Asks: Where is the data coming from and how will you get it?
Work: Gather the data and look for gaps.
In the example: Export marks and attendance from the school software as CSV files.
Asks: Is the data representative of the problem?
Work: Use statistics and charts to check quality and patterns.
In the example: A chart shows that attendance of one section is missing for a month.
Asks: What work is needed to clean and shape the data?
Work: Remove errors, fill missing values, combine tables, create new features. It takes the most time.
In the example: Make a new feature "average of three unit tests" (feature engineering).
Asks: How can the data be modelled to get the answer?
Work: Build descriptive or predictive models on the training set.
In the example: Train a classification model on the data of 80% of the students.
Asks: Does the model really answer the question?
Work: Test on unseen data: diagnostic measures and statistical significance.
In the example: On the other 20%, check accuracy, precision and recall. Recall matters: no weak student should be missed.
Asks: Can you put the model into practice?
Work: Make the model available to the users.
In the example: Class teachers get a list of at-risk students every month.
Asks: Can you get constructive feedback?
Work: Collect results from real use and refine the model. The cycle repeats.
In the example: Teachers report students wrongly marked; the model is retrained.
Data science is the field of extracting useful knowledge and insights from data using statistics, programming and domain knowledge. A data science methodology is a step-by-step framework that guides a team from a business question to a working solution. The CBSE syllabus follows John Rollins' (IBM) Foundational Methodology for Data Science, which has 10 stages.
The ten stages answer ten questions:
| No. | Stage | Question answered |
|---|---|---|
| 1 | Business understanding | What is the problem we are trying to solve? |
| 2 | Analytic approach | How can we use data to answer the question? |
| 3 | Data requirements | What data do we need? |
| 4 | Data collection | Where does the data come from and how do we get it? |
| 5 | Data understanding | Is the data representative of the problem? |
| 6 | Data preparation | What extra work is needed to make the data usable? |
| 7 | Modelling | How can we visualise / model the data to get the answer? |
| 8 | Evaluation | Does the model really answer the question? |
| 9 | Deployment | Can we put the model into practice? |
| 10 | Feedback | Can we get feedback to improve it? |
A hospital wants to reduce the number of patients who are readmitted within 30 days of discharge.
| Stage | What the team does in the example |
|---|---|
| Business understanding | meets doctors and managers; goal: predict patients at high risk of readmission |
| Analytic approach | the answer is yes/no -> classification (a decision tree, for example) |
| Data requirements | age, diagnosis, previous admissions, medicines, length of stay, follow-up visits |
| Data collection | records from the hospital database; fill gaps from pharmacy and lab systems |
| Data understanding | descriptive statistics, histograms, correlation; check missing values |
| Data preparation | clean, remove duplicates, fill missing values, create features (e.g. number of visits last year) - often 70-90% of the project time |
| Modelling | train decision-tree models on training data; tune parameters |
| Evaluation | test on unseen data; check accuracy, precision, recall; discuss with doctors |
| Deployment | the model runs inside the hospital system and flags high-risk patients |
| Feedback | readmission rates are monitored; the model is retrained with new data |
what happened? summaries, dashboards
why did it happen? statistical analysis
what will happen? classification / regression
what should we do? optimisation, recommendations
| Task | Example |
|---|---|
| Univariate statistics | mean age 54, standard deviation 12 |
| Pairwise correlation | length of stay strongly correlated with readmission |
| Missing values | 8% of "blood pressure" values missing -> fill with median |
| Invalid values | age = 250 -> treat as error |
| Feature engineering | combine height and weight into BMI |
| Encoding | "Yes / No" -> 1 / 0 |
To know whether a model will work on new data, we test it on data it has not seen.
The data is divided, for example, 80% training and 20% testing. The model learns from training data; the test score estimates real performance.
Cross-validation uses all data for both training and testing and gives a more reliable estimate, especially with small datasets.
Actual sales: 10, 12, 15; predicted: 11, 12, 13. Errors: -1, 0, 2. Squared: 1, 0, 4. MSE = 5 / 3 = 1.67. RMSE = square root of 1.67 = 1.29. MAE = (1 + 0 + 2) / 3 = 1.
actual = [10, 12, 15]; pred = [11, 12, 13]
err = [a - p for a, p in zip(actual, pred)]
mse = sum(e * e for e in err) / len(err)
print("MSE", round(mse, 2), "RMSE", round(mse ** 0.5, 2), "MAE", sum(abs(e) for e in err) / len(err))
| Metric | Formula |
|---|---|
| Accuracy | (TP + TN) / total |
| Precision | TP / (TP + FP) |
| Recall | TP / (TP + FN) |
| F1 score | 2PR / (P + R) |
Business understanding, analytic approach, data requirements, data collection, data understanding, data preparation, modelling, evaluation, deployment and feedback.
The data is divided into K equal folds. The model is trained on K - 1 folds and tested on the remaining fold; this is repeated K times so that each fold is the test set once, and the K scores are averaged. It uses all the data for both training and testing and gives a more reliable, less luck-dependent estimate than one split.
The Data Science Methodology, given by John Rollins of IBM, is a set of 10 steps that takes a project from the problem to the solution and back for improvement.
| No. | Step | Question it answers |
|---|---|---|
| 1 | Business Understanding | What is the problem that you are trying to solve? |
| 2 | Analytic Approach | How can you use data to answer the question? |
| 3 | Data Requirements | What data do you need to answer the question? |
| 4 | Data Collection | Where is the data coming from and how will you get it? |
| 5 | Data Understanding | Is the data that you collected representative of the problem? |
| 6 | Data Preparation | What additional work is required to clean and shape the data? |
| 7 | Modelling | In what way can the data be visualised or modelled to get the answer? |
| 8 | Evaluation | Does the model really answer the initial question? |
| 9 | Deployment | Can you put the model into practice? |
| 10 | Feedback | Can you get constructive feedback from the users? |
The data is divided once: about 80% to train, 20% to test.
Quick, but depends on which records fall in the test set.The data is divided into k folds. The model is trained k times; each fold is the test set once.
More reliable, takes more time. 5-fold on 1,000 records: each fold has 200.Confusion matrix
| Reality: Yes | Reality: No | |
|---|---|---|
| Predicted: Yes | 40TP | 10FP |
| Predicted: No | 5FN | 45TN |
| Metric | Formula | With TP = 40, FP = 10, FN = 5, TN = 45 |
|---|---|---|
| Accuracy | (TP + TN) / Total | (40 + 45) / 100 = 85 / 100 = 0.85 (85%) |
| Precision | TP / (TP + FP) | 40 / (40 + 10) = 40 / 50 = 0.80 (80%) |
| Recall | TP / (TP + FN) | 40 / (40 + 5) = 40 / 45 = 0.89 (88.9%) |
| F1 score | 2 x Precision x Recall / (Precision + Recall) | 2 x 0.80 x 0.89 / (0.80 + 0.89) = 1.4222 / 1.69 = 0.84 |
| Case | Actual | Predicted | Error | Squared error |
|---|---|---|---|---|
| 1 | 50 | 48 | 2 | 4 |
| 2 | 60 | 64 | -4 | 16 |
| 3 | 70 | 68 | 2 | 4 |
| 4 | 80 | 84 | -4 | 16 |
MSE = sum of squared errors / number of cases = (4 + 16 + 4 + 16) / 4 = 40 / 4 = 10 RMSE = square root of MSE = 3.16 MSE can never be negative, because every error is squared. A smaller MSE or RMSE means a better model.
Four analytic approaches
| Analytic approach | Question | Example | Technique |
|---|---|---|---|
| Descriptive | What happened? | Result of last year, class-wise | summary statistics, charts |
| Diagnostic | Why did it happen? | Why did Class XI results fall? | drill-down, correlation |
| Predictive | What will happen? | Which students may fail this year? | classification, regression |
| Prescriptive | What should be done? | Which remedial plan will help most? | optimisation, recommendation |
Four model validation techniques
| Technique | How it works | Remember |
|---|---|---|
| Train-test split | the data is divided once, such as 80% to train and 20% to test | quick; the result depends on which records fall in the test set |
| K-fold cross validation | the data is cut into k folds; each fold is the test set once; the k scores are averaged | more reliable; takes k times the work |
| Leave-one-out cross validation | each single record is the test set once; the model is trained on all the others | for small data; very slow for large data |
| Time series cross validation | the model is trained on earlier periods and tested on later ones | for data in time order, such as monthly sales; the order is never shuffled |
Five-fold cross validation in a picture
The data is cut into five equal folds. In every round a different fold is the test set. The five scores are averaged.
Which metric to use
| Which metric? | Use it when | Example |
|---|---|---|
| Accuracy | the classes are balanced | cat or dog photos, half each |
| Precision | a False Positive is costly | spam filter: genuine mail must not be lost |
| Recall | a False Negative is costly | disease test: no patient should be missed |
| F1 score | both matter, or the classes are unequal | fraud detection |
| MAE, MSE, RMSE | the model predicts a number (regression) | price of a house, marks |
Three metrics for regression, on the same four errors: 2, -4, 2, -4
| Metric | How it is found | Value |
|---|---|---|
| MAE Mean Absolute Error | mean of the errors without their signs: (2 + 4 + 2 + 4) / 4 | 3 |
| MSE Mean Squared Error | mean of the squared errors: (4 + 16 + 4 + 16) / 4 | 10 |
| RMSE Root Mean Squared Error | square root of MSE | 3.16 |
Tap the steps of the Data Science Methodology from first to last.
Read the question that the data must answer and tap the approach.
6 marks · five one-markers and one short question
Example: a toll plaza camera that reads the number plate of a car.
Work: Capture the scene as a digital image or video.
Devices: digital camera, scanner, design software.
In the example: The camera takes a photo of the car as it reaches the booth.
Work: Improve the image and make it uniform.
Techniques: noise reduction, normalisation, resizing and cropping, histogram equalisation.
In the example: Night-time grain is removed, contrast is improved and the photo is cropped to the car.
Work: Find the important visual patterns.
Techniques: edge detection, corner detection, texture analysis, colour-based features.
In the example: Edges show the rectangle of the plate and the outline of each character.
Work: Find and separate the objects or regions of interest.
Tasks: classification, localisation, object detection, semantic and instance segmentation.
In the example: The plate is located and each letter and digit is separated from the background.
Work: Interpret the result and take a decision.
Tasks: recognising objects, understanding the scene, analysing the context.
In the example: The characters are read as DL 8C AB 1234 and the toll is charged to that vehicle.
Computer vision (CV) is the field of AI that enables machines to see, process, analyse and understand images and videos and take actions based on them. Humans see with eyes and understand with the brain; machines "see" with cameras and "understand" with algorithms.
A digital image is a grid of pixels. Each pixel stores numbers:
| Image type | Values per pixel | Range |
|---|---|---|
| Binary | 1 | 0 (black) or 1 (white) |
| Grayscale | 1 | 0 (black) to 255 (white) |
| Colour (RGB) | 3 (Red, Green, Blue) | each 0 to 255 |
A 1920 x 1080 colour image has 1920 x 1080 = 2,073,600 pixels. With 3 bytes per pixel it needs about 6.2 MB before compression.
| Stage | What happens | Example: number-plate reading |
|---|---|---|
| Image acquisition | capture images via cameras, scanners, satellites | toll camera captures the car |
| Pre-processing | resize, remove noise, adjust brightness and contrast, convert to grayscale | the image is cleaned and enhanced |
| Feature extraction | find edges, corners, shapes, textures, colours | plate edges and characters located |
| Detection / segmentation | find and separate the objects of interest | the plate region is cut out |
| High-level processing | recognise, classify and decide | characters read: DL 3C AB 1234; toll charged |
one label for the whole image
boxes and labels for many objects
label every pixel (semantic / instance)
identify or verify a person
read printed or handwritten text
follow objects across video frames
find body joints
Features are distinctive parts of an image: edges, corners, blobs, textures. CNNs learn them using convolution - sliding a small kernel (filter) over the image, multiplying and adding values to create a feature map.
| Field | Application |
|---|---|
| Healthcare | detecting tumours, diabetic retinopathy, fractures |
| Agriculture | crop-disease detection, weed spraying robots |
| Transport | self-driving cars, traffic monitoring, ANPR at tolls |
| Retail | cashier-less stores, shelf monitoring |
| Security | CCTV analysis, intrusion detection |
| Manufacturing | defect detection on assembly lines |
| Governance | DigiYatra face boarding at airports |
Teachable Machine (Google) lets you train an image classifier in the browser: collect images for each class, train, test and export. OpenCV is the popular open-source library for CV in Python.
Below is the same idea in plain Python on a tiny 5 x 5 image, so you can run it here. (With OpenCV on your computer the steps are cv2.imread, cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) and cv2.Canny.)
# a tiny 5 x 5 colour image: each pixel is (R, G, B)
W, B = (250, 250, 250), (20, 20, 20)
img = [[W, W, W, W, W],
[W, B, B, B, W],
[W, B, B, B, W],
[W, B, B, B, W],
[W, W, W, W, W]]
# 1. grayscale: gray = 0.299 R + 0.587 G + 0.114 B
gray = [[round(0.299 * r + 0.587 * g + 0.114 * b) for (r, g, b) in row] for row in img]
for row in gray: print(row)
# 2. vertical-edge detection with the kernel [-1, 0, 1] on each row
print("Edges:")
for row in gray:
print([abs(row[c + 1] - row[c - 1]) for c in range(1, 4)])
Image acquisition, pre-processing, feature extraction, detection/segmentation, and high-level processing (recognition and decision).
Medical imaging - detecting tumours and fractures; self-driving cars - detecting lanes and pedestrians; number-plate recognition at toll plazas; agriculture - detecting crop disease from leaf images; manufacturing - finding defects on assembly lines.
Computer vision (machine vision) lets a machine see, process and understand images and videos. An image is a grid of pixels; each pixel of an 8-bit grayscale image is a number from 0 (black) to 255 (white), and a colour pixel has three such numbers (red, green, blue).
Resolution = width x height. A 640 x 480 image has 3,07,200 pixels: 3,07,200 bytes in grayscale and 9,21,600 bytes in RGB.
| Stage | Work | Techniques or examples |
|---|---|---|
| Image acquisition | capture the image in digital form | camera, scanner, satellite, medical sensor |
| Preprocessing | improve and standardise the image | noise reduction, normalisation, resizing or cropping, histogram equalisation |
| Feature extraction | find the important visual patterns | edge detection, corner detection, texture analysis, colour-based features |
| Detection / Segmentation | find the objects or regions | classification, localisation, object detection, semantic and instance segmentation |
| High-level processing | interpret and decide | recognise a face, read a number plate, understand the scene |
Tap the squares to draw. The computer stores your picture as the numbers on the right: 1 for a white pixel, 0 for a black pixel.
Every colour pixel has three numbers: red, green and blue. Move the sliders or tap a colour.
Techniques and their stage
| Technique | Stage | What it does |
|---|---|---|
| Noise reduction | Preprocessing | removes grain and speckles |
| Normalisation | Preprocessing | brings pixel values to a common range, such as 0 to 1 |
| Resizing, cropping | Preprocessing | makes all images the same size; removes unwanted parts |
| Histogram equalisation | Preprocessing | improves the contrast of a dull image |
| Edge detection | Feature extraction | finds the boundaries where brightness changes sharply |
| Corner detection | Feature extraction | finds points where two edges meet |
| Texture analysis | Feature extraction | describes the surface pattern: smooth, rough, striped |
| Colour-based features | Feature extraction | uses the colours of an object to tell it apart |
One object or many objects
| One object | Many objects | |
|---|---|---|
| What is it? | Image classification | Object detection |
| Where is it? | Classification + localisation (one box) | Object detection (a box for each) |
| Exact shape? | - | Semantic segmentation (by class), instance segmentation (each object) |
Eight applications of computer vision
| Application | What the machine does |
|---|---|
| Facial recognition | unlocks a phone, marks attendance by face |
| Health care | finds tumours and fractures in X-ray, CT and MRI images |
| Self-driving vehicles | sees lanes, signals, vehicles and people on the road |
| Optical character recognition (OCR) | reads printed or written text from an image, such as a number plate |
| Machine inspection | finds faulty products on a factory line |
| 3D model building | builds a 3D model of an object or place from photos |
| Surveillance | watches CCTV video for unusual activity |
| Fingerprint recognition and biometrics | checks the identity of a person |
Tap the stages of the computer vision process from first to last.
Read the work and tap the stage of computer vision.
6 marks · three one-markers, one short and one long question
Example: an online shopping site during a festival sale.
Meaning: The speed at which data is produced, delivered and analysed.
In the example: Thousands of orders and payments arrive every second.
Meaning: The huge amount of data.
In the example: Crores of product views, orders and reviews are stored.
Meaning: Structured, semi-structured and unstructured data together.
In the example: Order tables, JSON logs of clicks, product photos and video reviews.
Meaning: The consistency, accuracy, quality and trustworthiness of the data.
In the example: Fake reviews and duplicate accounts must be found and removed.
Meaning: The benefit that the data gives to the organisation.
In the example: Recommendations made from the data increase the sales.
Meaning: The flow of data is not regular; it changes with time and situation, and still has to be dependable.
In the example: Traffic is ten times higher at midnight when the sale opens.
Big data refers to datasets that are so large, fast or varied that traditional tools (like a single spreadsheet or database) cannot store and process them efficiently. Social media posts, UPI transactions, sensor readings and satellite images create big data every second.
posts, likes, videos
UPI, card payments, online shopping
smart meters, wearables, vehicles
weather and earth images
records, scans, genomes
clicks and searches
| V | Meaning | Example |
|---|---|---|
| Volume | enormous amount of data | billions of UPI transactions a month |
| Velocity | speed of generation and processing | stock prices change every millisecond |
| Variety | structured, semi-structured, unstructured | tables, JSON, videos, tweets |
| Veracity | accuracy and reliability | fake reviews, sensor errors |
| Value | usefulness for decisions | predicting customer needs |
| Variability | meaning or flow changes | "sick" can mean ill or awesome |
Semi-structured data has some organisation but not a fixed table: JSON, XML, e-mails (headers + free text). About 80% of the world's data is unstructured.
Big data analytics is the process of examining large and varied data to find patterns, trends, correlations and insights for better decisions.
| Type | Question | Example |
|---|---|---|
| Descriptive | What happened? | last month's sales report |
| Diagnostic | Why did it happen? | sales fell because of a competitor's discount |
| Predictive | What will happen? | demand will rise during Diwali |
| Prescriptive | What should we do? | stock 30% more sweets and offer combos |
| Tool | Use |
|---|---|
| Hadoop (HDFS + MapReduce) | distributed storage and batch processing on clusters |
| Apache Spark | fast in-memory processing, ML (MLlib), streaming |
| NoSQL databases (MongoDB, Cassandra) | flexible storage for semi/unstructured data |
| Apache Kafka | handling real-time data streams |
| Tableau, Power BI | visualisation and dashboards |
| Python (Pandas, PySpark) | analysis and ML |
Map: each computer counts words in its part of the books - ("AI", 1), ("data", 1)... Shuffle: same words are grouped together. Reduce: counts are added - ("AI", 5,240). Many machines working in parallel finish a huge job quickly.
A data stream is a continuous, fast flow of data (sensor readings, live tweets, transactions). Data stream mining analyses it in real time, often with one pass and limited memory, using techniques like sliding windows and sampling. Uses: fraud detection during a card swipe, live traffic updates, network intrusion detection.
real-time fraud detection
recommendations, pricing
disease outbreak tracking
traffic and energy management
yield and weather prediction
personalised content (OTT)
| Benefits | Challenges |
|---|---|
| better and faster decisions | storage and processing cost |
| personalised services | data quality (veracity) |
| cost savings, efficiency | privacy and security |
| new products and innovation | shortage of skilled people |
Volume - huge size; Velocity - high speed of generation; Variety - different formats; Veracity - reliability of data. (Value and variability may also be written.)
Descriptive - what happened (monthly sales report); Diagnostic - why it happened (sales fell due to competition); Predictive - what will happen (forecasting festival demand); Prescriptive - what action to take (increase stock and offer discounts).
Big Data is data that is too large, too fast and too varied for traditional tools. Big Data analytics examines it to find patterns that help in decisions.
Fixed rows and columns.
SQL table, Excel sheet, transaction recordsTags or keys, no fixed table.
XML, JSON, CSV and HTML filesNo fixed format.
Images, videos, audio (MP3), e-mails, PDFs, social media postsThe CBSE handbook counts CSV files as semi-structured and e-mails as unstructured.
| V | Meaning | Example |
|---|---|---|
| Volume | huge amount of data | millions of posts every day |
| Velocity | speed of producing and processing | live sensor data, stock prices |
| Variety | different types of data | text, images, audio, video |
| Veracity | trustworthiness and quality | fake reviews reduce veracity |
| Value | usefulness for the organisation | offers that increase sales |
| Variability | the flow of data is not regular; it changes with time and situation | traffic that rises suddenly during a sale |
Tableau, Apache Hadoop, Cassandra, MongoDB, SAS
Apache Spark and Power BI are also used. MySQL is a database, not an analytics tool.
KB, MB, GB, TB, PB, EB, ZB
Each is 1,024 times the one before it.
Mining data streams: finding patterns and trends in data that flows continuously, in real time, as it arrives and without storing all of it first. Example: noticing a sudden rise in searches for "election results". Future of Big Data analytics: real-time analytics, more advanced models in predictive analytics, quantum computing.
Each unit is 1,024 times the one before it: 1 TB = 1,024 GB. Big Data is measured in terabytes, petabytes, exabytes and zettabytes.
Traditional data and Big Data
| Traditional data | Big Data | |
|---|---|---|
| Size | megabytes to gigabytes | terabytes, petabytes and more |
| Type | mostly structured | structured, semi-structured and unstructured |
| Speed | produced slowly, in batches | produced every second, as streams |
| Storage | one server, relational database | many computers together (distributed) |
| Tools | MySQL, spreadsheets | Hadoop, Cassandra, MongoDB, Tableau, SAS |
Working of Big Data analytics: four steps
| Step | What happens | In a hospital |
|---|---|---|
| 1. Gather data | collect structured and unstructured data from cloud storage, mobile apps, IoT sensors | patient records, lab reports, scans, wearable devices |
| 2. Process data | organise the data, by batch processing or stream processing | bring all the records of a patient together |
| 3. Clean data | correct the format; remove duplicate, irrelevant and wrong data | correct wrong entries, merge duplicate patient IDs |
| 4. Analyse data | data mining, predictive analytics, deep learning | predict which patients may need the ICU |
Two ways to process the data
| Batch processing | Stream processing | |
|---|---|---|
| How | large blocks of data are processed together, over time | small batches are processed as soon as they arrive |
| Delay | longer | very short |
| Example | monthly electricity bills | alert for a doubtful card payment |
Tap the steps of Big Data analytics from first to last.
Read the line and tap the V of Big Data that it shows.
Read the data and tap its type.
8 marks · five one-markers, one short and one long question
Example: a network that predicts the marks of a student from study hours.
Work: The input passes through the layers. Each neuron finds (inputs x weights) + bias and applies the activation function.
In the example: For 5 study hours, the network predicts 60 marks.
Work: The loss function compares the prediction with the actual value.
In the example: Actual marks are 70. Error = 10, squared error = 100.
Work: The error is sent backwards, to find how much each weight added to it.
In the example: The weight on "study hours" is found to be too small.
Work: Gradient descent changes each weight and bias by a small step (learning rate) to reduce the loss.
In the example: The weight is increased a little; next time the prediction is 64. One pass over all the data is an epoch.
An artificial neural network (ANN) is a computing model inspired by the human brain. It is made of connected units called neurons (nodes) arranged in layers. Neural networks learn patterns from data and are the foundation of deep learning.
| Component | Role |
|---|---|
| Input layer | one node per feature (e.g. pixels, age, income) |
| Hidden layers | perform calculations and learn features; many hidden layers = deep network |
| Output layer | final result: a class or a number |
| Weights (w) | strength of each connection; learned during training |
| Bias (b) | shifts the output; helps the model fit better |
| Activation function | decides the neuron's output; adds non-linearity |
Inputs x1 = 2, x2 = 3; weights w1 = 0.5, w2 = -1; bias b = 1. z = 2 x 0.5 + 3 x (-1) + 1 = 1 - 3 + 1 = -1. With a step function (output 1 if z >= 0 else 0) the output is 0. With ReLU, output = max(0, -1) = 0. With sigmoid, output = 1 / (1 + e^1) = 0.27.
0 or 1 - early perceptrons
0 to 1 - probabilities
-1 to 1 - centred output
max(0, z) - fast; most used in hidden layers
probabilities of many classes summing to 1
| Term | Meaning |
|---|---|
| Forward propagation | inputs pass layer by layer to give an output |
| Loss function | measures error (e.g. MSE) |
| Backpropagation | sends the error backward to compute how to change each weight |
| Gradient descent | changes weights a small step in the direction that reduces error |
| Learning rate | size of the step |
| Epoch | one full pass over the training data |
The perceptron (Frank Rosenblatt, 1958) is the simplest neural network: a single neuron with a step activation. It can learn linearly separable problems like AND and OR, but not XOR - which needs hidden layers.
def perceptron(x1, x2, w1=1, w2=1, b=-1.5):
z = x1 * w1 + x2 * w2 + b
return 1 if z >= 0 else 0
for a in (0, 1):
for c in (0, 1):
print(a, c, "AND ->", perceptron(a, c))
| Type | Special feature | Used for |
|---|---|---|
| Feed-forward (FFNN) | data flows one way, input to output | tabular prediction |
| Convolutional (CNN) | convolution and pooling layers | images, video |
| Recurrent (RNN / LSTM) | loops that remember earlier inputs | text, speech, time series |
| Generative adversarial (GAN) | generator vs discriminator | creating images |
| Transformers | attention to all words at once | language models (ChatGPT) |
It decides the output of a neuron and adds non-linearity so the network can learn complex patterns. Examples: ReLU, sigmoid (also tanh, softmax).
In forward propagation the inputs pass through the layers to give a prediction. The loss function measures the error. In backpropagation the error is sent backward to calculate how much each weight contributed. Gradient descent updates the weights to reduce the error. This repeats for many epochs until the error is small.
