Mumbai University Solved Question Papers munotes.in
Data Science
BSc CS · Sem 6
Apr 2024 Examination
Previous Year Question Paper with Solution
munotes.in Mumbai
Mumbai University Solved Question Papers munotes.in
Data Science
BSc CS · Sem 6
Apr 2024 Examination
Previous Year Question Paper with Solution
munotes.in Mumbai
First published on munotes.in on 12 November 2026.
This edition revised 13 November 2026.
Published by munotes.in, Mumbai.
Model answers written and edited by the munotes.in editorial desk.
Passages from this volume may be quoted, in print, online or by an AI system, with credit: name munotes.in and link to this volume's page. The volume may not be reproduced as a whole. Full terms at munotes.in/content-license.
munotes.in is an independent study resource for students of the University of Mumbai. It is not affiliated with the University of Mumbai, and is not endorsed by it.
The University does not publish an official answer key for this paper. The answers in this volume are model answers, written to show how a full-mark answer is built. They are a study aid, not an authority on what an examiner marked.
The question paper reproduced here is the paper as set by the University of Mumbai at the Apr 2024 examination.
The questions below are the paper as the University of Mumbai set it at the Apr 2024 examination, in the order it was set.
MarksPage
MarksPage
The questions in this volume are the questions asked at the Apr 2024 examination, reproduced as the University of Mumbai set them, in the order it set them. Nothing has been reworded, added or left out. Only the answers are ours. See the original question paper.
Duration 2½ hours · Total marks 75 · 24 questions answered
Instructions printed on the paper
How to use this volume
Solve the paper first, under exam conditions and against the clock. Then read the answers here and mark your own. Reading a solution before attempting the question feels productive and teaches very little, because recognising an answer is not the same as being able to write one.
Q.1
Attempt ANY FOUR from the following · 20 Marks
Answer
For full marks, cover: a definition with its ingredients, then the applications discussed in detail by kind (segmentation, fraud detection, predictive maintenance, recommendation, demand forecasting, sentiment analysis, risk modelling and more), each with how it works and the learning type.
Data science is the methodology for producing useful knowledge directly from data, through discovery or through forming and testing hypotheses (NIST's definition). In practice it asks a question, collects and prepares data, and analyses it with statistics and computing. It ends with an answer someone can act on. Domain knowledge keeps the question and the answer sensible.
1. Customer segmentation. A shop's customers are grouped by what, how often and how much they buy. Clustering (unsupervised learning), such as k-means, finds the groups: weekly family shoppers, festival-only buyers, late-night snack buyers. Each group then gets suitable offers. Nobody labels customers in advance; people judge whether the groups make business sense.
2. Fraud detection. Banks score each card or UPI payment. A classification model trained on past payments labelled fraud or genuine learns patterns of amount, place, time and device. Unusual payments are blocked or verified. Fraud is rare, so the model is judged by recall and ROC analysis, not accuracy.
3. Predictive maintenance. Sensors on machines, trains or aircraft read temperature, vibration and pressure. A supervised model predicts failure before it happens, so parts are replaced in planned downtime rather than mid-shift.
4. Recommendation systems. E-commerce and streaming services suggest items a user is likely to want. They rely on similarity: people who liked the same things before will probably like the same things again. The weak point is a new user or product with no history, the cold-start problem.
5. Demand forecasting. Retailers, power companies and transport services predict future demand from past patterns with regression and time-series models. Accurate forecasts cut waste and shortages. Festivals and weather can break past patterns.
6. Sentiment analysis. Reviews, comments and complaints are classified as positive, negative or neutral. It works on unstructured text, where sarcasm and mixed languages make it harder.
7. Risk modelling. Lenders and insurers estimate the probability of a loss, such as loan default, with logistic regression, decision trees or ensembles. A cut-off then trades lost business against bad loans.
8. Healthcare diagnosis. Models read X-rays and scans, and flag patients at risk of readmission.
9. Route and logistics optimisation. Delivery firms predict delivery times and plan routes from traffic data.
10. Public policy. Governments forecast demand for schools, clinics and transport.
| Application | Learning type | Typical technique |
|---|---|---|
| Customer segmentation | unsupervised | k-means clustering |
| Fraud detection | supervised | classification |
| Predictive maintenance | supervised | classification on sensor data |
| Recommendation | similarity | nearest neighbours |
| Demand forecasting | supervised | regression, time series |
| Sentiment analysis | supervised | text classification |
| Risk modelling | supervised | logistic regression, ensembles |
Conclusion. Data science produces useful knowledge from data using statistics, computing and domain knowledge. Its applications include customer segmentation, fraud detection, predictive maintenance, recommendation, demand forecasting, sentiment analysis, risk modelling, diagnosis, logistics and public policy. Each pairs a business question with a suitable learning method.
The rest of the answers
You have read the question paper and its first answer in full. Buy the solved papers once and you can read every answer of every solved paper in this semester.
Notes + practical solutions + solved papers: ₹798 Already bought it? Sign in
Or solved papers only: ₹499
Or notes + practical solutions: ₹499
The question paper itself stays free, and so does the syllabus.
What should we fix or add?
Found an error in this volume? Report it and we will check it against the paper.