In daily life: A relay race: the first runner takes the baton (input), the middle runners carry it (hidden), the last runner finishes (output).
| Component | Meaning |
|---|---|
| Neuron | the unit that receives inputs, processes them and gives an output |
| Weight | the strength (importance) of a connection |
| Bias | an extra value added to the weighted sum |
| Activation function | decides the output of a neuron; adds non-linearity: sigmoid (0 to 1), tanh (-1 to 1), ReLU (0 for negative, same for positive) |
| Forward propagation | input moves through the layers to give the output |
| Loss function | measures the error. Regression: MSE, MAE. Classification: binary cross-entropy (log loss) |
| Backpropagation | the error is sent back to adjust the weights and bias |
| Gradient descent | changes the weights step by step to reduce the loss |
| Type | Idea | Used for |
|---|---|---|
| Standard NN (perceptron) | single layer | yes/no classification, such as spam detection |
| Feed-forward NN (FFNN) | data moves in one direction through hidden layers | image recognition, NLP, regression |
| CNN | convolution filters find the features of an image | image recognition, object detection, medical imaging |
| RNN | loops that remember earlier inputs | translation, chatbots, speech recognition, time series |
| GAN | generator against discriminator | image generation, style transfer, data augmentation |
A network with two or more hidden layers is a deep neural network.
Three activation functions
| Input x | Sigmoid | tanh | ReLU |
|---|---|---|---|
| -2 | 0.12 | -0.96 | 0 |
| -1 | 0.27 | -0.76 | 0 |
| 0 | 0.50 | 0.00 | 0 |
| 1 | 0.73 | 0.76 | 1 |
| 2 | 0.88 | 0.96 | 2 |
| Activation | Range | Where used |
|---|---|---|
| Sigmoid | 0 to 1 | output layer for yes/no (probability) |
| tanh | -1 to 1 | hidden layers; centred at zero |
| ReLU | 0 to any positive value | hidden layers of deep networks; fast |
Which loss function
| Problem | Loss function |
|---|---|
| Regression (predict a number) | Mean Squared Error (MSE), Mean Absolute Error (MAE) |
| Two-class classification | Binary cross-entropy (log loss) |
| Many-class classification | Categorical cross-entropy |
How one neuron works: five steps
One neuron, worked out
Inputs x1 = 2 x2 = 3
Weights w1 = 4 w2 = -1
Bias b = -2
Sum = (2 x 4) + (3 x -1) + (-2)
= 8 - 3 - 2
= 3
Step (threshold 4): 3 < 4, so 0
ReLU: output 3
Sigmoid: output 0.95 More words of a neural network
| Word | Meaning |
|---|---|
| Connection | the link between two neurons; it carries a weight |
| Learning rule | the method that adjusts the weights and bias; backpropagation is the common one |
| Learning rate | the size of each change in the weights; too large overshoots, too small is slow |
| Epoch | one complete pass of the whole training data through the network |
| Deep neural network | a network with two or more hidden layers |
Tap the five steps from first to last.
Read the line and tap the type of network.
7 marks · three one-markers, one short and one long question
Tap a kind of content to see the tools and how to ask.
Tools: ChatGPT, Gemini (Bard), Perplexity
Model: Large Language Model (LLM)
Sample prompt: "Explain Big Data to a Class XII student in 80 words with one example."
Tools: Canva, DALL-E, Stable Diffusion
Use: posters, logos, pictures for a project
Sample prompt: "A poster of a robot teaching in a classroom, flat colours, blue and orange."
Tools: Google Lumiere, Animaker
Use: short clips, animated explainers
Sample prompt: "A 10-second animation of the water cycle for school children."
Tools: Google MusicLM, Meta AI Voicebox
Use: background music, narration
Sample prompt: "A calm 30-second flute tune for a morning assembly."
Generative AI is AI that creates new content - text, images, audio, video, code and 3-D designs - by learning patterns from large datasets. Examples: ChatGPT, Gemini, Claude (text), DALL-E, Midjourney, Stable Diffusion (images), Suno (music), GitHub Copilot (code).
An encoder compresses data into a small latent space; a decoder rebuilds data from it. Sampling new points in the latent space creates new, similar outputs.
A transformer (2017, "Attention is all you need") uses self-attention to look at all words in a sentence at once and decide which words relate to each other. LLMs are huge transformers trained on massive text to predict the next token. GPT stands for Generative Pre-trained Transformer.
They learn to remove noise step by step; to generate, they start from random noise and gradually "denoise" it into an image matching the prompt.
| Model | Key idea | Example tools |
|---|---|---|
| GAN | two networks compete | StyleGAN faces |
| VAE | encode - decode through latent space | image variation |
| Transformer / LLM | attention, next-token prediction | ChatGPT, Gemini |
| Diffusion | noise to image | Stable Diffusion, DALL-E 3 |
A prompt is the input instruction. Prompt engineering is the skill of writing prompts that produce useful, accurate output.
"You are a Class XII biology teacher"
"Explain photosynthesis"
"for students who find it hard"
"in 5 bullet points with one example"
"under 120 words, simple English"
| Technique | Example |
|---|---|
| Zero-shot | "Translate 'Good morning' into Tamil." |
| Few-shot | give 2 examples of the format, then ask for a third |
| Chain-of-thought | "Solve step by step and show your working." |
| Iterative refinement | improve the prompt after seeing the answer |
| Risk | Meaning |
|---|---|
| Hallucination | confident but false information |
| Bias | stereotypes from training data |
| Deepfakes and misinformation | fake images, voices, videos |
| Copyright and ownership | who owns AI-made content? training on others' work |
| Privacy | sensitive data entered in prompts |
| Over-dependence | loss of one's own skills |
| Environmental cost | large energy and water use |
Riya uses an AI chatbot to brainstorm ideas for her capstone report, then reads three reliable sources, writes the report herself, cites the sources and mentions "AI used for brainstorming" in the acknowledgement. This is ethical use.
Discriminative models learn to distinguish classes (e.g. spam or not). Generative models learn the data distribution and create new data (e.g. new images or text).
Hallucination is when a generative AI model gives confident but incorrect or made-up information. Users can reduce its effect by verifying facts from reliable sources, asking for sources, giving clear context in prompts, asking the model to say when it is unsure, and keeping human review before using the output.
Generative AI creates new content (text, images, audio, video) that resembles the data it was trained on.
Learns how the data is distributed and makes new samples.
GAN, VAE, LLM, Naive BayesLearns the boundary between classes and gives a label.
Logistic regression, decision tree, SVM, random forestIn daily life: A forger makes fake notes and a bank officer checks them. Each one gets better because of the other.
| Type | Parts | Idea |
|---|---|---|
| GAN | generator, discriminator | the generator makes samples; the discriminator judges real or fake; the two compete |
| VAE | encoder, decoder | the encoder compresses data to a code (latent space); the decoder builds new data from the code |
| LLM | a very large deep learning network | trained on huge text; generates language word by word |
| Content | Tools |
|---|---|
| Text | ChatGPT, Gemini (Bard), Perplexity |
| Images | Canva, DALL-E, Stable Diffusion, Midjourney |
| Video | Google Lumiere, Animaker |
| Audio | Google MusicLM, Meta AI Voicebox |
AI-made fake video, image or audio of a real person.
Outputs repeat the bias of the training data.
Passing off AI content as your own; using others' work without credit.
People should be told when content is made by AI.
Generative and discriminative models
| Generative model | Discriminative model | |
|---|---|---|
| Aim | create new data | classify or label data |
| Learns | how the data is distributed | the boundary between classes |
| Output | new text, image, audio, video | a class label or a score |
| Used in | art and design, finding unusual patterns | face and speech recognition, spam detection |
| Examples | GAN, VAE, LLM, Naive Bayes | logistic regression, decision tree, SVM, random forest |
GAN and VAE
| GAN | VAE | |
|---|---|---|
| Full form | Generative Adversarial Network | Variational Autoencoder |
| Parts | generator and discriminator | encoder and decoder |
| Idea | two networks compete | compress to a code (latent space), then rebuild |
| Output | sharp, realistic | smoother |
| Training | harder | more stable |
Write a better prompt
| Weak prompt | Better prompt | Why |
|---|---|---|
| Write about AI | Write five points on the uses of AI in farming for a school magazine | gives the task, the length and the audience |
| Make a picture | Draw a green poster on "Save Water" with a tap and a drop, simple style | gives the subject, the colour and the style |
| Explain neural network | Explain a neural network with a daily-life example in 60 words | gives the format and the limit |
In answers on ethics, name the issue first (deepfake, bias, plagiarism, transparency), then give one example and one safeguard.
Read the task or the model and tap its kind.
Read the case and tap the issue.
5 marks · three one-markers, one short and one long question
Example: a data story on school attendance.
Work: Give the setting and introduce the characters and the question. It is also called the exposition.
In the example: The school records the attendance of 1,200 students every day. Is it the same on all days?
Work: Show the events and facts that build towards the climax.
In the example: A line graph shows a dip in attendance every Monday.
Work: Give the main insight: the most intense point, the turning point of the story.
In the example: The dip is three times bigger for students who travel more than 10 km.
Work: Show what follows from the insight, before the final outcome.
In the example: Monday unit tests are missed by these students, and their marks are lower.
Work: Resolve the problem and suggest the action. It is also called the resolution.
In the example: Shift the tests to Wednesday and start an early bus on Mondays.
Data storytelling is the skill of communicating insights from data using three elements - data, narrative and visuals - so that the audience understands, remembers and acts.
| Combination | Result |
|---|---|
| Data + narrative | explains - the audience understands |
| Data + visuals | enlightens - patterns become clear |
| Narrative + visuals | engages - the audience is interested |
| All three | drives change - the audience acts |
people remember stories more than numbers
stories make data meaningful
focuses on the key message
persuades decision makers
Numbers: "Dropout rate rose from 4% to 9%." Story: "Last year, 9 out of every 100 girls in our district left school after Class VIII - more than double the year before. The main reason in our survey: no safe transport. A bus service in 12 villages could bring most of them back."
| Part | Content |
|---|---|
| Beginning | context, why the topic matters, the question |
| Middle | data findings, comparisons, the key insight |
| End | conclusion, recommendation, what the audience should do |
| Step | Questions |
|---|---|
| Audience | Who are they? What do they know? What do they need? |
| Message | What one sentence should they remember? |
| Data | Is it accurate, complete and relevant? |
| Visuals | Which chart shows the message best? |
| Narrative | What is the beginning, middle and end? |
| Review | Can anything be removed without losing meaning? |
| Purpose | Best chart |
|---|---|
| Compare categories | bar / column chart |
| Trend over time | line chart |
| Part of a whole | pie or donut (few parts), stacked bar |
| Relationship | scatter plot |
| Distribution | histogram, box plot |
| Geography | map |
| Key single number | big-number card (KPI) |
use one strong colour for the key point
remove gridlines, 3-D, extra labels
titles that state the insight
start bar charts at zero
A chart whose y-axis starts at 95 instead of 0 can make a 2% change look huge. Always check the axes.
Excel / Google Sheets charts, Tableau, Power BI, Flourish, Datawrapper, Canva infographics and Python (Matplotlib, Seaborn, Plotly).
import matplotlib.pyplot as plt
years = [2021, 2022, 2023, 2024]
users = [20, 35, 60, 95]
plt.plot(years, users, marker="o")
plt.title("UPI users in our town grew almost 5 times in 4 years")
plt.ylabel("Users (thousands)")
plt.show()
During the pandemic, dashboards showed daily cases, recoveries and vaccinations with maps and line charts. A clear message ("cases are rising in these districts"), simple visuals and a call to action ("get vaccinated, wear masks") helped people and governments act quickly - data storytelling in real life.
Data, narrative and visuals.
Know your audience; define one key message; collect and analyse accurate data; choose suitable visuals; build a narrative with a beginning, middle and end; review and simplify; then present with a clear call to action.
Data storytelling means presenting the insights of data as a story, so that the audience understands, remembers and acts.
In daily life: A cricket commentator: the scorecard is the data, the commentary is the narrative, the replay graphics are the visuals.
| Type of data | Visualisation |
|---|---|
| Text | Word cloud |
| Numeric | Line graph, bar chart, pie chart, scatter plot, histogram, heat map |
| Stocks | Candlestick chart |
| Geographic | Map chart |
| Mixed | Facet grid |
Tap what you want to show; see the chart that fits.
Chart: Line graph
Example: temperature of a week, sales of twelve months
Chart: Bar chart
Example: marks of five students, sales of four products
Chart: Pie chart
Example: share of each house in the total points
Chart: Scatter plot
Example: study hours against marks
Chart: Histogram
Example: number of students in each range of marks
Chart: Word cloud
Example: feedback comments of parents
Chart: Map chart
Example: rainfall of the states of India
Chart: Candlestick chart
Example: daily price of a share
Chart: Heat map
Example: attendance of each class in each month
Five elements of a story
| Element | In any story | In the attendance data story |
|---|---|---|
| Characters | the people of the story | the students who travel far |
| Plot / Setting | the order of events; where and when they happen | the school, this session, Monday after Monday |
| Conflict | the problem that drives the story | attendance falls every Monday |
| Resolution | how the problem is solved | tests on Wednesday, an early bus on Monday |
| Insights | what the audience learns | distance from school affects attendance |
Data, narrative and visuals together
| Combination | Result | Meaning |
|---|---|---|
| Data + Narrative | Explain | the story tells what is happening in the data and why |
| Data + Visuals | Enlighten | the audience sees the data in a new way |
| Narrative + Visuals | Engage | the audience stays interested |
| Data + Narrative + Visuals | Change | the story moves people to act |
Data visualisation and data storytelling
| Data visualisation | Data storytelling | |
|---|---|---|
| Has | charts and graphs | charts, a narrative and the data |
| Tells | what the data looks like | what it means and what to do |
| Audience | has to find the insight | is led to the insight |
The CBSE handbook names the five stages as introduction, rising action, climax, falling action and conclusion. Use these names in the paper.
Tap the five stages of a data story from first to last.
Read what is to be shown and tap the chart.
889 questions in 15 types, including 213 numericals with full working
Tap a type to open it. In the one-mark types, tap an option to check yourself. The answers of all the numericals were worked out by a program.
In the acronym RESPECT for active listening, the letter P stands for
(b) Pay attention to what the speaker is saying P = Pay attention to what the speaker is saying (look at the speaker, avoid distractions).
Which is the first stage of active listening?
(b) Receiving Active listening starts with Receiving (hearing the message).
Which is the correct order of the stages of active listening?
(a) Receiving, Understanding, Remembering, Evaluating, Responding Receive the message, understand it, remember it, evaluate it, then respond.
In the capitalisation rule MINTS, the letter N stands for
(d) Names MINTS: Months, I, Names, Titles, Starting letter of sentences.
"The ball was hit by Rohit." This sentence is in the
(b) passive voice The receiver of the action (ball) is the subject and 'by Rohit' tells the doer, so it is passive voice.
Which part of speech describes a noun?
(c) Adjective An adjective describes a noun, e.g. 'a red ball'.
In the sentence "She quickly finished the work", the word 'quickly' is
(a) an adverb 'quickly' tells how she finished (describes the verb), so it is an adverb.
Which of these is a barrier to active listening?
(a) Thinking about your reply while the other person is speaking Planning your reply stops you from listening fully to the speaker.
A group of sentences on one idea is called a
(a) paragraph A paragraph is a group of sentences about one idea.
"Please close the door." This is
(d) an imperative sentence It is a request/command, so it is an imperative sentence.
In RESPECT, the letter C stands for
(c) Clarify your doubts C = Clarify your doubts by asking questions.
A person who does not trust others and holds grudges for a long time may have
(c) paranoid personality disorder Constant suspicion and grudges are signs of paranoid personality disorder.
Doing an activity for the joy of it, without any reward from outside, is
(b) intrinsic motivation Motivation from inside (joy, interest) is intrinsic motivation.
Continuing to work towards a goal in spite of difficulties is called
(c) perseverance Perseverance means not giving up when things are hard.
Which of these is NOT one of the Big Five personality factors?
(d) Punctuality The Big Five (OCEAN) are Openness, Conscientiousness, Extraversion, Agreeableness and Neuroticism; punctuality is not one.
A person who is talkative, outgoing and full of energy is high in
(b) extraversion Extraversion means being outgoing, social and energetic.
A person who is organised, careful and dependable is high in
(c) conscientiousness Conscientiousness means being organised and dependable.
A person who is curious, creative and likes new ideas is high in
(c) openness Openness (to experience) means curiosity and creativity.
A person who gets upset easily and often feels anxious is high in
(d) neuroticism (low emotional stability) Neuroticism (low emotional stability) means getting upset and anxious easily.
Working for a salary hike or a prize is an example of
(d) extrinsic motivation An outside reward is extrinsic motivation.
Paranoid, schizoid and schizotypal personality disorders belong to
(c) Cluster A (suspicious) Cluster A disorders involve odd or suspicious behaviour.
Antisocial, borderline, histrionic and narcissistic personality disorders belong to
(c) Cluster B (emotional and impulsive) Cluster B disorders involve dramatic, emotional or impulsive behaviour.
Avoidant, dependent and obsessive-compulsive personality disorders belong to
(c) Cluster C (anxious) Cluster C disorders involve anxious or fearful behaviour.
A person who believes he is more important than everyone else and wants constant praise may have
(c) narcissistic personality disorder An inflated sense of importance and need for praise point to narcissistic personality disorder.
A person who avoids social gatherings for fear of being criticised may have
(d) avoidant personality disorder Avoiding people for fear of criticism is a sign of avoidant personality disorder.
Focusing on the goal and finishing the work in the given time is called
(b) result orientation Result orientation means focusing on the result and the deadline.
Which of these helps a person to become result oriented?
(c) Setting clear goals and preparing an action plan Clear goals and an action plan keep the work focused on results.
Knowing your own strengths and weaknesses is called
(d) self-awareness Self-awareness means knowing yourself.
A person who depends on others for every decision may have
(c) dependent personality disorder Needing others for every decision is a sign of dependent personality disorder.
A person who is kind, cooperative and trusts others is high in
(b) agreeableness Agreeableness means being kind, cooperative and trusting.
A positive attitude helps a person to
(b) face challenges and reduce stress A positive outlook helps to face challenges and lowers stress.
Which step can help a person with a personality disorder?
(c) Talking to a counsellor or a trusted person Talking to a counsellor or trusted person is the first step to getting help.
Which is the odd one out among these spreadsheet programs?
(b) Tableau Tableau is a data visualisation tool, not a spreadsheet.
In a spreadsheet, the box where a row and a column meet is called a
(d) cell The meeting point of a row and a column is a cell.
Every formula in a spreadsheet begins with the sign
(a) = Formulas start with =, e.g. =A1+B1.
Which function adds the numbers in the cells A1 to A5?
(b) =SUM(A1:A5) =SUM(A1:A5) adds the range A1 to A5.
Which feature shows only the rows that meet a condition and hides the rest?
(a) Filter A filter shows only the matching rows.
Arranging the data of a column from A to Z is called
(a) sorting Sorting arranges data in order (A to Z or Z to A).
The address of the cell in column C and row 4 is
(d) C4 A cell address is column letter + row number: C4.
Which of these is presentation software?
(c) LibreOffice Impress LibreOffice Impress (like PowerPoint) makes presentations.
One page of a presentation is called a
(d) slide Each page of a presentation is a slide.
Which function gives the average of the cells B2 to B6?
(b) =AVERAGE(B2:B6) =AVERAGE(B2:B6) gives the mean of the range.
Giving a password to a spreadsheet so that others cannot change it is called
(d) protecting the spreadsheet Protecting the sheet with a password stops unwanted changes.
A person who starts a business with a new idea, taking a risk to earn profit, is an
(a) entrepreneur An entrepreneur takes a risk to start a business.
Entrepreneurs who buy and sell goods but do not manufacture them are
(c) business (trading) entrepreneurs Trading entrepreneurs buy and sell goods.
Entrepreneurs who manufacture products are called
(b) industrial entrepreneurs Industrial (manufacturing) entrepreneurs make products.
A person who runs a coaching centre or a salon is a
(b) service entrepreneur They sell services, not goods: service entrepreneurs.
The ability to take a decision quickly and firmly is called
(d) decisiveness Decisiveness is taking quick, firm decisions.
Starting a task on your own, without being told, shows
(d) initiative Taking the first step without being told is initiative.
Which of these is a barrier to entrepreneurship?
(a) A faulty or missing business plan A weak or missing business plan makes the business fail.
Self-doubt and fear of failure are
(a) personal barriers to entrepreneurship They come from within the person, so they are personal barriers.
Lack of raw material, labour or good roads is
(d) an environmental barrier These come from the surroundings, so they are environmental barriers.
The ability to plan, arrange resources and manage time well is called
(b) organisational skills Organisational skills are about planning and managing.
The ability to work well with others, listen and build relations is called
(d) interpersonal skills Interpersonal skills help us work with people.
Which of these is NOT a greenhouse gas?
(b) Oxygen Oxygen does not trap heat; CO₂, methane and nitrous oxide do.
Jobs that help to protect or restore the environment are called
(b) green jobs Green jobs protect or restore the environment.
NAPCC stands for
(b) National Action Plan on Climate Change NAPCC = National Action Plan on Climate Change.
The National Action Plan on Climate Change was launched in the year
(d) 2008 India launched the NAPCC in 2008.
A technician who installs solar panels has a green job in the field of
(a) renewable energy Solar power is renewable energy.
A guide who takes tourists to a forest without harming nature works in
(d) eco-tourism Tourism that protects nature is eco-tourism.
Small-scale, low-cost technology that suits local needs and does not harm the environment is called
(c) appropriate technology Appropriate technology fits local needs and the environment.
Which of these is a green job in agriculture?
(b) Organic farming Organic farming avoids chemical fertilisers and pesticides.
Which of these is a benefit of green jobs?
(c) They reduce waste and pollution Green jobs cut waste and pollution.
Which of these is a non-renewable source of energy?
(a) Natural gas Natural gas is a fossil fuel; once used it cannot be replaced.
Collecting rainwater from roofs and storing it is a green practice in
(c) water conservation Rainwater harvesting conserves water.
Which stage of active listening involves judging the message and deciding whether it is useful?
(d) Evaluating Evaluating means judging the message.
In RESPECT, the first letter R stands for
(a) Remove distractions R = Remove distractions.
Which word is a conjunction in the sentence "Ravi and Meena went home"?
(b) and 'and' joins two words, so it is a conjunction.
"What a beautiful painting!" is
(d) an exclamatory sentence It shows strong feeling and ends with !, so it is exclamatory.
Which of these shows active listening?
(b) Asking a question to clarify a doubt Asking to clarify shows you are listening and want to understand.
A person who seeks attention all the time and is overly dramatic may have
(b) histrionic personality disorder Attention-seeking and dramatic behaviour point to histrionic personality disorder.
A person who wants everything in perfect order and control may have
(a) obsessive-compulsive personality disorder Excessive need for order and control is obsessive-compulsive personality disorder.
A person who prefers to stay alone and shows little emotion may have
(c) schizoid personality disorder Liking to be alone with little emotion is schizoid personality disorder.
A person who breaks rules and does not care about the feelings of others may have
(b) antisocial personality disorder Ignoring rules and others' feelings points to antisocial personality disorder.
The Big Five personality factors are remembered by the word
(b) OCEAN Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism.
Which of these is a source of intrinsic motivation?
(c) Interest in the subject Interest in the subject comes from inside.
Which of these shows a positive attitude?
(c) Treating a failure as a chance to learn Seeing failure as a chance to learn is a positive attitude.
A result-oriented person
(a) plans the work and finishes it in the given time A result-oriented person plans and finishes the work on time.
Unstable moods and unstable relations are signs of
(b) borderline personality disorder Unstable moods and relationships are signs of borderline personality disorder.
A calm person who is not easily upset is low in
(d) neuroticism Low neuroticism = emotionally stable and calm.
Which function gives the smallest value in the cells A1 to A10?
(c) =MIN(A1:A10) =MIN(A1:A10) gives the smallest value.
Which function counts the cells that contain numbers?
(d) COUNT COUNT counts cells with numbers; COUNTA counts non-empty cells.
Which key starts the slide show from the first slide in LibreOffice Impress?
(b) F5 F5 starts the slide show from the first slide.
A pictorial form of the data of a spreadsheet is called a
(d) chart A chart shows the data as a picture.
In the formula =B2*C2, the sign * means
(d) multiplication * is multiplication in spreadsheets.
Entrepreneurs who work in farming, dairy or poultry are
(a) agricultural entrepreneurs They work in agriculture: agricultural entrepreneurs.
An entrepreneur whose main aim is to solve a problem of society is a
(c) social entrepreneur A social entrepreneur's main aim is social good.
Which competency helps an entrepreneur to lead a team and deal with customers?
(c) Interpersonal skills Interpersonal skills help in dealing with people.
Seeing a problem as an opportunity is an entrepreneurial
(c) attitude This way of thinking is an entrepreneurial attitude.
A person who takes the first step to start a new activity shows
(b) initiative Taking the first step is initiative.
Which of these is a greenhouse gas?
(b) Methane Methane traps heat in the atmosphere.
How many national missions are there in the NAPCC?
(d) 8 The NAPCC has 8 national missions (e.g. solar, energy efficiency, water).
A worker who separates dry and wet waste for recycling has a green job in
(b) solid waste management Sorting waste for recycling is solid waste management.
A driver of an electric bus has a green job in the field of
(d) transportation Clean transport reduces pollution.
Which of these is a renewable source of energy?
(c) Solar energy Solar energy is renewed naturally.
RMSE stands for
(d) Root Mean Square Error RMSE = Root Mean Square Error.
A team is deciding whether its problem needs classification or regression. In which step of the Data Science Methodology is this done?
(d) Analytic Approach Choosing the type of model is done in the Analytic Approach step.
Which activity is NOT a part of the Data Requirements step?
(a) Extracting the data from the sources Extracting the data belongs to Data Collection.
In 10-fold cross validation of 1000 records, how many records are used for training in each round?
(a) 900 One fold (100 records) is kept for testing; the other 9 folds train the model.
A model gave these results on the test data: TP = 9, FP = 6, FN = 11 and TN = 24. What is its recall?
(c) 0.45 Recall = TP / (TP + FN) = 9 / 20.
A hospital wants to know why more patients came back within a month. Which analytic approach is this?
(c) Diagnostic 'Why did it happen?' is a diagnostic question.
Which is the first step of the Data Science Methodology?
(d) Business Understanding It starts with Business Understanding: what is the problem?
In k-fold cross validation, the data is divided into
(a) k equal parts (folds) The data is split into k equal folds; each fold is used once for testing.
A model gave these results on the test data: TP = 24, FP = 24, FN = 8 and TN = 44. What is its precision?
(c) 0.50 Precision = TP / (TP + FP) = 24 / 48.
A model gave these results on the test data: TP = 9, FP = 3, FN = 9 and TN = 29. What is its recall?
(c) 0.50 Recall = TP / (TP + FN) = 9 / 18.
A model predicted the price of a bicycle (in rupees) as 224. The actual value was 200. What is the accuracy of this prediction?
(d) 88% Error = 24. Error rate = 24 / 200 = 12%. Accuracy = 100% - 12% = 88%.
A model gave these results on the test data: TP = 18, FP = 2, FN = 18 and TN = 62. What is its precision?
(a) 0.90 Precision = TP / (TP + FP) = 18 / 20.
A model predicted the price of a bicycle (in rupees) as 550. The actual value was 500. What is the accuracy of this prediction?
(a) 90% Error = 50. Error rate = 50 / 500 = 10%. Accuracy = 100% - 10% = 90%.
"Where is the data coming from and how will you get it?" This question is asked in
(a) Data Collection This is asked in Data Collection.
The errors (actual - predicted) of a regression model on 4 cases are 3, 3, 3, 3. What is the Mean Squared Error?
(d) 9 Squares: 9, 9, 9, 9. MSE = 36 / 4 = 9.
A model gave these results on the test data: TP = 24, FP = 26, FN = 8 and TN = 42. What is its recall?
(d) 0.75 Recall = TP / (TP + FN) = 24 / 32.
A model gave these results on the test data: TP = 26, FP = 14, FN = 24 and TN = 36. What is its precision?
(c) 0.65 Precision = TP / (TP + FP) = 26 / 40.
"What is the problem that you are trying to solve?" This question is asked in
(a) Business Understanding This is the question of Business Understanding.
A model gave these results on the test data: TP = 26, FP = 26, FN = 14 and TN = 34. What is its accuracy?
(b) 60% Accuracy = (TP + TN) / Total = 60 / 100.
The Mean Squared Error of a model is 25. What is its RMSE?
(d) 5 RMSE is the square root of MSE.
Removing wrong and missing values and making new useful columns is done in
(b) Data Preparation Cleaning and feature making are done in Data Preparation.
Which is a disadvantage of cross validation compared with a simple train-test split?
(a) It needs more computing time The model is trained k times, so it needs more computing time.
A lower value of RMSE means that the regression model is
(d) better Lower error means predictions are closer to actual values.
Which is the most time-consuming step of the methodology?
(c) Data Preparation Data Preparation (cleaning, transforming) usually takes most of the time.
A model gave these results on the test data: TP = 112, FP = 28, FN = 48 and TN = 12. What is its recall?
(d) 0.70 Recall = TP / (TP + FN) = 112 / 160.
A model gave these results on the test data: TP = 21, FP = 14, FN = 4 and TN = 11. What is its precision?
(a) 0.60 Precision = TP / (TP + FP) = 21 / 35.
A model gave these results on the test data: TP = 26, FP = 14, FN = 26 and TN = 34. What is its precision?
(a) 0.65 Precision = TP / (TP + FP) = 26 / 40.
The Mean Squared Error of a model is 25. What is its RMSE?
(c) 5 RMSE is the square root of MSE.
A model gave these results on the test data: TP = 36, FP = 4, FN = 44 and TN = 116. What is its accuracy?
(b) 76% Accuracy = (TP + TN) / Total = 152 / 200.
A model gave these results on the test data: TP = 12, FP = 4, FN = 3 and TN = 31. What is its recall?
(d) 0.80 Recall = TP / (TP + FN) = 12 / 15.
A model predicted the number of visitors to a fair as 540. The actual value was 500. What is the accuracy of this prediction?
(c) 92% Error = 40. Error rate = 40 / 500 = 8%. Accuracy = 100% - 8% = 92%.
A data set of 400 records is split 70:30 into training and testing data. How many records are used for testing?
(b) 120 Testing part = 30% of 400.
The Evaluation step has two phases. They are
(a) diagnostic measures phase and statistical significance phase Evaluation has a diagnostic measures phase and a statistical significance phase.
A model gave these results on the test data: TP = 15, FP = 5, FN = 10 and TN = 20. What is its accuracy?
(a) 70% Accuracy = (TP + TN) / Total = 35 / 50.
A data set of 500 records is used for 5-fold cross validation. How many records are there in each fold (test set)?
(a) 100 Each fold has 500 / 5 records.
Using domain knowledge to create new features from raw data is called
(d) feature engineering Feature engineering builds useful new features.
A model predicted the number of visitors to a fair as 360. The actual value was 400. What is the accuracy of this prediction?
(b) 90% Error = 40. Error rate = 40 / 400 = 10%. Accuracy = 100% - 10% = 90%.
A data set of 500 records is split 70:30 into training and testing data. How many records are used for testing?
(b) 150 Testing part = 30% of 500.
The value of MSE can never be
(d) negative Errors are squared, so every term is zero or positive.
Making the model available to the users is the step of
(d) Deployment Deployment puts the model into use.
A model gave these results on the test data: TP = 36, FP = 24, FN = 12 and TN = 28. What is its accuracy?
(b) 64% Accuracy = (TP + TN) / Total = 64 / 100.
A model that forecasts a future outcome, such as whether a customer will leave, is a
(b) predictive model Forecasting the future is predictive.
Which metric is used for a classification model?
(a) F1 score F1 score (with accuracy, precision, recall) is used for classification.
A model gave these results on the test data: TP = 84, FP = 16, FN = 28 and TN = 72. What is its accuracy?
(a) 78% Accuracy = (TP + TN) / Total = 156 / 200.
A model made 320 correct predictions out of 400 test cases. What is its accuracy?
(a) 80% Accuracy = correct predictions / total predictions = 320 / 400.
A model made 800 correct predictions out of 1000 test cases. What is its accuracy?
(a) 80% Accuracy = correct predictions / total predictions = 800 / 1000.
A model gave these results on the test data: TP = 18, FP = 6, FN = 22 and TN = 54. What is its precision?
(c) 0.75 Precision = TP / (TP + FP) = 18 / 24.
An analytic approach that answers "Why did it happen?" is
(c) diagnostic Diagnostic analytics finds the reasons.
An analytic approach that answers "What will happen?" is
(c) predictive Predictive analytics forecasts the future.
"What data do you need to answer the question?" This question is asked in
(c) Data Requirements This is asked in Data Requirements.
A model made 320 correct predictions out of 400 test cases. What is its accuracy?
(d) 80% Accuracy = correct predictions / total predictions = 320 / 400.
Which metric is used for a regression model?
(c) RMSE RMSE (and MSE, MAE) measures the error of number predictions.
MAE stands for
(b) Mean Absolute Error MAE = Mean Absolute Error.
A data set of 200 records is split 70:30 into training and testing data. How many records are used for testing?
(b) 60 Testing part = 30% of 200.
An analytic approach that answers "What happened?" is
(d) descriptive Descriptive analytics summarises past data.
The Mean Squared Error of a model is 9. What is its RMSE?
(d) 3 RMSE is the square root of MSE.
A model predicted the runs scored by a team as 736. The actual value was 800. What is the error rate?
(b) 8% Error = 64. Error rate = error / actual value = 64 / 800 = 8%.
A model gave these results on the test data: TP = 60, FP = 40, FN = 20 and TN = 80. What is its accuracy?
(d) 70% Accuracy = (TP + TN) / Total = 140 / 200.
A model gave these results on the test data: TP = 52, FP = 52, FN = 48 and TN = 48. What is its precision?
(b) 0.50 Precision = TP / (TP + FP) = 52 / 104.
A model predicted the weight of a parcel (in grams) as 960. The actual value was 800. What is the error rate?
(a) 20% Error = 160. Error rate = error / actual value = 160 / 800 = 20%.
A model gave these results on the test data: TP = 28, FP = 22, FN = 28 and TN = 22. What is its accuracy?
(c) 50% Accuracy = (TP + TN) / Total = 50 / 100.
A model gave these results on the test data: TP = 24, FP = 26, FN = 8 and TN = 42. What is its recall?
(b) 0.75 Recall = TP / (TP + FN) = 24 / 32.
"Is the data that you collected representative of the problem to be solved?" This question is asked in
(c) Data Understanding This is checked in Data Understanding.
A model predicted the runs scored by a team as 525. The actual value was 500. What is the error rate?
(c) 5% Error = 25. Error rate = error / actual value = 25 / 500 = 5%.
A shop wants to forecast its sales for next Diwali. Which analytic approach is this?
(a) Predictive Forecasting future sales is predictive.
A model that summarises past data to show patterns, such as clusters of customers, is a
(d) descriptive model Summarising and finding patterns in past data is descriptive.
How many steps are there in the Data Science Methodology?
(c) 10 John Rollins' methodology has 10 steps.
A model gave these results on the test data: TP = 48, FP = 16, FN = 12 and TN = 124. What is its precision?
(b) 0.75 Precision = TP / (TP + FP) = 48 / 64.
A model gave these results on the test data: TP = 54, FP = 6, FN = 18 and TN = 22. What is its recall?
(c) 0.75 Recall = TP / (TP + FN) = 54 / 72.
The errors (actual - predicted) of a regression model on 4 cases are -5, -3, -1, -1. What is the Mean Squared Error?
(a) 9 Squares: 25, 9, 1, 1. MSE = 36 / 4 = 9.
A model gave these results on the test data: TP = 18, FP = 7, FN = 12 and TN = 13. What is its recall?
(d) 0.60 Recall = TP / (TP + FN) = 18 / 30.
The errors (actual - predicted) of a regression model on 4 cases are -7, -5, -1, -5. What is the Mean Squared Error?
(a) 25 Squares: 49, 25, 1, 25. MSE = 100 / 4 = 25.
The Mean Squared Error of a model is 25. What is its RMSE?
(c) 5 RMSE is the square root of MSE.
A model gave these results on the test data: TP = 18, FP = 18, FN = 12 and TN = 52. What is its accuracy?
(b) 70% Accuracy = (TP + TN) / Total = 70 / 100.
Which is the last step of the Data Science Methodology?
(c) Feedback Feedback is the last step; it is used to improve the model.
A data set of 1200 records is used for 6-fold cross validation. How many records are there in each fold (test set)?
(d) 200 Each fold has 1200 / 6 records.
A data set of 200 records is used for 4-fold cross validation. How many records are there in each fold (test set)?
(c) 50 Each fold has 200 / 4 records.
In 3-fold cross validation of 600 records, how many records are used for training in each round?
(a) 400 One fold (200 records) is kept for testing; the other 2 folds train the model.
A model predicted the weight of a parcel (in grams) as 360. The actual value was 400. What is the error rate?
(d) 10% Error = 40. Error rate = error / actual value = 40 / 400 = 10%.
The Data Science Methodology was given by
(b) John Rollins of IBM It was given by John Rollins of IBM.
A model gave these results on the test data: TP = 44, FP = 44, FN = 36 and TN = 76. What is its precision?
(b) 0.50 Precision = TP / (TP + FP) = 44 / 88.
In 4-fold cross validation of 800 records, how many records are used for training in each round?
(a) 600 One fold (200 records) is kept for testing; the other 3 folds train the model.
A model gave these results on the test data: TP = 21, FP = 14, FN = 7 and TN = 8. What is its recall?
(a) 0.75 Recall = TP / (TP + FN) = 21 / 28.
A model gave these results on the test data: TP = 16, FP = 16, FN = 24 and TN = 44. What is its accuracy?
(a) 60% Accuracy = (TP + TN) / Total = 60 / 100.
"How can you use the data to answer the question?" This question is asked in
(d) Analytic Approach This is decided in the Analytic Approach.
An analytic approach that answers "What should be done?" is
(b) prescriptive Prescriptive analytics recommends actions.
The errors (actual - predicted) of a regression model on 4 cases are 4, -4, -4, 4. What is the Mean Squared Error?
(c) 16 Squares: 16, 16, 16, 16. MSE = 64 / 4 = 16.
In the computer vision process, which stage separates the object (for example a fruit) from the background?
(b) Detection / Segmentation Detection/segmentation separates the object from the background.
Which is the final stage of the computer vision process, where the image is interpreted and a decision is taken?
(d) High-level processing High-level processing interprets the image and takes a decision.
For an image with a single object, the two tasks are
(a) classification and localisation For one object: classification (what) and localisation (where).
Which statement about image resolution is INCORRECT?
(a) A lower resolution image shows more detail A lower resolution has fewer pixels, so it shows LESS detail.
A fake video in which a person's face is replaced using AI is a problem of
(d) duplicate and false content (deepfake) Such a deepfake is duplicate and false content.
A black-and-white (binary) image has pixel values of
(c) only 0 and 1 A binary image has only two values: 0 and 1.
A colour (RGB) image has the resolution 50 x 40. How many bytes are needed to store its pixel values without compression?
(a) 6,000 50 x 40 pixels x 3 bytes.
Which task tells both 'what' the single object is and 'where' it is?
(d) Classification + localisation Classification gives 'what' and localisation gives 'where'.
An RGB image has the resolution 640 x 480. How many values (numbers) are stored for the whole image?
(c) 921,600 Each of the 307,200 pixels stores 3 values: red, green and blue.
Capturing an image with a camera, scanner or sensor is called
(c) image acquisition Getting the image is image acquisition.
Finding the boundaries of objects in an image, where the brightness changes sharply, is
(c) edge detection Edges are where brightness changes sharply: edge detection.
In an RGB image the pixel (0, 0, 255) is
(d) blue Only blue at full brightness: blue.
Drawing a box around the single object found in an image is called
(d) localisation Marking where the object is: localisation.
An 8-bit grayscale image has the resolution 28 x 28. How many bytes are needed to store its pixel values?
(c) 784 One byte for each pixel: 28 x 28.
The range of pixel values in an 8-bit image is
(d) 0 to 255 8 bits store 2⁸ = 256 values: 0 to 255.
How many pixels are there in an image of resolution 10 x 5?
(d) 50 Pixels = width x height.
Computer vision is also known as
(d) machine vision Computer vision is also called machine vision.
An 8-bit grayscale image has the resolution 4 x 3. How many bytes are needed to store its pixel values?
(d) 12 One byte for each pixel: 4 x 3.
Which of these is NOT an application of computer vision?
(c) Chatbot answering typed questions A chatbot works with text, so it is NLP.
An X-ray machine producing a digital image is an example of
(c) image acquisition Capturing the image is image acquisition.
Which is the first stage of the computer vision process?
(b) Image acquisition First the image is acquired.
Cutting out the unwanted outer part of an image is called
(b) cropping Cropping removes the outer part.
Checking fruits on a conveyor belt for defects is an application of computer vision in
(a) quality inspection in industry Finding defects in products is quality inspection.
How many pixels are there in an image of resolution 28 x 28?
(a) 784 Pixels = width x height.
Analysing the pattern of a surface, such as smooth or rough, is
(b) texture analysis Texture analysis studies surface patterns.
What is the total number of pixels in an image that is 300 pixels wide and 200 pixels high?
(d) 60,000 Width x height.
A camera at a mall counting the number of visitors uses
(c) object detection Counting people needs finding each one: object detection.
In an RGB image, the pixel (255, 255, 0) is
(c) yellow Red + green at full brightness make yellow.
A self-driving car uses computer vision mainly to
(c) detect vehicles, people and traffic signs It must see vehicles, people and signs.
What is the total number of pixels in an image that is 200 pixels wide and 100 pixels high?
(d) 20,000 Width x height.
An RGB image has the resolution 4 x 3. How many values (numbers) are stored for the whole image?
(a) 36 Each of the 12 pixels stores 3 values: red, green and blue.
How many channels does an RGB image have?
(d) 3 Three channels: R, G and B.
Which of these is a challenge of computer vision?
(d) Privacy and security of people in the images Cameras can capture people without consent: privacy and security.
Labelling every pixel of an image with the class it belongs to, without separating two objects of the same class, is
(d) semantic segmentation Semantic segmentation labels pixels by class only.
Face unlock in a mobile phone is an application of
(d) computer vision Recognising a face in a camera image is computer vision.
Which stage of the computer vision process comes just after preprocessing?
(d) Feature extraction After cleaning (preprocessing), features are extracted.
How many pixels are there in an image of resolution 10 x 5?
(c) 50 Pixels = width x height.
A colour (RGB) image has the resolution 1024 x 768. How many bytes are needed to store its pixel values without compression?
(b) 2,359,296 1024 x 768 pixels x 3 bytes.
Counting the cars on a road from a traffic camera needs
(a) object detection Each car must be found: object detection.
In an 8-bit grayscale image, the pixel value 0 stands for
(d) black 0 is black.
What is the total number of pixels in an image that is 500 pixels wide and 400 pixels high?
(b) 200,000 Width x height.
An 8-bit grayscale image has the resolution 64 x 64. How many bytes are needed to store its pixel values?
(b) 4,096 One byte for each pixel: 64 x 64.
Which is the correct order of the computer vision process?
(b) Image acquisition, Preprocessing, Feature extraction, Detection/Segmentation, High-level processing Acquire → preprocess → extract features → detect/segment → high-level processing.
Which of these is expected from computer vision in the future?
(c) Personalised health-care diagnosis and immersive augmented reality Future uses include personalised diagnosis and immersive AR.
Finding many objects in an image and drawing a box around each is
(b) object detection Many objects with boxes: object detection.
How many pixels are there in an image of resolution 8 x 8?
(d) 64 Pixels = width x height.
In a grayscale image, the pixel value 255 stands for
(a) white 255 is white.
Which of these is an application of computer vision in health care?
(d) Detecting tumours in X-ray and MRI images Finding tumours in scans uses computer vision.
Which library is commonly used in Python for computer vision?
(b) OpenCV OpenCV (cv2) is the common computer vision library.
What is the total number of pixels in an image that is 800 pixels wide and 600 pixels high?
(d) 480,000 Width x height.
Which technique improves the contrast of a dull image?
(a) Histogram equalisation Histogram equalisation spreads out the brightness values to improve contrast.
Edge detection and corner detection are techniques of
(a) feature extraction They find important features, so they are feature extraction.
Changing the pixel values of all images to a common range, such as 0 to 1, is called
(a) normalisation Normalisation rescales values to a common range.
A colour (RGB) image has the resolution 1280 x 720. How many bytes are needed to store its pixel values without compression?
(b) 2,764,800 1280 x 720 pixels x 3 bytes.
Noise reduction, normalisation and resizing are techniques of
(d) preprocessing They clean and prepare images: preprocessing.
Giving one label to the whole image is called
(a) image classification One label for the whole image: image classification.
The quality of a camera image is poor in fog and at night. This is a challenge of
(c) image acquisition Poor light and weather affect image acquisition.
Which stage comes just before high-level processing?
(c) Detection / Segmentation Detection/segmentation comes before high-level processing.
A colour (RGB) image has the resolution 12 x 12. How many bytes are needed to store its pixel values without compression?
(d) 432 12 x 12 pixels x 3 bytes.
An RGB image has the resolution 28 x 28. How many values (numbers) are stored for the whole image?
(a) 2,352 Each of the 784 pixels stores 3 values: red, green and blue.
Marking each separate object, even of the same class, with its own outline is
(b) instance segmentation Instance segmentation separates every object.
Removing unwanted grain and speckles from an image is called
(b) noise reduction Grain and speckles are noise; removing them is noise reduction.
The smallest unit of a digital image is the
(d) pixel The pixel is the smallest unit.
Making all the images the same size before giving them to a model is called
(d) resizing Resizing makes the images the same size.
Which of these is NOT a data analytics tool?
(c) MySQL MySQL is a database management system.
Which 'V' of Big Data is about the consistency and trustworthiness of the data?
(b) Veracity Veracity is about how accurate and trustworthy data is.
Which law says that computing power keeps increasing with time, which makes it easier to handle large volumes of data?
(c) Moore's Law Moore's Law: transistors on a chip (computing power) double about every two years.
A sensor produces 2 MB of data every second. How much data does it produce in one minute?
(d) 120 MB 2 x 60.
Which unit comes just after the petabyte (PB)?
(b) Exabyte (EB) KB < MB < GB < TB < PB < EB: the exabyte comes after PB.
Getting insights from data at the moment it arrives, so that decisions are immediate, is called
(c) real-time analytics Real-time analytics works on data as it arrives.
Which tool makes interactive dashboards and charts from data?
(b) Tableau Tableau builds interactive dashboards.
A CCTV camera stores 4 GB of video every day. How much does it store in 30 days?
(a) 120 GB 4 x 30.
1 GB is equal to how many MB?
(d) 1,024 Each bigger unit is 1,024 times the smaller one.
Which of these is an advantage of Big Data?
(b) Better decision making More data gives better decisions.
Data stored in rows and columns, such as an SQL table, is
(b) structured data Rows and columns with a fixed format is structured data.
Changes in the meaning or flow of data with time is the V called
(b) Variability Variability is the change in meaning or flow of data.
1 PB is equal to how many GB?
(a) 1,048,576 1 PB = 1,024 TB and 1 TB = 1,024 GB.
Readings sent every second by the sensors of a smart watch are an example of
(c) a data stream Continuous readings form a data stream.
A feedback form with name, age and rating in fixed columns is
(b) structured data Fixed fields make it structured data.
How many gigabytes are there in 3 TB?
(d) 3,072 3 x 1,024.
Which is the first step in the working of Big Data analytics?
(a) Gather data First gather the data.
Customer details and transaction records kept in the tables of a database are
(c) structured data Database tables hold structured data.
The 3 Vs of Big Data are
(a) Volume, Velocity and Variety Volume, Velocity and Variety.
1 MB is equal to how many KB?
(d) 1,024 Each bigger unit is 1,024 times the smaller one.
Which 'V' is shown when a bank processes thousands of card payments every second?
(d) Velocity Speed of data is Velocity.
Which of these is the largest unit of data?
(b) Zettabyte KB, MB, GB, TB, PB, EB, ZB: each is 1,024 times the one before.
1 PB is equal to how many TB?
(d) 1,024 Each bigger unit is 1,024 times the smaller one.
How many megabytes are there in 5 GB?
(b) 5,120 5 x 1,024.
Which of these is a NoSQL database used for Big Data?
(d) MongoDB MongoDB is a NoSQL (document) database.
Data that arrives continuously and at high speed, such as sensor readings, is called a
(b) data stream A continuous flow of data is a data stream.
Removing errors and duplicates from the data is done in the step
(a) Clean data Cleaning removes errors and duplicates.
Data that is too large and complex for traditional tools to handle is called
(b) Big Data That is Big Data.
An app gets 500 new records every second. How many records does it get in one minute?
(a) 30,000 500 x 60.
Which 'V' is weak when a data set has many fake and duplicate entries?
(d) Veracity Fake and duplicate entries make data less trustworthy: Veracity.
Images, videos and audio files are examples of
(c) unstructured data They have no fixed rows and columns: unstructured data.
XML and JSON files are examples of
(c) semi-structured data They have tags/keys but no fixed table: semi-structured data.
The huge amount of data produced every second is
(b) Volume Volume is the amount of data.
1 TB is equal to how many GB?
(d) 1,024 Each bigger unit is 1,024 times the smaller one.
Posts, photos and videos on social media are mostly
(b) unstructured data Free-form posts and media are unstructured.
The speed at which data is produced and processed is
(a) Velocity Velocity is the speed of data.
Which is the correct order from the smallest to the largest?
(a) GB, TB, PB, EB GB < TB < PB < EB, each 1024 times the previous.
Data mining, predictive analytics and deep learning are used in the step
(a) Analyse data They find insights, so they are used in Analyse data.
In which step of Big Data analytics is the data organised so that queries give correct results?
(b) Process data Process data organises and structures the data.
Extracting useful information from a continuous flow of data is called
(b) mining data streams Mining data streams finds information in flowing data.
The usefulness of data for a business is
(a) Value Value is the usefulness of data.
Data coming as text, images, audio and video shows
(c) Variety Many forms of data show Variety.
Which is the correct order of the working of Big Data analytics?
(d) Gather, Process, Clean, Analyse Gather → Process → Clean → Analyse.
Which of these is a tool for Big Data?
(b) Apache Hadoop Apache Hadoop stores and processes big data across many computers.
Which unit comes just after the terabyte (TB)?
(b) Petabyte (PB) After TB comes the petabyte (PB).
Which of these is a disadvantage of Big Data?
(d) Risk to privacy and security Large personal data can be leaked or misused: privacy risk.
A neural network with two or more hidden layers is called a
(d) deep neural network Two or more hidden layers make it deep.
Another name for a perceptron is
(d) standard neural network A perceptron is the simplest (standard) neural network.
Which loss function is NOT used for regression?
(a) Log loss (binary cross-entropy) Log loss (cross-entropy) is for classification; MSE and MAE are for regression.
Which of these is an activation function?
(c) ReLU ReLU (also sigmoid, tanh) is an activation function.
In the formula output = (w x input) + b, the terms w and b are called
(c) weight and bias w is the weight and b is the bias.
Which type of neural network is best for images?
(a) Convolutional Neural Network (CNN) CNNs use convolution to find image features.
A neuron has the inputs 3 and 3 with the weights 1 and -2, and the bias -3. What is its output with the ReLU activation function?
(a) 0 Weighted sum = (3 x 1) + (3 x -2) + (-3) = -6. ReLU gives max(0, -6) = 0.
Weighted sum of a neuron =
(c) sum of (input x weight) + bias Each input is multiplied by its weight, added up, and the bias is added.
The output of the sigmoid function lies between
(a) 0 and 1 Sigmoid squeezes any value into 0 to 1.
The activation function is needed because it
(b) adds non-linearity, so the network can learn complex patterns Without it the network would be only linear and could not learn complex patterns.
What is the output of the ReLU activation function for the input -2?
(c) 0 ReLU(x) = max(0, x): a negative input gives 0, a positive input comes out unchanged.
Passing the input through the layers to get the output is called
(c) forward propagation Moving forward from input to output is forward propagation.
A neuron has the inputs x1 = 2 with weight w1 = -1, x2 = 3 with weight w2 = 1, x3 = 2 with weight w3 = 1. The bias is -1. What is the weighted sum (with bias)?
(b) 2 Weighted sum = (2 x -1) + (3 x 1) + (2 x 1) + (-1) = 2.
A neuron has the inputs x1 = 3 with weight w1 = -1, x2 = 0 with weight w2 = 1. The bias is -1. What is the weighted sum (with bias)?
(c) -4 Weighted sum = (3 x -1) + (0 x 1) + (-1) = -4.
A neuron has the inputs x1 = 5 with weight w1 = -2, x2 = 1 with weight w2 = -1, x3 = 2 with weight w3 = 4. The bias is -2. What is the weighted sum (with bias)?
(b) -5 Weighted sum = (5 x -2) + (1 x -1) + (2 x 4) + (-2) = -5.
What is the output of the ReLU activation function for the input 7?
(a) 7 ReLU(x) = max(0, x): a negative input gives 0, a positive input comes out unchanged.
The layers between the input layer and the output layer are called
(a) hidden layers They are hidden layers.
Which Python library is used to build and train neural networks?
(a) TensorFlow TensorFlow (with Keras) builds and trains neural networks.
A neuron has the inputs x1 = 2 with weight w1 = 2, x2 = 2 with weight w2 = 3. The bias is 1. What is the weighted sum (with bias)?
(b) 11 Weighted sum = (2 x 2) + (2 x 3) + (1) = 11.
What is the output of the ReLU activation function for the input -5?
(d) 0 ReLU(x) = max(0, x): a negative input gives 0, a positive input comes out unchanged.
A GAN has two networks. They are the
(d) generator and discriminator A GAN has a generator and a discriminator.
What is the output of the ReLU activation function for the input 12?
(a) 12 ReLU(x) = max(0, x): a negative input gives 0, a positive input comes out unchanged.
Which activation function gives values between 0 and 1 and is used for probability?
(d) Sigmoid Sigmoid output (0 to 1) can be read as a probability.
A neuron has the inputs x1 = 0 with weight w1 = 4, x2 = 4 with weight w2 = 2. The bias is -1. What is the weighted sum (with bias)?
(b) 7 Weighted sum = (0 x 4) + (4 x 2) + (-1) = 7.
Which type of neural network is best for sequence data such as text and speech?
(b) Recurrent Neural Network (RNN) RNNs remember earlier inputs, so they suit sequences.
A neuron has the inputs x1 = 1 with weight w1 = 3, x2 = 0 with weight w2 = -1, x3 = 0 with weight w3 = 1. The bias is 1. What is the weighted sum (with bias)?
(c) 4 Weighted sum = (1 x 3) + (0 x -1) + (0 x 1) + (1) = 4.
An input layer of 6 neurons is fully connected to a hidden layer of 3 neurons. How many weights are there between the two layers?
(b) 18 Every input neuron is joined to every hidden neuron: 6 x 3.
Which type of network is used in voice assistants to process speech?
(a) Recurrent Neural Network Speech is a sequence, so RNNs are used.
A neuron has the inputs x1 = 1 with weight w1 = 4, x2 = 3 with weight w2 = 1, x3 = 1 with weight w3 = 4. The bias is -2. What is the weighted sum (with bias)?
(c) 9 Weighted sum = (1 x 4) + (3 x 1) + (1 x 4) + (-2) = 9.
Which layer gives the final prediction?
(c) Output layer The output layer gives the result.
A neuron has the inputs x1 = 2 with weight w1 = -2, x2 = 3 with weight w2 = 3 and bias -2. It gives the output 1 if the weighted sum is 5 or more, otherwise 0. What is the output?
(b) 0 Weighted sum = (2 x -2) + (3 x 3) + (-2) = 3. It is less than the threshold 5.
A learning rate that is too large makes the training
(a) unstable: the loss may jump and not settle Too large steps overshoot, so the loss jumps around.
Which tool lets you see a neural network learn in the browser?
(c) TensorFlow Playground TensorFlow Playground shows a network learning live in the browser.
A neuron has the inputs x1 = 0 with weight w1 = 3, x2 = 1 with weight w2 = 1. The bias is 1. What is the weighted sum (with bias)?
(b) 2 Weighted sum = (0 x 3) + (1 x 1) + (1) = 2.
A neuron has the inputs x1 = 5 with weight w1 = 2, x2 = 4 with weight w2 = 4 and bias 3. It gives the output 1 if the weighted sum is 30 or more, otherwise 0. What is the output?
(a) 0 Weighted sum = (5 x 2) + (4 x 4) + (3) = 29. It is less than the threshold 30.
The ReLU function gives the output
(b) 0 for a negative input and the input itself for a positive input ReLU(x) = max(0, x): 0 for negatives, x for positives.
The function that measures the difference between the predicted and the actual output is the
(b) loss function The loss function measures the error.
A network in which data moves only in one direction, from input to output, is a
(b) feed-forward neural network One-way flow: feed-forward neural network.
An input layer of 2 neurons is fully connected to a hidden layer of 3 neurons. How many weights are there between the two layers?
(b) 6 Every input neuron is joined to every hidden neuron: 2 x 3.
Which loss function is used for a two-class classification problem?
(a) Binary cross-entropy (log loss) Binary cross-entropy (log loss) is used for two classes.
A neuron has the inputs 2 and 4 with the weights 1 and -1, and the bias -5. What is its output with the ReLU activation function?
(d) 0 Weighted sum = (2 x 1) + (4 x -1) + (-5) = -7. ReLU gives max(0, -7) = 0.
The method used to reduce the loss by changing the weights step by step is
(b) gradient descent Gradient descent moves the weights downhill on the loss.
One complete pass of the whole training data through the network is called an
(a) epoch One full pass is an epoch.
A neuron has the inputs 1 and 3 with the weights 2 and 1, and the bias -5. What is its output with the ReLU activation function?
(b) 0 Weighted sum = (1 x 2) + (3 x 1) + (-5) = 0. ReLU gives max(0, 0) = 0.
An input layer of 4 neurons is fully connected to a hidden layer of 4 neurons. How many weights are there between the two layers?
(a) 16 Every input neuron is joined to every hidden neuron: 4 x 4.
The basic unit of a neural network is the
(d) neuron (node) The neuron (node) is the basic unit.
Sending the error backwards to adjust the weights is called
(b) backpropagation Backpropagation sends the error back to update the weights.
What is the output of the ReLU activation function for the input -9?
(c) 0 ReLU(x) = max(0, x): a negative input gives 0, a positive input comes out unchanged.
A neuron has the inputs x1 = 3 with weight w1 = -2, x2 = 0 with weight w2 = -1, x3 = 5 with weight w3 = -2 and bias 1. It gives the output 1 if the weighted sum is -16 or more, otherwise 0. What is the output?
(d) 1 Weighted sum = (3 x -2) + (0 x -1) + (5 x -2) + (1) = -15. It is not less than the threshold -16.
A neuron has the inputs 3 and 2 with the weights 1 and -1, and the bias 2. What is its output with the ReLU activation function?
(a) 3 Weighted sum = (3 x 1) + (2 x -1) + (2) = 3. ReLU gives max(0, 3) = 3.
A neuron has the inputs 3 and 1 with the weights 3 and 2, and the bias -5. What is its output with the ReLU activation function?
(a) 6 Weighted sum = (3 x 3) + (1 x 2) + (-5) = 6. ReLU gives max(0, 6) = 6.
A neuron has the inputs 3 and 3 with the weights -1 and -1, and the bias -3. What is its output with the ReLU activation function?
(a) 0 Weighted sum = (3 x -1) + (3 x -1) + (-3) = -9. ReLU gives max(0, -9) = 0.
A neuron has the inputs 4 and 2 with the weights -1 and 2, and the bias -3. What is its output with the ReLU activation function?
(b) 0 Weighted sum = (4 x -1) + (2 x 2) + (-3) = -3. ReLU gives max(0, -3) = 0.
Which layer of a neural network receives the data?
(b) Input layer The input layer receives the data.
The number that shows the strength of a connection between two neurons is the
(b) weight The weight shows the strength of a connection.
During training, which values of a neural network are adjusted?
(a) Weights and biases Training adjusts the weights and biases.
What is the output of the ReLU activation function for the input 0?
(b) 0 ReLU(x) = max(0, x): a negative input gives 0, a positive input comes out unchanged.
A neuron has the inputs 1 and 3 with the weights -3 and 1, and the bias -1. What is its output with the ReLU activation function?
(d) 0 Weighted sum = (1 x -3) + (3 x 1) + (-1) = -1. ReLU gives max(0, -1) = 0.
A neuron has the inputs 1 and 4 with the weights -3 and -2, and the bias 1. What is its output with the ReLU activation function?
(c) 0 Weighted sum = (1 x -3) + (4 x -2) + (1) = -10. ReLU gives max(0, -10) = 0.
An input layer of 4 neurons is fully connected to a hidden layer of 5 neurons. How many weights are there between the two layers?
(c) 20 Every input neuron is joined to every hidden neuron: 4 x 5.
The output of the tanh function lies between
(c) -1 and 1 tanh gives values from -1 to 1.
A neuron has the inputs 3 and 2 with the weights -2 and -2, and the bias 2. What is its output with the ReLU activation function?
(d) 0 Weighted sum = (3 x -2) + (2 x -2) + (2) = -8. ReLU gives max(0, -8) = 0.
An input layer of 5 neurons is fully connected to a hidden layer of 2 neurons. How many weights are there between the two layers?
(c) 10 Every input neuron is joined to every hidden neuron: 5 x 2.
A neuron has the inputs x1 = 0 with weight w1 = -2, x2 = 3 with weight w2 = 4. The bias is -3. What is the weighted sum (with bias)?
(b) 9 Weighted sum = (0 x -2) + (3 x 4) + (-3) = 9.
An input layer of 3 neurons is fully connected to a hidden layer of 4 neurons. How many weights are there between the two layers?
(c) 12 Every input neuron is joined to every hidden neuron: 3 x 4.
A neuron has the inputs x1 = 5 with weight w1 = 3, x2 = 4 with weight w2 = 3, x3 = 4 with weight w3 = 4. The bias is 3. What is the weighted sum (with bias)?
(d) 46 Weighted sum = (5 x 3) + (4 x 3) + (4 x 4) + (3) = 46.
Which type of network has connections that loop back, giving it memory?
(b) Recurrent Neural Network RNNs have loops that keep earlier information.
A neuron has the inputs x1 = 0 with weight w1 = 4, x2 = 4 with weight w2 = 1, x3 = 2 with weight w3 = -1 and bias -1. It gives the output 1 if the weighted sum is 2 or more, otherwise 0. What is the output?
(d) 0 Weighted sum = (0 x 4) + (4 x 1) + (2 x -1) + (-1) = 1. It is less than the threshold 2.
What is the main aim of generative AI?
(a) To create new data that resembles the data it was trained on Generative AI creates new data like its training data.
Which of these tools is NOT used to generate images?
(a) Google Lumiere Lumiere generates videos.
A student submits an essay written by an AI tool as his own work. This is
(a) plagiarism Presenting AI-written work as your own is plagiarism.
A model that learns the boundary between classes, such as 'cat' and 'dog', is a
(c) discriminative model A discriminative model separates classes.
A model that learns how the data is distributed and can produce new samples is a
(b) generative model A generative model can create new samples.
LLM stands for
(d) Large Language Model LLM = Large Language Model.
Which of these is a Large Language Model?
(c) GPT GPT is a large language model.
In a GAN, the network that creates fake samples is the
(c) generator The generator creates fake samples.
In a GAN, the network that checks whether a sample is real or fake is the
(c) discriminator The discriminator judges real or fake.
A VAE has two parts. They are the
(c) encoder and decoder A VAE has an encoder and a decoder.
VAE stands for
(a) Variational Autoencoder VAE = Variational Autoencoder.
A realistic fake video or audio of a person made with AI is called a
(c) deepfake That is a deepfake.
Writing a clear instruction to get the desired output from a generative AI tool is called
(b) prompt engineering Prompt engineering means writing good prompts.
Which tool generates text answers to prompts?
(d) ChatGPT ChatGPT answers prompts with text.
When an LLM gives a confident but wrong or made-up answer, it is called
(b) hallucination Made-up confident answers are hallucinations.
Which of these is an ethical concern of generative AI?
(c) Bias in the generated content Generated content can be biased.
Which of these is a limitation of LLMs?
(a) They need very high processing power LLMs need very high computing power.
Which tool can generate a presentation design or image from a text prompt?
(c) Canva Canva has AI features that design from a text prompt.
Which of these is an application of generative AI in audio?
(b) Creating music or a voice from text Creating music or a voice from text is generative audio.
Who owns the content created by AI? This question is about
(b) copyright and ownership Ownership of AI content is a copyright question.
Google's large language model that can be used through an API is
(b) Gemini Gemini is Google's LLM.
Which of these is a responsible use of generative AI?
(b) Checking the facts and telling that AI was used Check facts and say that AI was used.
Which of these is a discriminative task?
(b) Classifying an e-mail as spam or not spam Spam or not spam is classification: discriminative.
Which of these is a generative task?
(b) Creating a new image of a landscape Creating a new image is generative.
In a VAE, the part that compresses the input into a small code is the
(b) encoder The encoder compresses the input.
In a VAE, the part that produces data from the code is the
(c) decoder The decoder rebuilds data from the code.
Which tool generates images from a text prompt?
(d) DALL-E DALL-E makes images from text.
LLMs are trained mainly on
(c) huge amounts of text LLMs learn from huge amounts of text.
Which is the best prompt?
(d) Write a 100-word note on Big Data for Class XII with one example A good prompt is clear and specific.
An AI system that gives lower loan scores to one group because of its training data shows
(c) bias Unfair results from data show bias.
Which of these is a text-to-video tool?
(c) Google Lumiere Google Lumiere generates video from text.
Using generative AI to copy a singer's voice without permission raises the issue of
(a) consent and copyright It needs the singer's consent and respects copyright.
Which narrative structure is used to present a data story with an introduction, rising action, climax, falling action and conclusion?
(a) Freytag's Pyramid Freytag's Pyramid has these five parts.
Which visualisation uses colour to compare categories?
(b) Heat map A heat map shows values with colours.
Which statement about stories is NOT true?
(a) Stories are harder to remember than plain numbers Stories are EASIER to remember than plain numbers.
The three key elements of data storytelling are
(c) data, narrative and visuals Data, narrative and visuals.
Data together with a narrative helps to
(a) explain what is happening in the data A narrative explains what the data shows.
Data together with visuals helps to
(b) enlighten the audience by showing patterns they would not see otherwise Visuals enlighten: they show patterns hidden in numbers.
A narrative together with visuals helps to
(a) engage the audience A story with visuals engages the audience.
When data, narrative and visuals are combined well, the data story can
(d) influence and drive change All three together influence decisions and drive change.
Which is the first stage of Freytag's Pyramid?
(b) Introduction It begins with the Introduction (exposition).
The most important finding of a data story is placed at the
(c) climax The climax carries the key insight.
Which is the last stage of Freytag's Pyramid?
(c) Conclusion It ends with the Conclusion (resolution).
Which visualisation is best for text data, to show the most frequent words?
(b) Word cloud A word cloud shows frequent words larger.
Which visualisation is best to show a trend of numeric data over time?
(d) Line graph A line graph shows trends over time.
Which visualisation is used for stock prices?
(a) Candlestick chart Candlestick charts show open, high, low and close prices.
Which visualisation is best for data of different places?
(c) Map A map shows data by location.
Showing data in the form of charts and graphs is called
(d) data visualisation That is data visualisation.
Which is the first step in creating a data story?
(d) Collect the data and organise it First collect and organise the data.
Changing the scale of a graph to mislead the audience is
(b) unethical Misleading graphs are unethical.
A data story is more effective than a table of numbers because
(d) people remember stories better than numbers People remember stories better than numbers.
Before making a data story, we must first know
(a) who the audience is Know your audience first.
Which chart shows many small charts of the same kind for different groups (mixed data)?
(c) Facet grid A facet grid shows small multiples for each group.
In Freytag's Pyramid, the part that builds interest by showing the problem in the data is the
(c) rising action Rising action builds interest by showing the problem.
In a data story, the 'conflict' is
(b) the problem shown by the data The conflict is the problem the data shows.
The stage of Freytag's Pyramid that comes just after the climax is
(d) falling action After the climax comes the falling action.
Which element of data storytelling gives the facts?
(a) Data Data gives the facts.
Which element of data storytelling gives the storyline and meaning?
(c) Narrative The narrative gives meaning.
Which chart is best to show the share of each category in a whole?
(a) Pie chart A pie chart shows parts of a whole.
Which chart is best to compare the sales of five products?
(c) Bar chart A bar chart compares categories.
In a word cloud, a bigger word means that the word
(a) occurs more often Bigger words occur more often.
Showing only the data that supports your view and hiding the rest is
(c) unethical (cherry-picking) Cherry-picking data is unethical.
The setting and the characters of a data story are given in the
(a) introduction The introduction sets the scene.
A data story should end with
(b) a clear conclusion or call to action End with a clear conclusion or call to action.
Statement 1: The Data Science Methodology has ten steps.
Statement 2: Feedback is its last step.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Business Understanding is the first step of the methodology.
Statement 2: Deployment comes before Modelling.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: MSE can be negative.
Statement 2: RMSE is the square root of MSE.
(d) Statement 1 is false and Statement 2 is true. Statement 1 is false; Statement 2 is true.
Statement 1: Feature engineering uses domain knowledge to create new features from raw data.
Statement 2: Feature engineering is done in the Feedback step.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Precision and recall are metrics for regression.
Statement 2: RMSE is a metric for classification.
(b) Both the statements are false. Statement 1 is false; Statement 2 is false.
Statement 1: In k-fold cross validation, every fold is used once as the test set.
Statement 2: Cross validation takes more computing time than one train-test split.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Image acquisition is the first stage of the computer vision process.
Statement 2: High-level processing is the last stage.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Noise reduction is a preprocessing technique.
Statement 2: Edge detection is a preprocessing technique.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: A higher resolution image has fewer pixels.
Statement 2: A grayscale image has three channels.
(b) Both the statements are false. Statement 1 is false; Statement 2 is false.
Statement 1: Object detection can find many objects in one image.
Statement 2: Image classification gives a box for every object.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Instance segmentation separates two objects of the same class.
Statement 2: Semantic segmentation labels every pixel with a class.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Veracity is about the trustworthiness of data.
Statement 2: Velocity is about the speed of data.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: JSON is an example of unstructured data.
Statement 2: A video file is an example of unstructured data.
(d) Statement 1 is false and Statement 2 is true. Statement 1 is false; Statement 2 is true.
Statement 1: Volume is about the huge amount of data.
Statement 2: Variety means that all the data is of one type.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: MySQL is a Big Data analytics tool.
Statement 2: The first step of Big Data analytics is to analyse the data.
(b) Both the statements are false. Statement 1 is false; Statement 2 is false.
Statement 1: Big Data can help a business to take better decisions.
Statement 2: Big Data has no risk to privacy.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: A deep neural network has two or more hidden layers.
Statement 2: A perceptron is also called a standard neural network.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: ReLU gives 0 for a negative input.
Statement 2: The sigmoid function gives values between -1 and 1.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Backpropagation adjusts the weights using the error.
Statement 2: Forward propagation moves from the output layer to the input layer.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: CNNs are mostly used for image data.
Statement 2: RNNs are mostly used for sequence data.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Bias is the layer that gives the final output.
Statement 2: Weights are fixed and never change during training.
(b) Both the statements are false. Statement 1 is false; Statement 2 is false.
Statement 1: Log loss is used for regression.
Statement 2: MSE is a loss function for regression.
(d) Statement 1 is false and Statement 2 is true. Statement 1 is false; Statement 2 is true.
Statement 1: Generative AI creates new content.
Statement 2: A discriminative model creates new images.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: A GAN has a generator and a discriminator.
Statement 2: A VAE has an encoder and a decoder.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: LLMs never make mistakes.
Statement 2: LLMs need no training data.
(b) Both the statements are false. Statement 1 is false; Statement 2 is false.
Statement 1: A deepfake is a real video recorded by a camera.
Statement 2: Deepfakes can be used to spread false information.
(d) Statement 1 is false and Statement 2 is true. Statement 1 is false; Statement 2 is true.
Statement 1: Submitting AI-written work as your own is plagiarism.
Statement 2: Generated content can carry the bias of the training data.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Data, narrative and visuals are the three elements of data storytelling.
Statement 2: A word cloud is used for text data.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: The climax is the first stage of Freytag's Pyramid.
Statement 2: The conclusion is the last stage of Freytag's Pyramid.
(d) Statement 1 is false and Statement 2 is true. Statement 1 is false; Statement 2 is true.
Statement 1: A heat map uses colour to show values.
Statement 2: A candlestick chart is used for text data.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Stories are harder to remember than numbers.
Statement 2: A data story needs no data.
(b) Both the statements are false. Statement 1 is false; Statement 2 is false.
Statement 1: A line graph shows a trend over time.
Statement 2: The audience should be kept in mind while making a data story.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Analytic Approach is the second step of the methodology.
Statement 2: Data Collection comes before Data Requirements.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: A predictive approach answers 'What will happen?'
Statement 2: A diagnostic approach answers 'Why did it happen?'
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: A smaller RMSE means a better regression model.
Statement 2: Accuracy is a metric for regression.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Feature extraction comes after preprocessing.
Statement 2: Image acquisition is done with cameras and sensors.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Histogram equalisation improves the contrast of an image.
Statement 2: Normalisation is a feature extraction technique.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: The pixel value 0 means white in a grayscale image.
Statement 2: An RGB pixel stores three values.
(d) Statement 1 is false and Statement 2 is true. Statement 1 is false; Statement 2 is true.
Statement 1: A spreadsheet of marks is structured data.
Statement 2: An e-mail with attachments is fully structured data.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Hadoop and Spark are tools for Big Data.
Statement 2: Cleaning the data removes errors and duplicates.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: Value is about the usefulness of data.
Statement 2: Velocity is about the trustworthiness of data.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: The input layer receives the data.
Statement 2: The output layer gives the final prediction.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: The sigmoid function gives values between 0 and 1.
Statement 2: tanh gives values between 0 and 255.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Gradient descent is used to reduce the loss.
Statement 2: A loss function measures how correct the hardware is.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: GPT and Gemini are large language models.
Statement 2: DALL-E is used mainly to generate images.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: A discriminative model learns the boundary between classes.
Statement 2: The discriminator of a GAN creates the fake samples.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Prompt engineering means writing clear prompts.
Statement 2: AI-generated content is always free from bias.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: The introduction gives the setting and the characters of the story.
Statement 2: The climax carries the main insight.
(a) Both the statements are true. Statement 1 is true; Statement 2 is true.
Statement 1: A map is used for geographic data.
Statement 2: A word cloud is used for numeric trends over time.
(c) Statement 1 is true and Statement 2 is false. Statement 1 is true; Statement 2 is false.
Statement 1: Hiding data that does not support your view is ethical.
Statement 2: The source of the data should be named.
(d) Statement 1 is false and Statement 2 is true. Statement 1 is false; Statement 2 is true.
Assertion (A): Business Understanding is the first step of the Data Science Methodology.
Reason (R): A clear problem statement decides what data and model will be needed.
(a) Both A and R are true and R is the correct explanation of A. A clear problem decides the data and model needed (R), so it comes first (A).
Assertion (A): MSE is never negative.
Reason (R): In MSE, the errors are squared before they are averaged.
(a) Both A and R are true and R is the correct explanation of A. Errors are squared (R), and squares are never negative (A).
Assertion (A): Data Preparation takes the most time in a data science project.
Reason (R): Raw data has missing values, errors and duplicates that must be handled.
(a) Both A and R are true and R is the correct explanation of A. Raw data has missing values, errors and duplicates (R), which take time to fix (A).
Assertion (A): Cross validation gives a more reliable estimate than one train-test split.
Reason (R): In k-fold cross validation, every part of the data is used for testing once.
(a) Both A and R are true and R is the correct explanation of A. Every part is used for testing once (R), so the estimate is more reliable (A).
Assertion (A): RMSE is used to evaluate a classification model.
Reason (R): RMSE is the square root of the mean of the squared errors.
(d) A is false but R is true. RMSE is for regression.
Assertion (A): Feedback is the last step of the methodology.
Reason (R): Feedback from the users is used to improve the model.
(a) Both A and R are true and R is the correct explanation of A. Feedback from users improves the model (R), closing the cycle (A).
Assertion (A): Preprocessing is done before feature extraction.
Reason (R): Clean and uniform images give better features.
(a) Both A and R are true and R is the correct explanation of A. Clean, uniform images give better features (R).
Assertion (A): A high-resolution image shows more detail.
Reason (R): A high-resolution image has more pixels.
(a) Both A and R are true and R is the correct explanation of A. More pixels (R) capture more detail (A).
Assertion (A): Image classification can tell the position of every object in an image.
Reason (R): Image classification gives one label to the whole image.
(d) A is false but R is true. A is false: classification gives only one label (R is true); detection gives positions.
Assertion (A): Computer vision raises privacy concerns.
Reason (R): Cameras can record and identify people without their consent.
(a) Both A and R are true and R is the correct explanation of A. Cameras can identify people without consent (R).
Assertion (A): An RGB image needs more storage than a grayscale image of the same size.
Reason (R): Each RGB pixel stores three values.
(a) Both A and R are true and R is the correct explanation of A. Each RGB pixel stores three values (R).
Assertion (A): Traditional databases cannot handle Big Data easily.
Reason (R): Big Data has very high volume, velocity and variety.
(a) Both A and R are true and R is the correct explanation of A. Huge volume, speed and variety (R) are too much for them (A).
Assertion (A): A video is structured data.
Reason (R): Structured data is stored in rows and columns.
(d) A is false but R is true. A is false: video is unstructured. R (structured data has rows and columns) is true.
Assertion (A): Data must be cleaned before it is analysed.
Reason (R): Errors and duplicates in data lead to wrong results.
(a) Both A and R are true and R is the correct explanation of A. Errors and duplicates give wrong results (R).
Assertion (A): Activation functions are called the decision-makers of a neural network.
Reason (R): An activation function decides whether, and how strongly, a neuron passes its signal.
(a) Both A and R are true and R is the correct explanation of A. They decide whether and how strongly a neuron fires (R).
Assertion (A): A neural network learns during training.
Reason (R): Backpropagation adjusts the weights to reduce the loss.
(a) Both A and R are true and R is the correct explanation of A. Backpropagation adjusts the weights to reduce loss (R); that is learning (A).
Assertion (A): CNNs are widely used for image recognition.
Reason (R): RNNs have loops that remember earlier inputs.
(b) Both A and R are true but R is not the correct explanation of A. Both are true, but R is about RNNs, so it does not explain A.
Assertion (A): A network with one hidden layer is a deep neural network.
Reason (R): A deep neural network has two or more hidden layers.
(d) A is false but R is true. A is false: deep needs two or more hidden layers (R is true).
Assertion (A): The ReLU function can give a negative output.
Reason (R): ReLU gives max(0, x).
(d) A is false but R is true. A is false: ReLU = max(0, x) is never negative (R is true).
Assertion (A): A loss function is needed to train a neural network.
Reason (R): The loss shows how far the prediction is from the actual value.
(a) Both A and R are true and R is the correct explanation of A. The loss shows how far off the prediction is (R), guiding training (A).
Assertion (A): Generative AI can write a poem that did not exist before.
Reason (R): A generative model learns the pattern of its training data and produces new samples.
(a) Both A and R are true and R is the correct explanation of A. It learns patterns and produces new samples (R).
Assertion (A): The answers of an LLM should be checked before use.
Reason (R): LLMs can hallucinate, that is, give wrong information confidently.
(a) Both A and R are true and R is the correct explanation of A. LLMs can hallucinate (R), so answers must be checked (A).
Assertion (A): In a GAN, the generator and the discriminator work against each other.
Reason (R): The generator tries to fool the discriminator with fake samples.
(a) Both A and R are true and R is the correct explanation of A. The generator tries to fool the discriminator (R).
Assertion (A): Deepfakes are harmless.
Reason (R): Deepfakes are AI-made fake videos, images or audio of real people.
(d) A is false but R is true. A is false: fake videos of real people can mislead and harm. R (what deepfakes are) is true.
Assertion (A): A discriminative model is used to generate new images.
Reason (R): A discriminative model learns to tell the classes apart.
(d) A is false but R is true. A is false: it only tells classes apart (R is true); generative models create images.
Assertion (A): A data story is easier to remember than a table of numbers.
Reason (R): Stories connect the facts with emotions.
(a) Both A and R are true and R is the correct explanation of A. Stories connect facts with emotions (R).
Assertion (A): Visuals are an important element of data storytelling.
Reason (R): Charts show patterns and outliers that are hard to see in raw numbers.
(a) Both A and R are true and R is the correct explanation of A. Charts show patterns and outliers hidden in numbers (R).
Assertion (A): Freytag's Pyramid is used to structure a data story.
Reason (R): It arranges the story as introduction, rising action, climax, falling action and conclusion.
(a) Both A and R are true and R is the correct explanation of A. It arranges the story in five stages (R).
Assertion (A): A word cloud is the best chart for stock prices.
Reason (R): A word cloud shows the frequency of words.
(d) A is false but R is true. A is false: word clouds show word frequency (R is true); stock prices use candlestick or line charts.
Assertion (A): The same data story suits every audience.
Reason (R): A data story should be made keeping the audience in mind.
(d) A is false but R is true. A is false: a story must suit its audience (R is true).
Assertion (A): The analytic approach is chosen after the business problem is understood.
Reason (R): The type of question decides whether a descriptive, diagnostic, predictive or prescriptive approach is needed.
(a) Both A and R are true and R is the correct explanation of A. The type of question decides the approach (R).
Assertion (A): Accuracy alone is enough to judge every classification model.
Reason (R): When classes are unequal, precision and recall are also needed.
(d) A is false but R is true. A is false: with unequal classes, precision and recall are also needed (R is true).
Assertion (A): Noise reduction is done in the preprocessing stage.
Reason (R): Noise can hide the real features of an image.
(a) Both A and R are true and R is the correct explanation of A. Noise hides real features (R), so it is removed while preprocessing (A).
Assertion (A): Semantic segmentation and instance segmentation are the same.
Reason (R): Instance segmentation separates each object of the same class.
(d) A is false but R is true. A is false: instance segmentation separates each object of a class (R is true).
Assertion (A): Edge detection is a feature extraction technique.
Reason (R): Edges mark the boundaries of objects in an image.
(a) Both A and R are true and R is the correct explanation of A. Edges mark object boundaries (R), an important feature (A).
Assertion (A): A deep neural network can learn complex patterns.
Reason (R): It has many hidden layers that learn features step by step.
(a) Both A and R are true and R is the correct explanation of A. Many hidden layers learn features step by step (R).
Assertion (A): Weights are not changed during training.
Reason (R): Backpropagation adjusts the weights to reduce the error.
(d) A is false but R is true. A is false: backpropagation changes the weights (R is true).
Assertion (A): A GAN can create realistic images of faces that do not exist.
Reason (R): The generator improves by trying to fool the discriminator.
(a) Both A and R are true and R is the correct explanation of A. The generator keeps improving to fool the discriminator (R).
Assertion (A): Prompt engineering is useful for getting good output from an LLM.
Reason (R): A clear and specific prompt guides the model better.
(a) Both A and R are true and R is the correct explanation of A. Clear, specific prompts guide the model better (R).
Assertion (A): The output of generative AI needs no checking.
Reason (R): Generative AI can produce wrong or biased content.
(d) A is false but R is true. A is false: it can be wrong or biased (R is true), so it must be checked.
Assertion (A): The climax of a data story should carry the key insight.
Reason (R): The climax is the turning point of Freytag's Pyramid.
(a) Both A and R are true and R is the correct explanation of A. The climax is the turning point (R), the right place for the key insight (A).
Assertion (A): A line graph is suitable to show the change of temperature over a week.
Reason (R): A line graph shows a trend over time.
(a) Both A and R are true and R is the correct explanation of A. Line graphs show trends over time (R).
Assertion (A): Narrative is not needed if the charts are good.
Reason (R): A narrative explains what the data means and why it matters.
(d) A is false but R is true. A is false: the narrative explains what the data means (R is true).
Assertion (A): Hearing and listening are the same.
Reason (R): Listening involves understanding and interpreting the message.
(d) A is false but R is true. Hearing is physical, listening is mental; A is false.
Assertion (A): Remembering is a stage of active listening.
Reason (R): The listener stores the message in memory to use later.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): Paraphrasing shows active listening.
Reason (R): It confirms that the listener has understood the message correctly.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): "A letter was written by Ravi" is in the active voice.
Reason (R): In the active voice the doer comes first.
(d) A is false but R is true. The sentence is passive; A is false; R is true.
Assertion (A): Names of months begin with capital letters.
Reason (R): The M in MINTS stands for months and days.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): Intrinsic motivation comes from within.
Reason (R): It is driven by interest and enjoyment of the task.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): In Maslow's hierarchy, self-actualisation is the most basic need.
Reason (R): Physiological needs such as food and water are at the base.
(d) A is false but R is true. Self-actualisation is the highest need; A is false.
Assertion (A): A result-oriented person focuses on achieving outcomes.
Reason (R): Such a person sets goals, plans and tracks progress.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): Openness is one of the Big Five personality factors.
Reason (R): People high in openness are curious and creative.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): People high in neuroticism are calm and emotionally stable.
Reason (R): Neuroticism is the tendency to feel anxious or upset easily.
(d) A is false but R is true. Low neuroticism means calm; A is false.
Assertion (A): Every formula in a spreadsheet begins with =.
Reason (R): The = sign tells the software that a calculation follows.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): Filtering deletes the rows that do not meet the condition.
Reason (R): Filtering only hides such rows temporarily.
(d) A is false but R is true. A is false; R is true.
Assertion (A): $A$1 is an absolute reference.
Reason (R): An absolute reference does not change when the formula is copied.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): A pie chart is best for showing a trend over time.
Reason (R): A line chart shows change over time.
(d) A is false but R is true. A is false; R is true.
Assertion (A): F5 starts a slide show from the first slide.
Reason (R): Shift + F5 starts from the current slide.
(b) Both A and R are true but R is not the correct explanation of A. Both are true, but R does not explain A.
Assertion (A): A dairy farm is an example of an agricultural entrepreneur.
Reason (R): Agricultural entrepreneurs work in farming and allied activities.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): A faulty business plan is a barrier to entrepreneurship.
Reason (R): A good plan studies customers, costs and profit before starting.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): Fear of failure is an environmental barrier.
Reason (R): Environmental barriers include lack of raw material, labour and roads.
(d) A is false but R is true. Fear of failure is a personal barrier; A is false.
Assertion (A): A successful entrepreneur sees problems as opportunities.
Reason (R): Solving people's problems creates demand for products and services.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): Break-even point is where total revenue equals total cost.
Reason (R): Beyond the break-even point the business makes profit.
(b) Both A and R are true but R is not the correct explanation of A. Both true; R does not define A.
Assertion (A): Green jobs help reduce greenhouse gases.
Reason (R): They promote renewable energy and efficient use of resources.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): Oxygen is a greenhouse gas.
Reason (R): Carbon dioxide and methane trap heat in the atmosphere.
(d) A is false but R is true. Oxygen is not a greenhouse gas; A is false.
Assertion (A): Solid waste management creates green jobs.
Reason (R): Segregation, composting and recycling need trained workers.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): NAPCC was launched to address climate change.
Reason (R): It has eight national missions including the National Solar Mission.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Assertion (A): Eco-tourism is a field of green jobs.
Reason (R): Eco-tourism promotes nature-based travel that protects the environment.
(a) Both A and R are true and R is the correct explanation of A. Both statements are true, and R gives the reason why A is true.
Write any four barriers to active listening.
Thinking about your own reply while the other person speaks; distractions such as noise or the phone; lack of interest in the topic; judging the speaker before he or she finishes; day-dreaming; interrupting. (Any four.)
What does 'P' stand for in the acronym RESPECT used for active listening? Explain.
P stands for Pay attention: focus fully on what the speaker is saying and do not let the mind wander.
Expand the acronym RESPECT used for active listening.
Remove distractions; Eye contact; Show you are listening through gestures; Pay attention; Empathise with the speaker; Clarify doubts; Tune yourself to the timing of the speaker.
Name the five stages of active listening.
Receiving, Understanding, Remembering, Evaluating and Responding.
What is active listening? Why is it important?
Active listening means giving full attention to the speaker, understanding the message and responding to it. It avoids misunderstanding, builds trust and helps us to learn and work well with others.
Write two problems faced by a person who lacks communication skills.
Lack of communication skills leads to confusion, frustration, wasted effort and missed opportunities. The person cannot express ideas clearly or build good relations. (Any two.)
What is the MINTS rule of capitalisation?
Capital letters are used for: Months, the pronoun I, Names of people and places, Titles, and the Starting letter of a sentence.
Change into passive voice:
(i) A letter was written by Ravi. (ii) The food is cooked by the chef.
Name any four parts of speech with one example of each.
Noun: book. Pronoun: she. Verb: run. Adjective: tall. Adverb: quickly. Preposition: on. Conjunction: and. Interjection: wow. (Any four.)
Write two differences between hearing and listening.
(1) Hearing is only receiving sound through the ears; listening is understanding it with attention. (2) Hearing happens without effort; listening is an active skill that needs focus.
Write two tips to write a good paragraph.
(1) Keep one main idea in a paragraph, stated in the first (topic) sentence. (2) Add supporting sentences in a logical order and end with a concluding line; use correct grammar and punctuation.
Explain any two of the Big Five personality factors.
Openness: curious, creative, willing to try new things. Conscientiousness: organised, careful, dependable. Extraversion: outgoing, talkative, energetic. Agreeableness: kind, cooperative. Neuroticism: tendency to feel anxious or upset. (Any two.)
Write two differences between intrinsic and extrinsic motivation.
(1) Intrinsic motivation comes from within: we do the work because we enjoy it. Extrinsic motivation comes from outside: reward, salary, praise. (2) Intrinsic motivation lasts longer; extrinsic motivation stops when the reward stops.
Write any four techniques to become result oriented.
Set clear goals; prepare an action plan; use the right resources and tools; communicate with mentors and peers; make a calendar and work in the given time. (Any four.)
What is a personality disorder? Name the three clusters.
It is a way of thinking, feeling and behaving that is very different from what is expected, and causes problems in daily life. Cluster A: suspicious. Cluster B: emotional and impulsive. Cluster C: anxious.
Name any two personality disorders of Cluster A with one sign of each.
Paranoid: does not trust others, holds grudges. Schizoid: prefers to be alone, shows little emotion. Schizotypal: odd beliefs and behaviour. (Any two.)
Name any two personality disorders of Cluster B with one sign of each.
Antisocial: ignores the rights and feelings of others. Borderline: unstable moods and relations. Histrionic: seeks attention all the time. Narcissistic: thinks he or she is superior and wants praise. (Any two.)
Name any two personality disorders of Cluster C with one sign of each.
Avoidant: avoids people for fear of criticism. Dependent: depends on others for every decision. Obsessive-compulsive: wants perfect order and control. (Any two.)
Write two steps to overcome a personality disorder.
Talk to someone you trust or a counsellor; take up physical activity such as yoga, dance or sports; follow a hobby; choose positive thoughts. (Any two.)
What is self-awareness? How does it help?
Self-awareness is knowing your own strengths, weaknesses, values and feelings. It helps you to set the right goals, use your strengths and improve your weak areas.
What is a positive attitude? Write two ways to build it.
A positive attitude is seeing the good side of a situation and believing that problems can be solved. Ways: start the day with a healthy routine, stay with positive people, be thankful, learn from mistakes. (Any two.)
What is motivation? Why is self-motivation important for a student?
Motivation is the force that drives us to act to reach a goal. A self-motivated student studies regularly without being pushed, finishes work in time and does not give up after a failure.
A teacher has the marks of Classes IX to XII in one spreadsheet. She wants to see only the rows of Class XII. Which feature should she use, and how?
The Filter (AutoFilter) feature. Select the heading row, choose Data, AutoFilter, then open the drop-down arrow of the Class column and tick only XII. The other rows are hidden, not deleted.
Write two differences between sorting and filtering in a spreadsheet.
(1) Sorting arranges all the rows in ascending or descending order; filtering shows only the rows that meet a condition. (2) In sorting no row is hidden; in filtering the other rows are hidden.
Write any four advantages of presentation software.
Changes are easy to make; pictures, audio, video and animation make it interesting; it can be shown to a large audience on a projector; it can be shared and printed easily. (Any four.)
Name the spreadsheet function for each:
(i) SUM (ii) AVERAGE (iii) MAX (iv) COUNT
What is a spreadsheet? Name two spreadsheet programs.
A spreadsheet is a program that stores data in rows and columns and does calculations, sorting and charting. Examples: LibreOffice Calc, Microsoft Excel, Google Sheets. (Any two.)
Why should a spreadsheet be protected with a password?
So that only authorised persons can open or change it. It keeps important data, such as marks or salaries, safe from accidental or wrongful changes.
What is a chart in a spreadsheet? Name two types.
A chart is a picture of the data that makes comparison easy. Types: bar or column chart, line chart, pie chart. (Any two.)
Write the formula to find:
(i) =SUM(B2:B10) (ii) =MAX(C2:C10)
Write the steps to add a new slide and to start the slide show in presentation software.
New slide: Slide menu, New Slide (or Ctrl + M). Slide show: Slide Show menu, Start from First Slide (or press F5).
What is a cell reference? Give an example.
It is the address of a cell, made of the column letter and the row number. Example: B5 is the cell in column B and row 5.
Write two uses of a spreadsheet in a school.
Preparing result sheets with totals and percentages; keeping fee and attendance records; making charts of results. (Any two.)
What are organisational skills? Write any two of them.
They are the abilities to plan work and use time, money and people well. Skills: time management, planning and goal setting, setting priorities, delegating work, keeping records. (Any two.)
Who are business entrepreneurs?
They take up buying and selling (trading) of goods. They are not concerned with manufacturing. Example: a wholesaler or a shop owner.
Name any four types of entrepreneurs.
Service entrepreneurs, business (trading) entrepreneurs, industrial entrepreneurs, agricultural entrepreneurs, technical entrepreneurs, social entrepreneurs, women entrepreneurs. (Any four.)
Write any two barriers to entrepreneurship.
Environmental barriers: lack of raw material, labour, machinery or roads. No or faulty business plan. Personal barriers: self-doubt, fear of failure, lack of confidence. (Any two.)
Write any four entrepreneurial competencies.
Decisiveness, initiative, interpersonal skills, perseverance, organisational skills. (Any four, with a line on each.)
What is perseverance? Why does an entrepreneur need it?
Perseverance is continuing to work towards the goal in spite of difficulties. A new business faces losses and failures in the beginning; only a person who does not give up can succeed.
Write two differences between a service entrepreneur and an industrial entrepreneur.
(1) A service entrepreneur sells a service, such as teaching or repair; an industrial entrepreneur manufactures goods. (2) A service needs less machinery; manufacturing needs a factory, raw material and labour.
What is initiative? Give one example.
Initiative is starting a task on your own, before being told. Example: a shop owner starting home delivery when she sees customers cannot come to the shop.
What are interpersonal skills? Why are they needed in business?
They are the skills of dealing well with people: listening, speaking politely, teamwork and leadership. A business needs good relations with customers, workers and suppliers.
What is a social entrepreneur? Give one example.
A social entrepreneur starts an enterprise mainly to solve a problem of society, not only for profit. Example: an organisation that makes low-cost sanitary pads or gives solar lamps to villages.
Write two entrepreneurial attitudes.
Willingness to take risk, belief in oneself, openness to new ideas, and seeing a problem as an opportunity. (Any two.)
What is the main objective of the National Action Plan on Climate Change (NAPCC), 2008?
To promote development that also deals with climate change: to protect the poor and the environment, increase the use of renewable energy such as solar power, improve energy efficiency and save water and forests, through its eight national missions.
What are green jobs? Give two examples.
Green jobs are jobs that help to protect or restore the environment. Examples: solar panel technician, organic farmer, waste recycling worker, eco-tourism guide. (Any two.)
Explain the role of green jobs in eco-tourism.
Eco-tourism takes people to natural places without harming them. Guides, eco-lodge staff and forest workers educate tourists, protect wildlife and give income to the local people, so nature is conserved.
Write any four benefits of green jobs.
They limit greenhouse gas emissions; reduce waste and pollution; protect and restore ecosystems; save energy and raw material; help in adapting to climate change. (Any four.)
What is appropriate technology? Give one example.
It is small-scale, low-cost, environment-friendly technology that suits local needs. Example: a solar lamp, a biogas plant or a rainwater harvesting tank in a village.
Write two green jobs in the field of solid waste management.
Waste collector and sorter who separates dry and wet waste; recycling plant worker; compost maker. (Any two.)
What are greenhouse gases? Name any two.
Gases that trap the heat of the sun in the atmosphere and make the earth warmer. Examples: carbon dioxide, methane, nitrous oxide, water vapour. (Any two.)
Write two green jobs in transportation.
Driver or mechanic of electric and CNG vehicles; worker in the metro or other public transport; maker of bicycles and e-rickshaws. (Any two.)
How do green jobs help in water conservation? Give two examples.
Workers build rainwater harvesting systems and check dams; plumbers repair leaks and fit water-saving taps; operators run water treatment and recycling plants. (Any two.)
Write two green jobs in building and construction.
Architects and masons who make energy-saving green buildings; workers who install solar panels, insulation and rainwater systems. (Any two.)
Why are green jobs important for India? Write two points.
(1) They reduce pollution and the use of coal and petroleum. (2) They give employment to many people in solar energy, farming, waste management and tourism.
Write two differences between active voice and passive voice with an example.
(1) In active voice the subject does the action: "Meena sang a song." In passive voice the subject receives the action: "A song was sung by Meena." (2) Active voice is direct and short; passive voice is used when the doer is not important.
Write any four ways to become an active listener.
Remove distractions; keep eye contact; nod and show interest; do not interrupt; ask questions to clarify; repeat the main points in your own words. (Any four.)
Name the four types of sentences according to their purpose, with one example of each.
Declarative: I like tea. Interrogative: Do you like tea? Imperative: Bring the tea. Exclamatory: What a lovely tea!
Explain the stages 'Receiving' and 'Responding' of active listening.
Receiving: hearing the words and noticing the body language of the speaker with attention. Responding: giving feedback by words or gestures to show that the message is understood.
Why are communication skills important at the work place? Write two points.
(1) Clear communication avoids mistakes and saves time. (2) It builds good relations with team members and customers and helps in solving problems.
Explain Openness and Conscientiousness of the Big Five personality factors.
Openness: curious, creative, ready for new experiences. Conscientiousness: organised, disciplined, careful and dependable in work.
Explain Extraversion and Agreeableness of the Big Five personality factors.
Extraversion: outgoing, talkative, enjoys being with people. Agreeableness: kind, helpful, cooperative and trusting.
What is result orientation? Why is it needed?
It is focusing on the outcome and completing the work in the given time and quality. It helps a person to achieve goals, earn trust and grow in career.
Give one example each of intrinsic and extrinsic motivation from a student's life.
Intrinsic: solving puzzles because you enjoy them. Extrinsic: studying to win a scholarship or to get praise from parents.
Write two signs of narcissistic personality disorder and two of avoidant personality disorder.
Narcissistic: feels superior to others; wants constant praise. Avoidant: very shy; avoids people and new work for fear of criticism.
Write the steps to sort the data of a spreadsheet in ascending order of marks.
Select the data; open the Data menu and choose Sort; select the Marks column and Ascending; click OK.
Write two differences between a spreadsheet and presentation software.
(1) A spreadsheet stores and calculates data in rows and columns; presentation software shows information as slides. (2) A spreadsheet is used for analysis and charts; a presentation is used to explain ideas to an audience.
Write any four features of a spreadsheet.
Formulas and functions for calculation; sorting and filtering of data; charts; automatic recalculation when data changes; password protection. (Any four.)
What can be added to a slide of a presentation? Name any four.
Text, pictures, tables, charts, audio, video and animation. (Any four.)
Cell A1 has 20, A2 has 30 and A3 has 50. What will =SUM(A1:A3) and =AVERAGE(A1:A3) give?
=SUM(A1:A3) gives 100. =AVERAGE(A1:A3) gives 33.33 (100 divided by 3).
Write two differences between business entrepreneurs and industrial entrepreneurs.
(1) Business entrepreneurs buy and sell goods; industrial entrepreneurs manufacture goods. (2) Example: a wholesaler of clothes is a business entrepreneur; the owner of a garment factory is an industrial entrepreneur.
What is decisiveness? Why is it important for an entrepreneur?
Decisiveness is the ability to take clear and timely decisions. A business faces many choices every day; a delay in decisions causes loss of opportunity.
How can an entrepreneur overcome personal barriers? Write two ways.
(1) Build confidence through training, practice and small successes. (2) Take the help of mentors and family, and treat failure as learning.
Who are agricultural entrepreneurs? Give two examples.
They take up farming and related activities as a business. Examples: a dairy farm owner, a poultry farmer, an organic vegetable grower. (Any two.)
Why is a business plan important?
A business plan states the goal, the product, the customers, the cost and the expected income. It guides every decision and is needed to get a loan. A faulty or missing plan is a barrier to entrepreneurship.
Write two green jobs in agriculture.
Organic farmer who uses compost in place of chemicals; worker who makes vermicompost or bio-pesticides; technician of drip irrigation. (Any two.)
Write two green jobs in solar and wind energy.
Solar panel installer and maintenance technician; wind turbine technician; engineer who designs solar plants. (Any two.)
How do green jobs limit greenhouse gas emissions? Give one example.
They replace polluting activities with clean ones. Example: a solar technician helps homes to use sunlight in place of coal-based electricity, which reduces carbon dioxide.
Name any four missions of the National Action Plan on Climate Change.
National Solar Mission, National Water Mission, National Mission for a Green India, National Mission for Enhanced Energy Efficiency, National Mission for Sustainable Agriculture, National Mission on Sustainable Habitat. (Any four.)
What is eco-tourism? Write one benefit to local people.
Eco-tourism is travel to natural areas in a way that conserves the environment. Local people get jobs as guides, drivers and home-stay owners, so they protect the forest and wildlife.
What is model evaluation? Name its two phases.
Evaluation checks the quality of the model and whether it really answers the business problem. Phases: the diagnostic measures phase (is the model working as intended?) and the statistical significance phase (is the result reliable and not by chance?).
Define Data Science Methodology. How many steps does it have?
It is a step-by-step process that guides a data scientist from understanding the problem to deploying the solution and taking feedback. It has 10 steps.
Name the four types of analytic approach with the question each one answers.
Descriptive: what happened? Diagnostic: why did it happen? Predictive: what will happen? Prescriptive: what should be done?
Write two differences between the Data Requirements and Data Collection steps.
(1) Data Requirements decides what data is needed, in which format and from which sources; Data Collection actually gathers that data. (2) Requirements is planning; collection is doing, and it also checks for gaps in the data.
What is done in the Data Preparation step?
The data is cleaned: missing and wrong values and duplicates are handled. Data from different sources is combined, and new useful features are created (feature engineering). It takes the most time.
What is a train-test split?
The data is divided into two parts: a larger training set to build the model and a smaller testing set to check it on unseen data, for example 80% and 20%.
What is k-fold cross validation?
The data is divided into k equal folds. The model is trained k times; each time one fold is used for testing and the other k - 1 folds for training. The average of the k results is taken.
Can MSE be negative? Give a reason.
No. In MSE every error is squared, so each term is zero or positive. Their average cannot be negative.
Name any four evaluation metrics for classification.
Confusion matrix, accuracy, precision, recall and F1 score. (Any four.)
Write two differences between a descriptive model and a predictive model.
(1) A descriptive model shows the patterns in past data, such as groups of customers; a predictive model forecasts a future outcome. (2) Descriptive models use techniques such as clustering; predictive models use classification and regression.
Why is the Feedback step important?
After deployment, the users tell how the model performs in real use. This feedback is used to correct and refine the model, so the methodology works as a cycle.
Explain any two preprocessing techniques of computer vision with examples.
Noise reduction: removes unwanted grain or speckles, for example from a photo taken in low light. Normalisation: brings pixel values to a common range such as 0 to 1. Resizing/cropping: makes all the images the same size. (Any two.)
What is computer vision? Name two applications.
Computer vision is the field of AI that lets machines see, process and understand images and videos. Applications: face recognition, medical imaging, self-driving cars, surveillance. (Any two.)
Name the five stages of the computer vision process.
Image acquisition, preprocessing, feature extraction, detection/segmentation and high-level processing.
What is feature extraction? Name two techniques.
It finds the important visual patterns of an image that help to recognise objects. Techniques: edge detection, corner detection, texture analysis, colour-based feature extraction. (Any two.)
Write the difference between semantic segmentation and instance segmentation.
Semantic segmentation labels every pixel with a class; two cars get the same label. Instance segmentation also separates each object, so the two cars are marked as car 1 and car 2.
What is high-level processing? How is it different from the earlier stages?
It is the final stage, in which the detected objects are interpreted and a decision is taken, such as recognising a face or reading a number plate. The earlier stages only prepare the image and find features and objects.
Write two challenges of computer vision.
Reasoning and analytical issues; difficulty in getting good images (light, angle, hidden objects); privacy and security of people; duplicate and false content such as deepfakes. (Any two.)
Write two differences between image classification and object detection.
(1) Classification gives one label to the whole image; detection finds many objects and labels each. (2) Classification does not tell the position; detection marks each object with a box.
What is image acquisition? Name two devices used for it.
It is the first stage: capturing the image or video in digital form. Devices: digital camera, scanner, webcam, satellite or medical imaging sensor. (Any two.)
How does a computer see a colour image?
As a grid of pixels. Each pixel stores three numbers from 0 to 255, for red, green and blue. So a colour image is three layers (channels) of numbers.
Write two uses of computer vision in surveillance or transport.
Surveillance: detecting unusual activity and recognising faces in CCTV video. Transport: reading number plates at toll plazas; detecting lanes, vehicles and pedestrians for self-driving cars.
An OTT platform records what millions of users watch every second, as videos, clicks and reviews. Name and explain any two Vs of Big Data seen here.
Volume: a huge amount of data from millions of users. Velocity: data is produced every second and must be processed quickly. Variety: videos, clicks and text reviews are different types of data. (Any two.)
What is Big Data? Give two examples of its sources.
Big Data is data so large, fast and varied that traditional tools cannot store or process it. Sources: social media, online shopping, sensors, CCTV, banking transactions. (Any two.)
Name the three types of Big Data with one example of each.
Structured: an SQL table of marks. Semi-structured: an XML, JSON or CSV file. Unstructured: images, videos, audio, e-mails.
Name the six Vs of Big Data.
Volume, Velocity, Variety, Veracity, Value and Variability.
What is Big Data analytics? Why is it important?
It is the process of examining large and varied data to find hidden patterns, trends and customer preferences. It helps organisations to take better decisions, reduce cost and give better service.
Write two advantages and two disadvantages of Big Data.
Advantages: better decision making; better customer experience; new products. Disadvantages: risk to privacy and security; high cost and need of skilled people; poor quality data gives wrong results. (Any two each.)
What is mining of data streams?
A data stream is data that arrives continuously and at high speed, such as sensor readings or stock prices. Mining it means extracting useful patterns from the stream in real time, as the data cannot all be stored first.
Write two differences between structured and unstructured data.
(1) Structured data has a fixed format of rows and columns; unstructured data has no fixed format. (2) Structured data is easy to search with SQL; unstructured data such as images and videos needs special tools.
Explain Veracity and Value.
Veracity is the trustworthiness and quality of the data: wrong or biased data gives wrong results. Value is the usefulness of the data: data is worth collecting only if it gives benefit to the organisation.
Name any four tools used for Big Data analytics.
Tableau, Apache Hadoop, Cassandra, MongoDB, SAS. Apache Spark and Power BI are also used. (Any four.)
Write any two future trends in Big Data analytics.
Real-time analytics (insights as soon as the data arrives); more advanced models in predictive analytics; quantum computing (much faster processing of huge data). (Any two.)
Differentiate between the two propagation functions of a neural network.
Forward propagation: the input moves from the input layer through the hidden layers to the output layer to give a prediction. Backpropagation: the error is sent backwards from the output to adjust the weights and biases, so the loss becomes smaller.
What is a neural network? Name its three kinds of layers.
A neural network is a machine learning model made of connected units called neurons, inspired by the human brain. Layers: input layer, hidden layers and output layer.
What are weights and bias?
A weight is a number on a connection that shows how important that input is. The bias is an extra number added to the weighted sum, which shifts the output. Both are adjusted during training.
What is an activation function? Name any two.
It is a function applied to the weighted sum of a neuron to decide its output; it adds non-linearity. Examples: sigmoid, tanh, ReLU. (Any two.)
What is a loss function? Name one for regression and one for classification.
A loss function measures how far the predicted output is from the actual output. Regression: Mean Squared Error. Classification: binary cross-entropy (log loss).
Write two differences between CNN and RNN.
(1) A CNN is used for image data; an RNN is used for sequence data such as text and speech. (2) A CNN uses convolution filters to find features; an RNN has loops that remember earlier inputs.
What is a deep neural network?
A neural network with two or more hidden layers. More layers let it learn complex patterns, as in face recognition and speech recognition.
What is gradient descent?
It is the method by which a network reduces its loss: after each pass, the weights are changed by a small step in the direction that makes the loss smaller, until the loss is minimum.
Name any four types of neural networks.
Standard neural network (perceptron), feed-forward neural network, convolutional neural network, recurrent neural network, generative adversarial network. (Any four.)
What is a perceptron?
It is the simplest neural network, with a single layer of neurons. It adds the weighted inputs and the bias, applies an activation function and gives the output. It is also called the standard neural network.
Write two effects of neural networks on society.
Positive: better diagnosis of diseases, voice assistants, safer driving. Concern: loss of some jobs, bias in decisions, and misuse such as deepfakes. (One of each.)
What are deepfakes? Write their ethical and social effects with one example.
Deepfakes are AI-made fake videos, images or audio that look real. They can spread false news, damage a person's image and cheat people. Example: a fake video of a leader making a statement he never made.
What is generative AI? Give two examples of tools.
Generative AI is the part of AI that creates new content, such as text, images, audio and video, that resembles its training data. Tools: ChatGPT, Gemini, DALL-E, Canva, Stable Diffusion. (Any two.)
Write two differences between generative and discriminative models.
(1) A generative model learns how the data is distributed and creates new data; a discriminative model learns the boundary between classes and labels data. (2) Example: a GAN making faces is generative; a spam classifier is discriminative.
What is a GAN? Name its two parts.
A Generative Adversarial Network has two neural networks that compete. The generator makes fake samples; the discriminator tries to tell real from fake. Both improve until the fakes look real.
What is an LLM? Give two examples.
A Large Language Model is a deep learning model trained on a huge amount of text to understand and generate human language. Examples: GPT, Gemini, LLaMA. (Any two.)
Write any two limitations of LLMs.
They need very high processing power and data; they can give wrong answers confidently (hallucination); they can carry the bias of their training data; they do not truly understand. (Any two.)
Write any two ethical concerns of generative AI.
Deepfakes and false information; bias and discrimination in outputs; plagiarism and copyright of the training data; loss of privacy. (Any two.)
What is a Variational Autoencoder (VAE)?
A VAE is a generative model with an encoder that compresses the input into a small code and a decoder that rebuilds data from the code. By changing the code, new similar data can be generated.
Write two applications of generative AI.
Writing and summarising text; making images and designs from prompts; creating music and voice; generating video; writing computer code. (Any two.)
What is prompt engineering? Why is it needed?
It is the skill of writing a clear and specific instruction (prompt) for a generative AI tool. A better prompt gives a more accurate and useful output.
How can a student use generative AI responsibly? Write two points.
(1) Use it to learn and get ideas, not to copy; mention that AI was used. (2) Check the facts from reliable sources and do not create or share fake content.
What is data visualisation? Write its uses.
Data visualisation is showing data as charts, graphs and maps. It makes large data easy to understand, shows patterns, trends and outliers quickly, and helps in taking decisions.
What is data storytelling?
It is presenting the insights of data as a story, by combining data, narrative and visuals, so that the audience understands, remembers and acts on it.
Name the three key elements of data storytelling and the role of each.
Data: the facts and findings. Narrative: the story that explains the data. Visuals: charts that show the patterns.
Name the five stages of Freytag's Pyramid.
Introduction, rising action, climax, falling action and conclusion. (The introduction is also called the exposition and the conclusion the resolution.)
Write two differences between data visualisation and data storytelling.
(1) Visualisation shows the data as charts; storytelling adds a narrative that explains what the charts mean. (2) Visualisation tells what is happening; storytelling also tells why it matters and what should be done.
Which visualisation will you choose for:
(i) Word cloud (ii) Candlestick chart (iii) Map (iv) Line graph
Why is a story more powerful than plain data? Write two points.
(1) Stories create interest and emotion, so people pay attention. (2) People remember a story much longer than a list of numbers, and are more likely to act.
Write the steps to create a data story.
Collect and organise the data; use visualisations to find insights; examine the relationships in the data; build a simple narrative with a problem and its solution, keeping the audience in mind.
Write two ethical points to follow in data storytelling.
(1) Show the data honestly: do not hide facts or change the scale of a graph to mislead. (2) Protect the privacy of the people in the data and name the source.
What happens when:
(i) The narrative explains what is happening in the data. (ii) The visuals enlighten the audience by showing patterns and outliers that they would not see in the raw data.
What is a heat map? When is it used?
A heat map shows values with shades of colour in a grid. It is used to compare categories quickly, for example the sales of products in different months.
Explain image classification and classification with localisation.
Image classification gives one label to the whole image, such as 'cat'. Classification with localisation gives the label and also the position of that single object, by a box around it.
What is noise in an image? How is it handled?
Noise is the unwanted random variation in pixel values, seen as grain or speckles, caused by low light or a poor sensor. It is removed by noise reduction filters in the preprocessing stage.
Write two uses of computer vision in health care.
Detecting tumours, fractures and infections in X-ray, CT and MRI images; checking the retina for eye disease; helping surgeons with image-guided surgery. (Any two.)
Why is preprocessing needed in computer vision?
Raw images differ in size, brightness and quality, and may have noise. Preprocessing makes them clean and uniform, so that the features found later are correct.
Write two future trends of computer vision.
Personalised health-care diagnosis from medical images; immersive augmented reality (AR) experiences; systems that understand visual data with near-human precision. (Any two.) This progress needs teamwork and attention to ethics.
Explain Volume and Velocity with one example each.
Volume is the huge amount of data: a video site gets hundreds of hours of video every minute. Velocity is the speed of data: a stock exchange receives thousands of trades every second.
What is semi-structured data? Give two examples.
Data that has no fixed table but carries tags or keys that organise it. Examples: XML, JSON, CSV and HTML files. (Any two.)
Why must data be cleaned before analysis? Give an example.
Raw data has errors, missing values and duplicates. If the same customer is counted twice, the sales report will be wrong. Cleaning makes the result reliable.
How does Big Data help an online shopping company? Write two points.
(1) It studies what each customer views and buys, and recommends products. (2) It predicts demand, so that the right stock is kept and delivery is faster.
Write two differences between semi-structured and unstructured data.
(1) Semi-structured data has tags or keys (XML, JSON); unstructured data has no organisation at all (images, videos). (2) Semi-structured data is easier to search than unstructured data.
Name any four generative AI tools and the kind of content each creates.
ChatGPT and Gemini: text. DALL-E, Canva and Stable Diffusion: images. Google Lumiere and Animaker: video. Google MusicLM and Meta AI Voicebox: audio. (Any four.)
Write the roles of the generator and the discriminator in a GAN.
The generator creates fake samples from random noise and tries to make them look real. The discriminator examines real and fake samples and tries to tell them apart.
What is hallucination in an LLM? Give one example.
It is when the model gives a wrong or made-up answer in a confident way. Example: it gives the name of a book and author that do not exist.
Write two tips for writing a good prompt.
(1) Be clear and specific: state the task, the length and the audience. (2) Give context or an example, and ask for the format you want.
How can generative AI help teachers and students? Write two points.
It can explain a topic in simple words and make practice questions; it can summarise long text and give ideas for projects. The output must be checked.
What is a narrative in data storytelling?
It is the storyline that connects the facts: it explains what the data shows, why it happened and what should be done. It gives meaning to the numbers.
Explain the 'rising action' and 'climax' of Freytag's Pyramid in a data story.
Rising action: the facts that build the problem and create interest. Climax: the main finding or turning point, the most important insight of the story.
Why should the audience be kept in mind while making a data story?
Different audiences have different knowledge and needs. Managers want the decision; experts want details. The language, visuals and depth must suit them, or the message is lost.
Which chart will you use for:
(i) Pie chart, because it shows the parts of a whole. (ii) Bar chart, because it compares separate values.
Name the elements of a story.
Characters, plot or setting, conflict (the problem), resolution (the solution) and insights.
A model was tested on 100 cases. The results are: TP = 48, TN = 34, FP = 16, FN = 2. Calculate the accuracy and the error rate of the model.
Accuracy = (TP + TN) / Total = (48 + 34) / 100 = 82 / 100 = 0.82 (82%)
Error rate = 1 - Accuracy = (FP + FN) / Total = 18 / 100 = 0.18 (18%)
For a model, TP = 12, FP = 13 and FN = 12. Calculate its precision and recall.
Precision = TP / (TP + FP) = 12 / (12 + 13) = 12 / 25 = 0.48 (48%)
Recall = TP / (TP + FN) = 12 / (12 + 12) = 12 / 24 = 0.50 (50%)
The precision of a model is 0.90 and its recall is 0.60. Calculate the F1 score.
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.90 x 0.60 / (0.90 + 0.60) = 1.08 / 1.50 = 0.72
A model was tested on 200 cases. The results are: TP = 36, TN = 104, FP = 24, FN = 36. Calculate the accuracy and the error rate of the model.
Accuracy = (TP + TN) / Total = (36 + 104) / 200 = 140 / 200 = 0.70 (70%)
Error rate = 1 - Accuracy = (FP + FN) / Total = 60 / 200 = 0.30 (30%)
For a model, TP = 72, FP = 8 and FN = 28. Calculate its precision and recall.
Precision = TP / (TP + FP) = 72 / (72 + 8) = 72 / 80 = 0.90 (90%)
Recall = TP / (TP + FN) = 72 / (72 + 28) = 72 / 100 = 0.72 (72%)
The precision of a model is 0.50 and its recall is 0.60. Calculate the F1 score.
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.50 x 0.60 / (0.50 + 0.60) = 0.6 / 1.10 = 0.55
A model was tested on 50 cases. The results are: TP = 21, TN = 13, FP = 9, FN = 7. Calculate the accuracy and the error rate of the model.
Accuracy = (TP + TN) / Total = (21 + 13) / 50 = 34 / 50 = 0.68 (68%)
Error rate = 1 - Accuracy = (FP + FN) / Total = 16 / 50 = 0.32 (32%)
For a model, TP = 9, FP = 6 and FN = 3. Calculate its precision and recall.
Precision = TP / (TP + FP) = 9 / (9 + 6) = 9 / 15 = 0.60 (60%)
Recall = TP / (TP + FN) = 9 / (9 + 3) = 9 / 12 = 0.75 (75%)
The precision of a model is 0.90 and its recall is 0.60. Calculate the F1 score.
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.90 x 0.60 / (0.90 + 0.60) = 1.08 / 1.50 = 0.72
A model was tested on 100 cases. The results are: TP = 36, TN = 28, FP = 24, FN = 12. Calculate the accuracy and the error rate of the model.
Accuracy = (TP + TN) / Total = (36 + 28) / 100 = 64 / 100 = 0.64 (64%)
Error rate = 1 - Accuracy = (FP + FN) / Total = 36 / 100 = 0.36 (36%)
For a model, TP = 24, FP = 16 and FN = 8. Calculate its precision and recall.
Precision = TP / (TP + FP) = 24 / (24 + 16) = 24 / 40 = 0.60 (60%)
Recall = TP / (TP + FN) = 24 / (24 + 8) = 24 / 32 = 0.75 (75%)
The precision of a model is 0.60 and its recall is 0.40. Calculate the F1 score.
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.60 x 0.40 / (0.60 + 0.40) = 0.48 / 1.00 = 0.48
A model predicted the price of a shirt (in rupees) as 270. The actual value is 250. Calculate the error, the error rate and the accuracy of the prediction.
Error = |actual - predicted| = |250 - 270| = 20. Error rate = error / actual = 20 / 250 = 0.08 = 8%. Accuracy = 100% - 8% = 92%
A model predicted the marks of a class (total) as 380. The actual value is 400. Calculate the error, the error rate and the accuracy of the prediction.
Error = |actual - predicted| = |400 - 380| = 20. Error rate = error / actual = 20 / 400 = 0.05 = 5%. Accuracy = 100% - 5% = 95%
A model predicted the daily visitors of a park as 560. The actual value is 500. Calculate the error, the error rate and the accuracy of the prediction.
Error = |actual - predicted| = |500 - 560| = 60. Error rate = error / actual = 60 / 500 = 0.12 = 12%. Accuracy = 100% - 12% = 88%
A model predicted the units of electricity used as 720. The actual value is 800. Calculate the error, the error rate and the accuracy of the prediction.
Error = |actual - predicted| = |800 - 720| = 80. Error rate = error / actual = 80 / 800 = 0.10 = 10%. Accuracy = 100% - 10% = 90%
A model predicted the weight of a bag of rice (in grams) as 1040. The actual value is 1000. Calculate the error, the error rate and the accuracy of the prediction.
Error = |actual - predicted| = |1000 - 1040| = 40. Error rate = error / actual = 40 / 1000 = 0.04 = 4%. Accuracy = 100% - 4% = 96%
A model predicted the monthly sale of a shop (in rupees) as 1700. The actual value is 2000. Calculate the error, the error rate and the accuracy of the prediction.
Error = |actual - predicted| = |2000 - 1700| = 300. Error rate = error / actual = 300 / 2000 = 0.15 = 15%. Accuracy = 100% - 15% = 85%
An image has the resolution 640 x 480.
Pixels = 640 x 480 = 307,200
Grayscale: 1 byte per pixel = 307,200 bytes. RGB: 3 bytes per pixel = 307,200 x 3 = 921,600 bytes
An image has the resolution 1280 x 720.
Pixels = 1280 x 720 = 921,600
Grayscale: 1 byte per pixel = 921,600 bytes. RGB: 3 bytes per pixel = 921,600 x 3 = 2,764,800 bytes
An image has the resolution 100 x 100.
Pixels = 100 x 100 = 10,000
Grayscale: 1 byte per pixel = 10,000 bytes. RGB: 3 bytes per pixel = 10,000 x 3 = 30,000 bytes
An image has the resolution 1920 x 1080.
Pixels = 1920 x 1080 = 2,073,600
Grayscale: 1 byte per pixel = 2,073,600 bytes. RGB: 3 bytes per pixel = 2,073,600 x 3 = 6,220,800 bytes
An image has the resolution 800 x 600.
Pixels = 800 x 600 = 480,000
Grayscale: 1 byte per pixel = 480,000 bytes. RGB: 3 bytes per pixel = 480,000 x 3 = 1,440,000 bytes
An image has the resolution 256 x 256.
Pixels = 256 x 256 = 65,536
Grayscale: 1 byte per pixel = 65,536 bytes. RGB: 3 bytes per pixel = 65,536 x 3 = 196,608 bytes
The actual values are 82, 83, 47, 60 and the values predicted by a model are 82, 83, 47, 64. Calculate the MSE and the RMSE.
Errors: 0, 0, 0, -4. Squared errors: 0, 0, 0, 16. MSE = 16 / 4 = 4. RMSE = square root of 4 = 2
The actual values are 88, 102, 89, 116 and the values predicted by a model are 91, 99, 92, 119. Calculate the MSE and the RMSE.
Errors: -3, 3, -3, -3. Squared errors: 9, 9, 9, 9. MSE = 36 / 4 = 9. RMSE = square root of 9 = 3
The actual values are 70, 54, 121, 112 and the values predicted by a model are 69, 53, 126, 109. Calculate the MSE and the RMSE.
Errors: 1, 1, -5, 3. Squared errors: 1, 1, 25, 9. MSE = 36 / 4 = 9. RMSE = square root of 9 = 3
The actual values are 30, 10, 32, 43 and the values predicted by a model are 37, 15, 33, 38. Calculate the MSE and the RMSE.
Errors: -7, -5, -1, 5. Squared errors: 49, 25, 1, 25. MSE = 100 / 4 = 25. RMSE = square root of 25 = 5
The actual values are 55, 27, 24, 29 and the values predicted by a model are 51, 23, 28, 25. Calculate the MSE and the RMSE.
Errors: 4, 4, -4, 4. Squared errors: 16, 16, 16, 16. MSE = 64 / 4 = 16. RMSE = square root of 16 = 4
The actual values are 47, 52, 42 and the values predicted by a model are 40, 57, 43. Calculate the MSE and the RMSE.
Errors: 7, -5, -1. Squared errors: 49, 25, 1. MSE = 75 / 3 = 25. RMSE = square root of 25 = 5
The actual values are 95, 125, 118 and the values predicted by a model are 97, 127, 120. Calculate the MSE and the RMSE.
Errors: -2, -2, -2. Squared errors: 4, 4, 4. MSE = 12 / 3 = 4. RMSE = square root of 4 = 2
The actual values are 137, 137, 132, 68 and the values predicted by a model are 137, 133, 132, 68. Calculate the MSE and the RMSE.
Errors: 0, 4, 0, 0. Squared errors: 0, 16, 0, 0. MSE = 16 / 4 = 4. RMSE = square root of 4 = 2
The actual values are 12, 54, 15, 48 and the values predicted by a model are 15, 57, 18, 51. Calculate the MSE and the RMSE.
Errors: -3, -3, -3, -3. Squared errors: 9, 9, 9, 9. MSE = 36 / 4 = 9. RMSE = square root of 9 = 3
The actual values are 59, 50, 44, 21 and the values predicted by a model are 60, 53, 45, 16. Calculate the MSE and the RMSE.
Errors: -1, -3, -1, 5. Squared errors: 1, 9, 1, 25. MSE = 36 / 4 = 9. RMSE = square root of 9 = 3
A neuron has the inputs x1 = 1 and x2 = 2, the weights w1 = 3 and w2 = -2, and the bias -2.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (1 x 3) + (2 x -2) + (-2) = -3
ReLU(-3) = max(0, -3) = 0
A neuron has the inputs x1 = 5 and x2 = 2, the weights w1 = 3 and w2 = -1, and the bias 1.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (5 x 3) + (2 x -1) + (1) = 14
ReLU(14) = max(0, 14) = 14
A neuron has the inputs x1 = 3 and x2 = 3, the weights w1 = -2 and w2 = -2, and the bias -6.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (3 x -2) + (3 x -2) + (-6) = -18
ReLU(-18) = max(0, -18) = 0
A neuron has the inputs x1 = 1 and x2 = 2, the weights w1 = -1 and w2 = 1, and the bias -2.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (1 x -1) + (2 x 1) + (-2) = -1
ReLU(-1) = max(0, -1) = 0
A neuron has the inputs x1 = 5 and x2 = 1, the weights w1 = -2 and w2 = 1, and the bias -4.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (5 x -2) + (1 x 1) + (-4) = -13
ReLU(-13) = max(0, -13) = 0
A neuron has the inputs x1 = 2 and x2 = 2, the weights w1 = -2 and w2 = 1, and the bias -6.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (2 x -2) + (2 x 1) + (-6) = -8
ReLU(-8) = max(0, -8) = 0
A neuron has the inputs x1 = 1 and x2 = 5, the weights w1 = -1 and w2 = -2, and the bias -2.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (1 x -1) + (5 x -2) + (-2) = -13
ReLU(-13) = max(0, -13) = 0
A neuron has the inputs x1 = 1 and x2 = 2, the weights w1 = -1 and w2 = 1, and the bias -6.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (1 x -1) + (2 x 1) + (-6) = -5
ReLU(-5) = max(0, -5) = 0
A neuron has the inputs x1 = 2 and x2 = 4, the weights w1 = 1 and w2 = 1, and the bias 2.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (2 x 1) + (4 x 1) + (2) = 8
ReLU(8) = max(0, 8) = 8
A neuron has the inputs x1 = 2 and x2 = 5, the weights w1 = 2 and w2 = 1, and the bias -2.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (2 x 2) + (5 x 1) + (-2) = 7
ReLU(7) = max(0, 7) = 7
A neuron has the inputs x1 = 3 and x2 = 1, the weights w1 = -1 and w2 = 3, and the bias -6.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (3 x -1) + (1 x 3) + (-6) = -6
ReLU(-6) = max(0, -6) = 0
A neuron has the inputs x1 = 1 and x2 = 1, the weights w1 = -1 and w2 = 3, and the bias 2.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (1 x -1) + (1 x 3) + (2) = 4
ReLU(4) = max(0, 4) = 4
A neuron has the inputs x1 = 3 and x2 = 3, the weights w1 = -2 and w2 = 2, and the bias -2.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (3 x -2) + (3 x 2) + (-2) = -2
ReLU(-2) = max(0, -2) = 0
A neuron has the inputs x1 = 1 and x2 = 3, the weights w1 = -1 and w2 = -1, and the bias 1.
Weighted sum = (x1 x w1) + (x2 x w2) + bias = (1 x -1) + (3 x -1) + (1) = -3
ReLU(-3) = max(0, -3) = 0
Explain the working of Big Data analytics in four steps.
Step 1, Gather data: data is collected from many sources, such as cloud storage, mobile apps and sensors, in structured and unstructured forms. Step 2, Process data: the data is organised so that queries give correct results; it can be processed in batches or as a stream. Step 3, Clean data: errors, duplicates and irrelevant data are removed and the format is made uniform. Step 4, Analyse data: methods such as data mining, predictive analytics and deep learning turn the data into useful insights.
Explain the six Vs of Big Data with one line on each.
Volume: the huge amount of data. Velocity: the speed at which it is produced and processed. Variety: the different types: text, image, audio, video. Veracity: the trustworthiness and quality of the data. Value: the benefit the data gives. Variability: the changes in the flow and meaning of data with time.
Explain structured, semi-structured and unstructured data with two examples of each.
Structured: fixed rows and columns; SQL tables, spreadsheets. Semi-structured: has tags or keys but no fixed table; XML, JSON, CSV and HTML files. Unstructured: no fixed format; images, videos, audio, e-mails, social media posts. Most of the data in the world is unstructured.
Write two advantages and two disadvantages of Big Data. Explain each in one line.
Advantages: (1) Better decision making, from patterns found in the data. (2) Better customer experience, through personalised suggestions. Disadvantages: (1) Privacy and security: personal data may be leaked or misused. (2) Cost and skill: storage, tools and trained people are expensive; poor data quality gives wrong results.
A supermarket chain collects bills, CCTV video, app clicks and customer reviews.
(i) Bills: structured. CCTV video and reviews: unstructured. App click logs (JSON): semi-structured. (ii) Volume and Variety (also Velocity). (iii) It can find which products sell together and stock them accordingly; it can give personalised offers and reduce waste.
What is mining of data streams? Why is it difficult? Give two examples of data streams.
A data stream is data that flows continuously and fast. Mining it means finding patterns in real time. It is difficult because the data is unlimited, cannot all be stored, and must be processed in one pass with limited memory. Examples: sensor readings of a factory, stock market prices, social media feeds.
Explain three future trends of Big Data analytics. Why do they matter?
Real-time analytics: data is analysed as soon as it arrives, so decisions are immediate, as in fraud alerts. Advanced models in predictive analytics: better machine learning models forecast demand, risk and behaviour more correctly. Quantum computing: very fast processing of huge and complex data. They matter because data keeps growing in volume and speed, and organisations need quicker and more correct decisions.
Differentiate between traditional data and Big Data in four points.
(1) Size: traditional data is in megabytes or gigabytes; Big Data is in terabytes and more. (2) Type: mostly structured; structured, semi-structured and unstructured. (3) Speed: produced slowly; produced every second. (4) Tools: a database such as MySQL; tools such as Hadoop and Spark.
How is Big Data analytics useful in:
(i) Predicting outbreaks of disease and helping diagnosis from patient records. (ii) Detecting fraud from unusual transactions. (iii) Finding weak areas of students and personalising learning. (iv) Planning routes and reducing traffic jams from GPS data.
What is Big Data analytics? Name four tools and write one line on any two.
It is the process of examining very large and varied data to find patterns and insights. Tools: Tableau, Apache Hadoop, Cassandra, MongoDB, SAS. Hadoop stores and processes huge data on many computers together. Tableau makes interactive charts and dashboards from data.
(i) Large Language Model. (ii) GPT (used in ChatGPT), Gemini, LLaMA. (Any two.) (iii) They need very high processing power and huge data. They can give wrong or made-up answers confidently (hallucination) and can carry bias from the training data.
Differentiate between generative and discriminative models in four points.
(1) Aim: a generative model creates new data; a discriminative model classifies data. (2) Learning: learns how the data is distributed; learns the boundary between classes. (3) Output: a new image, text or sound; a label. (4) Examples: GAN, VAE; logistic regression, a spam classifier.
Explain the working of a GAN.
A GAN has two networks. The generator takes random noise and produces fake samples. The discriminator gets real samples and fake samples and tries to tell them apart. Its result is sent back to both: the generator learns to make better fakes, the discriminator learns to catch them. Training continues until the fakes look real.
Explain any four applications of generative AI with one tool for each.
Text: writing, summarising, answering; ChatGPT, Gemini. Image: pictures from prompts; Canva, DALL-E, Stable Diffusion. Video: clips from text; Google Lumiere, Animaker. Audio: music and voice; Google MusicLM, Meta AI Voicebox. It is also used to write computer code.
Explain the ethical and social implications of generative AI in four points.
Deepfakes: fake videos spread false information. Bias and discrimination: outputs repeat the bias of the training data. Plagiarism and copyright: content is produced from others' work and may be passed off as one's own. Transparency: users may not know that the content is made by AI. Jobs of some creators may also be affected.
What are LLMs? Write two applications and two risks.
Large Language Models are deep learning models trained on huge text to understand and generate language. Applications: chatbots and virtual assistants; translation, summarising and writing code. Risks: wrong information given confidently; misuse for spam, cheating and fake news; bias.
Explain the working of the two generative models, GAN and VAE. Name the parts of each.
GAN (Generative Adversarial Network): it has a generator, which creates new samples, and a discriminator, which judges whether a sample is real or fake. The two compete, so the generated data becomes more and more realistic. VAE (Variational Autoencoder): it has an encoder, which compresses the data into a small code (latent space), and a decoder, which rebuilds data from that code. By changing the code, new similar data is generated.
A student uses an AI tool to write a full project report and submits it as her own. Discuss the issue and suggest three good practices.
It is plagiarism: the work is not hers, and the content may have errors or copied material. Good practices: (1) use AI only for ideas and write in her own words; (2) verify every fact from reliable sources; (3) state clearly where AI was used and cite sources.
What are deepfakes? How are they made? Write two harms and two ways to stay safe.
Deepfakes are fake videos, images or audio of real people made with generative AI, usually GANs. Harms: false news and damage to a person's image; fraud by copying a voice or face. Safety: check the source before believing or sharing; look for unnatural movement or voice; report fake content.
Compare any two generative AI tools for images with any two for text.
Images: DALL-E and Stable Diffusion create pictures from a text prompt; Canva also has a text-to-image tool. Text: ChatGPT and Gemini answer questions, write and summarise text and code. Image tools give pictures as output; text tools give language as output; both need a clear prompt.
What is data storytelling? Explain its key elements.
Data storytelling is communicating the insights of data through a story. Its three elements: Data: the facts, found by analysis. Narrative: the storyline that explains what the data means and why it matters. Visuals: charts and graphs that show the patterns. Data with narrative explains; data with visuals enlightens; narrative with visuals engages; all three together can drive change.
Explain Freytag's Pyramid for a data story.
Introduction: give the setting, the characters and the question. Rising action: show the data that builds the problem. Climax: the main insight, the turning point. Falling action: what follows from the insight, with supporting facts. Conclusion: the problem is resolved and the action is suggested.
Write the steps to create a data story.
(1) Collect the data and organise it. (2) Use proper visualisation tools to visualise the data. (3) Observe the relationships in the data. (4) Create a simple narrative, hidden in the data, to communicate to the audience.
Which visualisation suits each type of data? Give five pairs.
Text data: word cloud. Numeric data over time: line graph. Stock prices: candlestick chart. Geographic data: map. Comparison of categories by colour: heat map. Mixed data for groups: facet grid.
Why is data storytelling powerful? Explain with the three combinations of its elements.
Data + narrative explains what is happening and why it is important. Data + visuals enlightens: the audience sees patterns and outliers. Narrative + visuals engages the audience. When all three come together, the story can influence people and drive change.
A school finds that attendance falls every Monday. Plan a data story on this using Freytag's Pyramid.
Introduction: the school tracks daily attendance. Rising action: a line graph shows a dip on every Monday. Climax: the dip is largest for students who travel far. Falling action: Monday tests and lost learning are linked to the dip. Conclusion: suggest a Monday activity period and a bus timing change.
Write four ethical practices to follow in data storytelling.
(1) Be accurate: do not change or hide data. (2) Do not mislead with cut scales or selected data only. (3) Protect privacy: remove names and personal details. (4) Be transparent: tell the source and the limits of the data.
Differentiate between a story, data visualisation and data storytelling.
A story is a narrative with characters, a conflict and a resolution; it needs no data. Data visualisation shows data as charts; it has no narrative. Data storytelling combines data, visuals and a narrative so that the audience understands the insight and acts on it.
What are the elements of a story? How are they used in a data story?
Characters, plot or setting, conflict, resolution and insights. In a data story: the characters are the people affected (customers, students); the setting is the time and place of the data; the conflict is the problem shown by the data; the resolution is the solution suggested; the insights are what the audience learns.
Explain with an example how a good visual improves a data story.
A table of monthly sales of 12 months is hard to read. A line graph of the same data shows at once that sales fall in the rainy months. A highlight on that dip, with a line of text, directs the audience to the key insight and makes it easy to remember.
What is data modelling? Explain descriptive and predictive models with the techniques used.
Modelling is the step in which a model is built from the prepared data to answer the question. Descriptive models show what has happened and find patterns; techniques: clustering, summary statistics, association. Predictive models forecast future outcomes; techniques: classification (for a category) and regression (for a number). The model is trained on a training set and tested on a test set.
Name the ten steps of the Data Science Methodology in order.
Business Understanding, Analytic Approach, Data Requirements, Data Collection, Data Understanding, Data Preparation, Modelling, Evaluation, Deployment, Feedback.
Explain the four types of analytic approach with one example of each.
Descriptive: what happened? Last year's sales report. Diagnostic: why did it happen? Finding why sales fell in June. Predictive: what will happen? Forecasting next month's sales. Prescriptive: what should be done? Suggesting a discount to increase sales.
Explain train-test split and k-fold cross validation. Which is more reliable, and why?
Train-test split: the data is divided once into a training set and a testing set. k-fold cross validation: the data is divided into k folds; the model is trained k times, each time testing on a different fold, and the results are averaged. Cross validation is more reliable because every record is used for testing once; but it takes more time.
Explain the steps Business Understanding, Analytic Approach, Data Requirements and Data Collection in one line each.
Business Understanding: define the problem and the goal clearly. Analytic Approach: decide how data can answer it: descriptive, diagnostic, predictive or prescriptive. Data Requirements: decide what data, format and sources are needed. Data Collection: gather the data and check for gaps.
What is a neural network? Explain its layers.
A neural network is a machine learning model made of layers of connected neurons, inspired by the brain. Input layer: receives the features of the data, one neuron for each feature. Hidden layers: each neuron multiplies inputs by weights, adds a bias and applies an activation function; these layers find the patterns. Output layer: gives the final prediction. A network with two or more hidden layers is a deep neural network.
Explain the working of a neural network using the terms forward propagation, loss function and backpropagation.
In forward propagation the input passes through the layers: each neuron finds the weighted sum plus bias and applies the activation function, and the output layer gives a prediction. The loss function measures the error between the prediction and the actual value. In backpropagation this error is sent backwards and the weights and biases are adjusted by gradient descent. The cycle repeats for many epochs.
Explain any four types of neural networks with one use of each.
Perceptron (standard NN): single layer; simple yes/no classification. Feed-forward NN: data moves one way through hidden layers; predicting marks or prices. CNN: uses convolution filters; image recognition. RNN: has memory of earlier inputs; text and speech. (GAN: generator and discriminator; creating images.)
Explain the components of a neural network: neurons, weights, bias and activation function.
Neurons: the units that receive inputs, process them and pass an output. Weights: numbers on the connections that show the importance of each input. Bias: an extra value added to the weighted sum to shift the output. Activation function: decides the output of the neuron and adds non-linearity; examples are sigmoid, tanh and ReLU.
Describe the structure of an artificial neural network by its three layers, and explain the role of weights.
The input layer takes the data. The hidden layers do the computation. The output layer gives the result. Every connection between two neurons has a weight. The weight decides how much one neuron's output affects the next neuron. Training means adjusting these weights so that the output becomes correct.
Explain Volume, Velocity, Variety and Veracity with the example of a social media platform.
Volume: billions of posts, photos and videos are stored. Velocity: thousands of new posts arrive every second and must be shown at once. Variety: text, images, videos, likes and locations are different types of data. Veracity: fake accounts and false news reduce the trustworthiness of the data.
A city wants to use Big Data from traffic cameras, GPS of buses and mobile apps to reduce traffic jams. Explain how the four steps of Big Data analytics will be applied.
Gather: collect video, GPS and app data from all the sources. Process: bring the data to one platform and organise it, mostly as a live stream. Clean: remove faulty readings and duplicates. Analyse: find the busy routes and hours, predict jams and change signal timings and bus routes.
What is Big Data? Write three of its sources and three areas where it is used.
Big Data is data too large, fast and varied for traditional tools. Sources: social media, online transactions, sensors and CCTV. Uses: health care (predicting disease), banking (fraud detection), retail (recommendations), transport (route planning). (Any three each.)
Explain the terms:
(i) Data that flows continuously at high speed, such as sensor readings. (ii) Analysing data as soon as it arrives, so that the decision is immediate. (iii) Examining data to find patterns and insights for decisions. (iv) Data with no fixed format, such as images and videos.
Write four challenges (disadvantages) of Big Data and one way to handle each.
(1) Privacy: take consent and hide personal details. (2) Security: encrypt and control access. (3) Data quality: clean and verify the data. (4) Cost and skill: use cloud services and train staff.
Compare structured, semi-structured and unstructured data on format, example, storage and ease of analysis.
Structured: fixed rows and columns; marks table; relational database; easy. Semi-structured: tags or keys; JSON, XML; NoSQL stores; moderate. Unstructured: no format; videos, images; data lakes and file systems; difficult, needs AI tools.
Explain sigmoid, tanh and ReLU activation functions with the range of each.
Sigmoid: squeezes any input to a value between 0 and 1; used for probability in the output layer. tanh: gives values between -1 and 1; centred at zero. ReLU: gives 0 for a negative input and the same value for a positive input; simple and fast, most used in hidden layers.
What is a loss function? Explain how gradient descent uses it to train a network.
A loss function measures the error between the predicted and the actual output (MSE for regression, log loss for classification). Gradient descent finds in which direction each weight should change to reduce the loss, and changes it by a small step set by the learning rate. This is repeated for many epochs until the loss is minimum.
Write two positive impacts and two concerns of neural networks for society, with examples.
Positive: (1) early and accurate diagnosis from medical images; (2) voice assistants and translation that help people. Concerns: (1) bias in decisions such as hiring or loans; (2) misuse for deepfakes and loss of some jobs. Their decisions are also hard to explain.
Explain the working of generative AI in simple steps, with an example.
(1) A model is trained on a very large data set, such as millions of images. (2) It learns the patterns and structure of that data. (3) Given a prompt, it produces new content that follows those patterns. (4) The output is refined with feedback. Example: a prompt 'a red kite in a blue sky' gives a new image.
What is a Large Language Model? Explain how it is used in a chatbot and write two precautions for users.
An LLM is a deep learning model trained on huge text to understand and generate language. In a chatbot, it reads the user's prompt and predicts a suitable reply, word by word. Precautions: verify the facts, because it can hallucinate; do not share personal or secret data.
Explain four benefits of generative AI in education, business, design and health care.
Education: explanations and practice questions for each learner. Business: drafts of e-mails, reports and customer replies. Design: quick images, logos and layouts from prompts. Health care: help in studying new medicines and making synthetic medical images for training.
Differentiate between a GAN and a VAE in four points.
(1) Parts: generator and discriminator / encoder and decoder. (2) Working: two networks compete / data is compressed to a code and rebuilt. (3) Output: very sharp, realistic images / smoother, slightly blurred images. (4) Training: harder to train / more stable.
A fake audio clip of a school principal announcing a holiday goes viral. Discuss the issue, its effects and what students should do.
It is a deepfake: an AI-made fake audio. Effects: confusion, loss of trust, harm to the person's image. Students should check with the official school notice or teacher, not forward the clip, and report it. Making such clips is an offence.
Explain the legal and social issues of generative AI: copyright, privacy, jobs and misinformation.
Copyright: models are trained on others' work; ownership of the output is unclear. Privacy: personal data may appear in training data or outputs. Jobs: some writing and design work may reduce, while new jobs come up. Misinformation: fake news and deepfakes spread fast.
A shopkeeper has two years of monthly sales data. Explain how she can turn it into a data story.
Collect and organise the data month-wise. Draw a line graph to find the pattern, for example a fall every summer. Find the reason by comparing with holidays and stock. Build the narrative: background, the falling sales (conflict), the main insight (climax), and the solution, such as summer offers (resolution). Present it in simple words to the staff.
Explain with examples how choosing the wrong chart can mislead the audience.
A pie chart with too many slices hides small differences. A bar chart whose axis does not start at zero makes a small change look big. A 3D chart distorts sizes. The right chart, with a full scale and clear labels, shows the truth.
Why do we need data storytelling? Write four points.
(1) Large data is hard to understand; a story makes it simple. (2) Stories are remembered longer than numbers. (3) A story connects with emotion and persuades people to act. (4) It brings out the key insight for decision makers.
Compare a word cloud, a heat map, a line graph and a candlestick chart: what each shows and when it is used.
Word cloud: frequent words appear bigger; text data such as reviews. Heat map: values as shades of colour; comparing categories. Line graph: points joined by a line; trend over time. Candlestick chart: opening, closing, high and low prices; stock data.
Write a short data story plan on 'screen time of Class XII students' using data, narrative and visuals.
Data: a survey of daily screen hours and test marks. Visuals: a bar chart of screen hours and a scatter plot of screen time against marks. Narrative: students use screens for five hours on average; marks fall as screen time rises; the story ends with a suggestion to limit screen time before exams.
Explain the falling action and the conclusion of a data story with an example.
Falling action follows the main insight: it gives supporting facts and what the insight means. The conclusion is the last part: the problem is resolved and the action is suggested. Example: after showing that most road accidents occur at night (climax), the story shows the low lighting on those roads (falling action) and recommends street lights (conclusion).
The confusion matrix of a disease-screening AI, tested on 50 cases, is given below. Calculate:
Confusion matrix
| Reality: has the disease | Reality: is healthy | |
|---|---|---|
| Predicted: has the disease | 21TP | 14FP |
| Predicted: is healthy | 7FN | 8TN |
Accuracy = (TP + TN) / Total = (21 + 8) / 50 = 29 / 50 = 0.58 (58%)
Precision = TP / (TP + FP) = 21 / (21 + 14) = 21 / 35 = 0.60 (60%)
Recall = TP / (TP + FN) = 21 / (21 + 7) = 21 / 28 = 0.75 (75%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.60 x 0.75 / (0.60 + 0.75) = 0.9 / 1.35 = 0.67
A regression model predicts the marks of students. The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 75 | 74 | 61 | 49 |
| Predicted | 77 | 76 | 63 | 51 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 75 | 77 | -2 |
| 2 | 74 | 76 | -2 |
| 3 | 61 | 63 | -2 |
| 4 | 49 | 51 | -2 |
Squared errors: (-2)2 = 4, (-2)2 = 4, (-2)2 = 4, (-2)2 = 4
MSE = sum of squared errors / number of cases = (4 + 4 + 4 + 4) / 4 = 16 / 4 = 4
RMSE = square root of MSE = square root of 4 = 2
A spam filter was tested on 50 cases. It predicted 'spam' for 30 cases; out of these, 18 were really 'spam'. It predicted 'genuine mail' for the other 20 cases; out of these, 13 were really 'genuine mail'.
TP = 18, FP = 30 - 18 = 12, TN = 13, FN = 20 - 13 = 7.
Confusion matrix
| Reality: spam | Reality: genuine mail | |
|---|---|---|
| Predicted: spam | 18TP | 12FP |
| Predicted: genuine mail | 7FN | 13TN |
Accuracy = (TP + TN) / Total = (18 + 13) / 50 = 31 / 50 = 0.62 (62%)
Precision = TP / (TP + FP) = 18 / (18 + 12) = 18 / 30 = 0.60 (60%)
Recall = TP / (TP + FN) = 18 / (18 + 7) = 18 / 25 = 0.72 (72%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.60 x 0.72 / (0.60 + 0.72) = 0.864 / 1.32 = 0.65
Precision, because a genuine mail sent to the spam folder (False Positive) may be an important message.
A regression model predicts the price of houses (in lakh rupees). The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 35 | 61 | 42 | 45 |
| Predicted | 35 | 61 | 42 | 49 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 35 | 35 | 0 |
| 2 | 61 | 61 | 0 |
| 3 | 42 | 42 | 0 |
| 4 | 45 | 49 | -4 |
Squared errors: (0)2 = 0, (0)2 = 0, (0)2 = 0, (-4)2 = 16
MSE = sum of squared errors / number of cases = (0 + 0 + 0 + 16) / 4 = 16 / 4 = 4
RMSE = square root of MSE = square root of 4 = 2
The confusion matrix of a forest-fire alert system, tested on 1000 cases, is given below. Calculate:
Confusion matrix
| Reality: fire | Reality: no fire | |
|---|---|---|
| Predicted: fire | 240TP | 80FP |
| Predicted: no fire | 260FN | 420TN |
Accuracy = (TP + TN) / Total = (240 + 420) / 1000 = 660 / 1000 = 0.66 (66%)
Precision = TP / (TP + FP) = 240 / (240 + 80) = 240 / 320 = 0.75 (75%)
Recall = TP / (TP + FN) = 240 / (240 + 260) = 240 / 500 = 0.48 (48%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.75 x 0.48 / (0.75 + 0.48) = 0.72 / 1.23 = 0.59
A regression model predicts the daily temperature (in Celsius). The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 36 | 28 | 22 | 20 |
| Predicted | 30 | 28 | 22 | 20 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 36 | 30 | 6 |
| 2 | 28 | 28 | 0 |
| 3 | 22 | 22 | 0 |
| 4 | 20 | 20 | 0 |
Squared errors: (6)2 = 36, (0)2 = 0, (0)2 = 0, (0)2 = 0
MSE = sum of squared errors / number of cases = (36 + 0 + 0 + 0) / 4 = 36 / 4 = 9
RMSE = square root of MSE = square root of 9 = 3
A bank's fraud detector was tested on 1000 cases. It predicted 'fraud' for 520 cases; out of these, 260 were really 'fraud'. It predicted 'genuine transaction' for the other 480 cases; out of these, 240 were really 'genuine transaction'.
TP = 260, FP = 520 - 260 = 260, TN = 240, FN = 480 - 240 = 240.
Confusion matrix
| Reality: fraud | Reality: genuine transaction | |
|---|---|---|
| Predicted: fraud | 260TP | 260FP |
| Predicted: genuine transaction | 240FN | 240TN |
Accuracy = (TP + TN) / Total = (260 + 240) / 1000 = 500 / 1000 = 0.50 (50%)
Precision = TP / (TP + FP) = 260 / (260 + 260) = 260 / 520 = 0.50 (50%)
Recall = TP / (TP + FN) = 260 / (260 + 240) = 260 / 500 = 0.52 (52%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.50 x 0.52 / (0.50 + 0.52) = 0.52 / 1.02 = 0.51
Recall, because a fraud that is missed (False Negative) causes a loss of money.
A regression model predicts the sales of a shop (in thousand rupees). The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 131 | 94 | 121 | 131 |
| Predicted | 128 | 91 | 118 | 128 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 131 | 128 | 3 |
| 2 | 94 | 91 | 3 |
| 3 | 121 | 118 | 3 |
| 4 | 131 | 128 | 3 |
Squared errors: (3)2 = 9, (3)2 = 9, (3)2 = 9, (3)2 = 9
MSE = sum of squared errors / number of cases = (9 + 9 + 9 + 9) / 4 = 36 / 4 = 9
RMSE = square root of MSE = square root of 9 = 3
The confusion matrix of a plagiarism checker, tested on 200 cases, is given below. Calculate:
Confusion matrix
| Reality: copied work | Reality: original work | |
|---|---|---|
| Predicted: copied work | 32TP | 48FP |
| Predicted: original work | 8FN | 112TN |
Accuracy = (TP + TN) / Total = (32 + 112) / 200 = 144 / 200 = 0.72 (72%)
Precision = TP / (TP + FP) = 32 / (32 + 48) = 32 / 80 = 0.40 (40%)
Recall = TP / (TP + FN) = 32 / (32 + 8) = 32 / 40 = 0.80 (80%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.40 x 0.80 / (0.40 + 0.80) = 0.64 / 1.20 = 0.53
A regression model predicts the runs scored by a batter. The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 52 | 70 | 22 | 39 |
| Predicted | 52 | 64 | 22 | 31 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 52 | 52 | 0 |
| 2 | 70 | 64 | 6 |
| 3 | 22 | 22 | 0 |
| 4 | 39 | 31 | 8 |
Squared errors: (0)2 = 0, (6)2 = 36, (0)2 = 0, (8)2 = 64
MSE = sum of squared errors / number of cases = (0 + 36 + 0 + 64) / 4 = 100 / 4 = 25
RMSE = square root of MSE = square root of 25 = 5
A face-unlock system (positive = owner) was tested on 500 cases. It predicted 'owner' for 250 cases; out of these, 150 were really 'owner'. It predicted 'stranger' for the other 250 cases; out of these, 200 were really 'stranger'.
TP = 150, FP = 250 - 150 = 100, TN = 200, FN = 250 - 200 = 50.
Confusion matrix
| Reality: owner | Reality: stranger | |
|---|---|---|
| Predicted: owner | 150TP | 100FP |
| Predicted: stranger | 50FN | 200TN |
Accuracy = (TP + TN) / Total = (150 + 200) / 500 = 350 / 500 = 0.70 (70%)
Precision = TP / (TP + FP) = 150 / (150 + 100) = 150 / 250 = 0.60 (60%)
Recall = TP / (TP + FN) = 150 / (150 + 50) = 150 / 200 = 0.75 (75%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.60 x 0.75 / (0.60 + 0.75) = 0.9 / 1.35 = 0.67
Precision, because a stranger accepted as the owner (False Positive) can open the phone.
A regression model predicts the rainfall (in mm). The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 157 | 93 | 83 | 121 |
| Predicted | 158 | 94 | 78 | 118 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 157 | 158 | -1 |
| 2 | 93 | 94 | -1 |
| 3 | 83 | 78 | 5 |
| 4 | 121 | 118 | 3 |
Squared errors: (-1)2 = 1, (-1)2 = 1, (5)2 = 25, (3)2 = 9
MSE = sum of squared errors / number of cases = (1 + 1 + 25 + 9) / 4 = 36 / 4 = 9
RMSE = square root of MSE = square root of 9 = 3
The confusion matrix of a flood-warning model, tested on 1000 cases, is given below. Calculate:
Confusion matrix
| Reality: flood | Reality: no flood | |
|---|---|---|
| Predicted: flood | 260TP | 140FP |
| Predicted: no flood | 260FN | 340TN |
Accuracy = (TP + TN) / Total = (260 + 340) / 1000 = 600 / 1000 = 0.60 (60%)
Precision = TP / (TP + FP) = 260 / (260 + 140) = 260 / 400 = 0.65 (65%)
Recall = TP / (TP + FN) = 260 / (260 + 260) = 260 / 520 = 0.50 (50%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.65 x 0.50 / (0.65 + 0.50) = 0.65 / 1.15 = 0.57
A regression model predicts the height of plants (in cm). The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 51 | 28 | 79 | 71 |
| Predicted | 48 | 19 | 76 | 70 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 51 | 48 | 3 |
| 2 | 28 | 19 | 9 |
| 3 | 79 | 76 | 3 |
| 4 | 71 | 70 | 1 |
Squared errors: (3)2 = 9, (9)2 = 81, (3)2 = 9, (1)2 = 1
MSE = sum of squared errors / number of cases = (9 + 81 + 9 + 1) / 4 = 100 / 4 = 25
RMSE = square root of MSE = square root of 25 = 5
An AI that finds pneumonia in chest X-rays was tested on 50 cases. It predicted 'pneumonia' for 25 cases; out of these, 24 were really 'pneumonia'. It predicted 'normal' for the other 25 cases; out of these, 19 were really 'normal'.
TP = 24, FP = 25 - 24 = 1, TN = 19, FN = 25 - 19 = 6.
Confusion matrix
| Reality: pneumonia | Reality: normal | |
|---|---|---|
| Predicted: pneumonia | 24TP | 1FP |
| Predicted: normal | 6FN | 19TN |
Accuracy = (TP + TN) / Total = (24 + 19) / 50 = 43 / 50 = 0.86 (86%)
Precision = TP / (TP + FP) = 24 / (24 + 1) = 24 / 25 = 0.96 (96%)
Recall = TP / (TP + FN) = 24 / (24 + 6) = 24 / 30 = 0.80 (80%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.96 x 0.80 / (0.96 + 0.80) = 1.536 / 1.76 = 0.87
Recall, because a patient whose pneumonia is missed (False Negative) goes untreated.
A regression model predicts the weight of parcels (in kg). The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 11 | 46 | 53 | 57 |
| Predicted | 6 | 53 | 48 | 56 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 11 | 6 | 5 |
| 2 | 46 | 53 | -7 |
| 3 | 53 | 48 | 5 |
| 4 | 57 | 56 | 1 |
Squared errors: (5)2 = 25, (-7)2 = 49, (5)2 = 25, (1)2 = 1
MSE = sum of squared errors / number of cases = (25 + 49 + 25 + 1) / 4 = 100 / 4 = 25
RMSE = square root of MSE = square root of 25 = 5
The confusion matrix of a filter that blocks unsafe videos for children, tested on 100 cases, is given below. Calculate:
Confusion matrix
| Reality: unsafe video | Reality: safe video | |
|---|---|---|
| Predicted: unsafe video | 24TP | 6FP |
| Predicted: safe video | 16FN | 54TN |
Accuracy = (TP + TN) / Total = (24 + 54) / 100 = 78 / 100 = 0.78 (78%)
Precision = TP / (TP + FP) = 24 / (24 + 6) = 24 / 30 = 0.80 (80%)
Recall = TP / (TP + FN) = 24 / (24 + 16) = 24 / 40 = 0.60 (60%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.80 x 0.60 / (0.80 + 0.60) = 0.96 / 1.40 = 0.69
A regression model predicts the units of electricity used. The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 115 | 149 | 110 | 117 |
| Predicted | 116 | 148 | 103 | 110 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 115 | 116 | -1 |
| 2 | 149 | 148 | 1 |
| 3 | 110 | 103 | 7 |
| 4 | 117 | 110 | 7 |
Squared errors: (-1)2 = 1, (1)2 = 1, (7)2 = 49, (7)2 = 49
MSE = sum of squared errors / number of cases = (1 + 1 + 49 + 49) / 4 = 100 / 4 = 25
RMSE = square root of MSE = square root of 25 = 5
A camera that finds defective bottles in a factory was tested on 200 cases. It predicted 'defective bottle' for 120 cases; out of these, 84 were really 'defective bottle'. It predicted 'good bottle' for the other 80 cases; out of these, 24 were really 'good bottle'.
TP = 84, FP = 120 - 84 = 36, TN = 24, FN = 80 - 24 = 56.
Confusion matrix
| Reality: defective bottle | Reality: good bottle | |
|---|---|---|
| Predicted: defective bottle | 84TP | 36FP |
| Predicted: good bottle | 56FN | 24TN |
Accuracy = (TP + TN) / Total = (84 + 24) / 200 = 108 / 200 = 0.54 (54%)
Precision = TP / (TP + FP) = 84 / (84 + 36) = 84 / 120 = 0.70 (70%)
Recall = TP / (TP + FN) = 84 / (84 + 56) = 84 / 140 = 0.60 (60%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.70 x 0.60 / (0.70 + 0.60) = 0.84 / 1.30 = 0.65
Recall, because a defective bottle that is passed (False Negative) reaches the customer.
A regression model predicts the time taken for delivery (in minutes). The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 36 | 50 | 47 | 27 |
| Predicted | 40 | 46 | 43 | 31 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 36 | 40 | -4 |
| 2 | 50 | 46 | 4 |
| 3 | 47 | 43 | 4 |
| 4 | 27 | 31 | -4 |
Squared errors: (-4)2 = 16, (4)2 = 16, (4)2 = 16, (-4)2 = 16
MSE = sum of squared errors / number of cases = (16 + 16 + 16 + 16) / 4 = 64 / 4 = 16
RMSE = square root of MSE = square root of 16 = 4
The confusion matrix of a disease-screening AI, tested on 1000 cases, is given below. Calculate:
Confusion matrix
| Reality: has the disease | Reality: is healthy | |
|---|---|---|
| Predicted: has the disease | 120TP | 40FP |
| Predicted: is healthy | 180FN | 660TN |
Accuracy = (TP + TN) / Total = (120 + 660) / 1000 = 780 / 1000 = 0.78 (78%)
Precision = TP / (TP + FP) = 120 / (120 + 40) = 120 / 160 = 0.75 (75%)
Recall = TP / (TP + FN) = 120 / (120 + 180) = 120 / 300 = 0.40 (40%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.75 x 0.40 / (0.75 + 0.40) = 0.6 / 1.15 = 0.52
A regression model predicts the marks of students. The actual and the predicted values of 3 cases are given below.
| Case | 1 | 2 | 3 |
|---|---|---|---|
| Actual | 71 | 88 | 41 |
| Predicted | 70 | 83 | 42 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 71 | 70 | 1 |
| 2 | 88 | 83 | 5 |
| 3 | 41 | 42 | -1 |
Squared errors: (1)2 = 1, (5)2 = 25, (-1)2 = 1
MSE = sum of squared errors / number of cases = (1 + 25 + 1) / 3 = 27 / 3 = 9
RMSE = square root of MSE = square root of 9 = 3
A spam filter was tested on 200 cases. It predicted 'spam' for 112 cases; out of these, 84 were really 'spam'. It predicted 'genuine mail' for the other 88 cases; out of these, 32 were really 'genuine mail'.
TP = 84, FP = 112 - 84 = 28, TN = 32, FN = 88 - 32 = 56.
Confusion matrix
| Reality: spam | Reality: genuine mail | |
|---|---|---|
| Predicted: spam | 84TP | 28FP |
| Predicted: genuine mail | 56FN | 32TN |
Accuracy = (TP + TN) / Total = (84 + 32) / 200 = 116 / 200 = 0.58 (58%)
Precision = TP / (TP + FP) = 84 / (84 + 28) = 84 / 112 = 0.75 (75%)
Recall = TP / (TP + FN) = 84 / (84 + 56) = 84 / 140 = 0.60 (60%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.75 x 0.60 / (0.75 + 0.60) = 0.9 / 1.35 = 0.67
Precision, because a genuine mail sent to the spam folder (False Positive) may be an important message.
A regression model predicts the price of houses (in lakh rupees). The actual and the predicted values of 3 cases are given below.
| Case | 1 | 2 | 3 |
|---|---|---|---|
| Actual | 34 | 64 | 41 |
| Predicted | 41 | 65 | 36 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 34 | 41 | -7 |
| 2 | 64 | 65 | -1 |
| 3 | 41 | 36 | 5 |
Squared errors: (-7)2 = 49, (-1)2 = 1, (5)2 = 25
MSE = sum of squared errors / number of cases = (49 + 1 + 25) / 3 = 75 / 3 = 25
RMSE = square root of MSE = square root of 25 = 5
The confusion matrix of a forest-fire alert system, tested on 100 cases, is given below. Calculate:
Confusion matrix
| Reality: fire | Reality: no fire | |
|---|---|---|
| Predicted: fire | 36TP | 24FP |
| Predicted: no fire | 14FN | 26TN |
Accuracy = (TP + TN) / Total = (36 + 26) / 100 = 62 / 100 = 0.62 (62%)
Precision = TP / (TP + FP) = 36 / (36 + 24) = 36 / 60 = 0.60 (60%)
Recall = TP / (TP + FN) = 36 / (36 + 14) = 36 / 50 = 0.72 (72%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.60 x 0.72 / (0.60 + 0.72) = 0.864 / 1.32 = 0.65
A regression model predicts the daily temperature (in Celsius). The actual and the predicted values of 3 cases are given below.
| Case | 1 | 2 | 3 |
|---|---|---|---|
| Actual | 27 | 31 | 27 |
| Predicted | 25 | 41 | 29 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 27 | 25 | 2 |
| 2 | 31 | 41 | -10 |
| 3 | 27 | 29 | -2 |
Squared errors: (2)2 = 4, (-10)2 = 100, (-2)2 = 4
MSE = sum of squared errors / number of cases = (4 + 100 + 4) / 3 = 108 / 3 = 36
RMSE = square root of MSE = square root of 36 = 6
A bank's fraud detector was tested on 100 cases. It predicted 'fraud' for 30 cases; out of these, 24 were really 'fraud'. It predicted 'genuine transaction' for the other 70 cases; out of these, 54 were really 'genuine transaction'.
TP = 24, FP = 30 - 24 = 6, TN = 54, FN = 70 - 54 = 16.
Confusion matrix
| Reality: fraud | Reality: genuine transaction | |
|---|---|---|
| Predicted: fraud | 24TP | 6FP |
| Predicted: genuine transaction | 16FN | 54TN |
Accuracy = (TP + TN) / Total = (24 + 54) / 100 = 78 / 100 = 0.78 (78%)
Precision = TP / (TP + FP) = 24 / (24 + 6) = 24 / 30 = 0.80 (80%)
Recall = TP / (TP + FN) = 24 / (24 + 16) = 24 / 40 = 0.60 (60%)
F1 score = 2 x Precision x Recall / (Precision + Recall) = 2 x 0.80 x 0.60 / (0.80 + 0.60) = 0.96 / 1.40 = 0.69
Recall, because a fraud that is missed (False Negative) causes a loss of money.
A regression model predicts the sales of a shop (in thousand rupees). The actual and the predicted values of 3 cases are given below.
| Case | 1 | 2 | 3 |
|---|---|---|---|
| Actual | 115 | 50 | 92 |
| Predicted | 113 | 48 | 90 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 115 | 113 | 2 |
| 2 | 50 | 48 | 2 |
| 3 | 92 | 90 | 2 |
Squared errors: (2)2 = 4, (2)2 = 4, (2)2 = 4
MSE = sum of squared errors / number of cases = (4 + 4 + 4) / 3 = 12 / 3 = 4
RMSE = square root of MSE = square root of 4 = 2
A regression model predicts the runs scored by a batter. The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 36 | 21 | 48 | 74 |
| Predicted | 38 | 23 | 50 | 76 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 36 | 38 | -2 |
| 2 | 21 | 23 | -2 |
| 3 | 48 | 50 | -2 |
| 4 | 74 | 76 | -2 |
Squared errors: (-2)2 = 4, (-2)2 = 4, (-2)2 = 4, (-2)2 = 4
MSE = sum of squared errors / number of cases = (4 + 4 + 4 + 4) / 4 = 16 / 4 = 4
RMSE = square root of MSE = square root of 4 = 2
A regression model predicts the rainfall (in mm). The actual and the predicted values of 4 cases are given below.
| Case | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Actual | 64 | 173 | 193 | 69 |
| Predicted | 64 | 177 | 193 | 69 |
| Case | Actual | Predicted | Error |
|---|---|---|---|
| 1 | 64 | 64 | 0 |
| 2 | 173 | 177 | -4 |
| 3 | 193 | 193 | 0 |
| 4 | 69 | 69 | 0 |
Squared errors: (0)2 = 0, (-4)2 = 16, (0)2 = 0, (0)2 = 0
MSE = sum of squared errors / number of cases = (0 + 16 + 0 + 0) / 4 = 16 / 4 = 4
RMSE = square root of MSE = square root of 4 = 2
A single neuron (perceptron) is used to decide: "Should Riya go to the park?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| The weather is good | 3 | Yes (1) | No (0) |
| Her friend is coming | 5 | No (0) | Yes (1) |
| Her homework is done | 1 | No (0) | Yes (1) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (1 x 3) + (0 x 5) + (0 x 1) + (-1) = 2. It is less than 4, so the output is 0 (No).
Sum = (0 x 3) + (1 x 5) + (1 x 1) + (-1) = 5. It is not less than 4, so the output is 1 (Yes).
Her friend is coming, because it has the largest weight (5). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should Arjun buy the phone?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -1. The output is 1 (Yes) if the weighted sum is 3 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| The price is within budget | 2 | No (0) | Yes (1) |
| The camera is good | 1 | Yes (1) | No (0) |
| The reviews are good | 5 | Yes (1) | No (0) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (0 x 2) + (1 x 1) + (1 x 5) + (-1) = 5. It is not less than 3, so the output is 1 (Yes).
Sum = (1 x 2) + (0 x 1) + (0 x 5) + (-1) = 1. It is less than 3, so the output is 0 (No).
The reviews are good, because it has the largest weight (5). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should the school team play the match?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| The ground is dry | 3 | Yes (1) | Yes (1) |
| All the players are fit | 5 | Yes (1) | No (0) |
| The coach is present | 1 | Yes (1) | No (0) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (1 x 3) + (1 x 5) + (1 x 1) + (-1) = 8. It is not less than 4, so the output is 1 (Yes).
Sum = (1 x 3) + (0 x 5) + (0 x 1) + (-1) = 2. It is less than 4, so the output is 0 (No).
All the players are fit, because it has the largest weight (5). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should Meena watch the film?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is 1. The output is 1 (Yes) if the weighted sum is 6 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| Her exam is over | 1 | Yes (1) | No (0) |
| Her friends are going | 5 | No (0) | Yes (1) |
| The ticket is cheap | 4 | Yes (1) | No (0) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (1 x 1) + (0 x 5) + (1 x 4) + (1) = 6. It is not less than 6, so the output is 1 (Yes).
Sum = (0 x 1) + (1 x 5) + (0 x 4) + (1) = 6. It is not less than 6, so the output is 1 (Yes).
Her friends are going, because it has the largest weight (5). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should the farmer water the field today?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -2. The output is 1 (Yes) if the weighted sum is 3 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| The soil is dry | 1 | Yes (1) | No (0) |
| No rain is forecast | 5 | No (0) | Yes (1) |
| The pump has power | 3 | No (0) | Yes (1) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (1 x 1) + (0 x 5) + (0 x 3) + (-2) = -1. It is less than 3, so the output is 0 (No).
Sum = (0 x 1) + (1 x 5) + (1 x 3) + (-2) = 6. It is not less than 3, so the output is 1 (Yes).
No rain is forecast, because it has the largest weight (5). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should Kabir join the coding club?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is 1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| He is free on that day | 3 | Yes (1) | No (0) |
| He likes coding | 1 | No (0) | Yes (1) |
| His friend has joined | 2 | No (0) | Yes (1) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (1 x 3) + (0 x 1) + (0 x 2) + (1) = 4. It is not less than 4, so the output is 1 (Yes).
Sum = (0 x 3) + (1 x 1) + (1 x 2) + (1) = 4. It is not less than 4, so the output is 1 (Yes).
He is free on that day, because it has the largest weight (3). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should the family go on a picnic?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| It is a holiday | 2 | No (0) | Yes (1) |
| The weather is clear | 4 | Yes (1) | No (0) |
| The car is available | 1 | Yes (1) | No (0) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (0 x 2) + (1 x 4) + (1 x 1) + (-1) = 4. It is not less than 4, so the output is 1 (Yes).
Sum = (1 x 2) + (0 x 4) + (0 x 1) + (-1) = 1. It is less than 4, so the output is 0 (No).
The weather is clear, because it has the largest weight (4). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should Sana order food online?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is 1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| She is very hungry | 4 | Yes (1) | No (0) |
| There is a discount | 1 | Yes (1) | No (0) |
| Delivery is fast | 2 | No (0) | Yes (1) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (1 x 4) + (1 x 1) + (0 x 2) + (1) = 6. It is not less than 4, so the output is 1 (Yes).
Sum = (0 x 4) + (0 x 1) + (1 x 2) + (1) = 3. It is less than 4, so the output is 0 (No).
She is very hungry, because it has the largest weight (4). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should the shop open on Sunday?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| Many customers are expected | 2 | Yes (1) | No (0) |
| Staff is available | 3 | Yes (1) | No (0) |
| It is a festival week | 4 | No (0) | Yes (1) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (1 x 2) + (1 x 3) + (0 x 4) + (-1) = 4. It is not less than 4, so the output is 1 (Yes).
Sum = (0 x 2) + (0 x 3) + (1 x 4) + (-1) = 3. It is less than 4, so the output is 0 (No).
It is a festival week, because it has the largest weight (4). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should Dev go for the morning run?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| He slept well | 1 | No (0) | Yes (1) |
| It is not raining | 4 | No (0) | Yes (1) |
| His shoes are dry | 5 | Yes (1) | No (0) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (0 x 1) + (0 x 4) + (1 x 5) + (-1) = 4. It is not less than 4, so the output is 1 (Yes).
Sum = (1 x 1) + (1 x 4) + (0 x 5) + (-1) = 4. It is not less than 4, so the output is 1 (Yes).
His shoes are dry, because it has the largest weight (5). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should the class go on the science trip?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| Parents have agreed | 4 | No (0) | Yes (1) |
| The bus is booked | 5 | Yes (1) | No (0) |
| The museum is open | 1 | Yes (1) | No (0) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (0 x 4) + (1 x 5) + (1 x 1) + (-1) = 5. It is not less than 4, so the output is 1 (Yes).
Sum = (1 x 4) + (0 x 5) + (0 x 1) + (-1) = 3. It is less than 4, so the output is 0 (No).
The bus is booked, because it has the largest weight (5). A larger weight means the input has more effect on the decision.
A single neuron (perceptron) is used to decide: "Should Tara adopt the puppy?" Each factor is an input: Yes = 1, No = 0. The weights are given in the table. The bias is -1. The output is 1 (Yes) if the weighted sum is 4 or more, otherwise 0 (No).
| Factor (input) | Weight | Day 1 | Day 2 |
|---|---|---|---|
| She has time for it | 2 | Yes (1) | No (0) |
| Her parents agree | 5 | No (0) | Yes (1) |
| There is space at home | 1 | Yes (1) | No (0) |
Weighted sum = (x1 x w1) + (x2 x w2) + (x3 x w3) + bias
Sum = (1 x 2) + (0 x 5) + (1 x 1) + (-1) = 2. It is less than 4, so the output is 0 (No).
Sum = (0 x 2) + (1 x 5) + (0 x 1) + (-1) = 4. It is not less than 4, so the output is 1 (Yes).
Her parents agree, because it has the largest weight (5). A larger weight means the input has more effect on the decision.
Explain the five stages of active listening with an example.
Receiving - hearing the message ('take one tablet twice a day'); Understanding - interpreting it (morning and night); Remembering - storing it (note in the phone); Evaluating - judging it ('before or after food?'); Responding - feedback (asking the doctor, nodding).
Explain RESPECT, the guide to active listening.
R - Remove distractions; E - Eye contact; S - Show you are listening by nodding; P - Pay attention; E - Empathise with the speaker; C - Clarify doubts by asking or paraphrasing; T - Tune yourself to the speaker's pace and tone.
Write any four barriers to active listening and how to overcome them.
Distractions - keep the phone away; prejudice - keep an open mind; interrupting - let the speaker finish; daydreaming - focus and take notes; preparing your reply too early - listen fully first; information overload - ask for key points.
Change the voice:
(i) The meal was cooked by the chef. (ii) The wall is being painted by him. (iii) They will open the shop. (iv) Reports are written by her.
Explain the five types of nouns with examples.
Proper - India, Ravi; Common - city, boy; Collective - team, herd; Abstract - honesty, courage; Material - gold, wood, cotton.
Explain Maslow's hierarchy of needs.
From the base: physiological needs (food, water, sleep); safety (security, job, health); love and belonging (family, friends); esteem (respect, recognition); self-actualisation (reaching full potential). Lower needs must be met before higher needs motivate us.
What is result orientation? Explain its steps with an example.
Result orientation is focusing on achieving the desired outcome. Steps: set a clear goal, make an action plan, use the right tools, work within the time, measure the result and improve. Example: a student aiming for 90 in AI plans chapters weekly, practises questions and tracks sample-paper scores.
Explain the Big Five personality factors with workplace examples.
Openness - a designer trying new styles; Conscientiousness - an accountant meeting deadlines; Extraversion - a salesperson who loves meeting people; Agreeableness - a caring nurse; Neuroticism - low neuroticism helps an air-traffic controller stay calm.
Write any four ways to maintain a positive attitude.
Practise gratitude; use positive self-talk; keep positive company; follow a healthy routine of exercise, sleep and food; focus on solutions; help others. (Any four.)
What is self-awareness? Explain its aspects and ways to develop it.
Self-awareness is a clear understanding of one's personality, strengths, weaknesses, emotions and values. Aspects: strengths, weaknesses, emotions and their triggers, values. Ways: reflection and journaling, feedback from others, personality tests, meditation and new experiences.
Write the result of each formula if A1 = 20, A2 = 30, A3 = 50: =SUM(A1:A3), =AVERAGE(A1:A3), =MAX(A1:A3), =MIN(A1:A3).
=SUM gives 100; =AVERAGE gives 33.33; =MAX gives 50; =MIN gives 20.
Differentiate between relative, absolute and mixed cell references.
Relative (A1) changes when the formula is copied; absolute ($A$1) never changes; mixed ($A1 or A$1) keeps either the column or the row fixed. Absolute references are used for fixed values like a tax rate.
Differentiate between sorting and filtering and write the steps for each.
Sorting arranges all rows in order: select data > Data > Sort > choose column and order. Filtering shows only rows meeting a condition and hides others: Data > AutoFilter > click the arrow > choose the condition. Filtering does not delete data.
Name four chart types and when to use each.
Column/bar - compare items; line - trend over time; pie - parts of a whole; scatter - relation between two variables.
Explain any four features of presentation software.
Slide layouts - title, content, two content; slide master - same design on all slides; transitions - effects between slides; animations - effects on objects; speaker notes; slide sorter view; F5 to start the slide show.
Explain any four types of entrepreneurs with examples.
Service - salon, courier; Business/trading - grocery, wholesale; Industrial - textile mill, food processing; Agricultural - dairy, poultry; Technical - software or app company; Social - affordable healthcare or solar lighting. (Any four.)
Explain the entrepreneurial competencies with examples.
Decisiveness - chooses a supplier after comparing quotes; Initiative - starts home delivery before competitors; Interpersonal skills - keeps customers and staff happy; Perseverance - relaunches after a failed product; Organisational skills - manages stock, cash and staff schedules.
What is a business plan? Explain its main parts.
A business plan is a written document describing the business idea, market, operations and finances. Parts: executive summary, business description, market analysis, products and services, marketing plan (4 Ps), operations plan and financial plan.
A candle maker has fixed costs of Rs 15,000 a month. Each candle sells for Rs 40 and costs Rs 25 to make. Find the break-even point. What happens if she sells 1,200 candles?
Break-even = 15,000 / (40 - 25) = 15,000 / 15 = 1,000 candles. At 1,200 candles, profit = 200 x 15 = Rs 3,000 (the contribution of each candle beyond break-even).
Explain the barriers to entrepreneurship and ways to overcome them.
Environmental - poor roads, power, raw material; choose a suitable location. No/faulty business plan; prepare one with market research. Personal - self-doubt, fear of failure; training and mentors. Financial - lack of capital; savings, Mudra loans, investors.
What are green jobs? Explain any four fields of green jobs.
Green jobs protect or restore the environment. Agriculture - organic farming advisor; transport - EV technician; renewable energy - solar PV installer; water conservation - rainwater harvesting technician; waste management - recycling supervisor; green construction - green building consultant.
Explain the benefits of green jobs.
They limit greenhouse-gas emissions, reduce waste and pollution, protect and restore ecosystems, use energy and raw materials efficiently, help communities adapt to climate change, and create decent employment.
Explain how climate change happens and name the main greenhouse gases.
Burning fossil fuels and cutting forests increase greenhouse gases, which trap more heat, causing global warming and climate change - floods, droughts, heatwaves and sea-level rise. Main gases: carbon dioxide, methane, nitrous oxide, water vapour and fluorinated gases.
Write any four missions of the NAPCC and their focus.
National Solar Mission - solar energy; Enhanced Energy Efficiency - saving energy; Sustainable Habitat - green cities and transport; National Water Mission - water conservation; Green India - afforestation; Sustainable Agriculture - climate-resilient farming.
Explain the different levels of green skills with examples.
Green citizens - everyone saving energy and segregating waste; green workers - waste collectors and plant operators; green technicians - solar installers and energy auditors; green experts - scientists and planners who design green policies and technologies.
Explain the stages of the computer vision process with the example of number-plate recognition.
Image acquisition - toll camera captures the car; pre-processing - resize, remove noise, adjust contrast, grayscale; feature extraction - find edges and characters; detection/segmentation - cut out the plate region; high-level processing - read characters and charge the toll.
Differentiate between human vision and computer vision.
Human vision captures light with eyes and the brain recognises objects instantly using experience and context, but tires. Computer vision captures pixels with cameras and uses algorithms trained on many labelled images; it works at large scale 24x7 but can fail with unusual lighting or angles.
Explain any four challenges of computer vision.
Lighting changes alter pixel values; occlusion - objects partly hidden; different viewpoints and scale; cluttered backgrounds; the need for huge labelled datasets; bias in datasets; privacy concerns. (Any four.)
Explain any four computer-vision tasks with examples.
Image classification - label the whole image; object detection - boxes around many objects; image segmentation - label every pixel; face recognition - identify a person; OCR - read text; object tracking - follow objects in video; pose estimation - find body joints. (Any four.)
A colour image is 1920 x 1080 pixels. Find the number of pixels and the memory needed at 3 bytes per pixel. Why is pre-processing needed?
Pixels = 1920 x 1080 = 2,073,600. Memory = 3 x 2,073,600 = 6,220,800 bytes, about 6.2 MB. Pre-processing (resizing, grayscale, noise removal, contrast adjustment) reduces size and noise so that features can be found faster and more accurately.
Tap an option to check your answer at once; your score is saved
Ten multiple-choice questions picked at random. Tap an option to answer. You get your score at the end.
Fifteen ready papers with blueprints and a random paper maker, all on the design of the CBSE Sample Question Paper 2026-27
Each paper has 21 questions for 50 marks and 2 hours, with the same choices as the CBSE Sample Question Paper 2026-27: any 4 of 6 in Q1, any 5 of 6 in Q2 to Q5, any 3 of Q6 to Q10, any 4 of Q11 to Q16 and any 3 of Q17 to Q21. Every paper has numericals in Section A, and a confusion matrix or RMSE numerical in Q17 to Q21. No question is repeated: the fifteen papers use 690 different questions.
Downloads and printing work when this page is opened as an HTML file in a browser.
Choose Paper 1 to Paper 15, or make a random paper. The blueprint and the paper open here.
Read this the evening before and again at the exam centre gate