Module 1 Questions

RBT levels: L1 – Remember, L2 – Understand, L3 – Apply, L4 – Analyze, L5 – Evaluate

Q. RBT
Question CO
No Level
Define Data Science, its relation to Big Data, and why it's more than Big Data
technology. Illustrate with a real-world example.
OR
A retail analytics startup collects terabytes of transaction, click-stream, and
customer demographic data using cloud platforms. However, management
claims that “Big Data alone will automatically generate business insights.”
a) Critically analyze this statement.
1. L2 1
b) Explain how Data Science complements Big Data
technologies to generate actionable intelligence.
c) Illustrate your answer with a real-world application.
OR
Define Data Science and apply its components (statistics, ML, domain
knowledge) to illustrate how it would be used to predict student performance
using institutional data.
Explain the evolution of Data Science. Why is data science considered
interdisciplinary? Illustrate with a real-world example from industry.
2. OR L2 1
Explain Data Science evolution, its interdisciplinary nature, and moving
beyond hype for real value (with industry example).
List common reasons why Data Science is sometimes considered “hyped” and
apply practical steps an organization can use to implement a small pilot project
for sales prediction.
OR
Data Science is often described as a “hyped” field in recent years. Analyze the
reasons behind this perception and explain how organizations can move
beyond the hype to achieve real value.
OR
3. Differentiate between Big Data hype and real value creation. Given a L3 1
healthcare dataset with patient logs, explain how “getting past the hype” leads
to actionable insights.
OR
“Big Data alone does not generate business value.”
Critically analyze this statement using a healthcare or retail analytics scenario.
OR
How organizations can move beyond the hype to achieve actionable insights.
Justify your answer with a suitable real-time example.
Explain the role of a Data Scientist in:
a) Academia b) Industry
Highlight key differences in responsibilities and outcomes.
4. L2 1
OR
Explain Data Scientist roles in academia vs. industry (differences in
responsibilities/outcomes).
5. What is "Datafication"? Using a smart city traffic system as an example, L3 1
explain how converting real-world data into digital form enables AI. Also,
discuss the conflict between personalized services and privacy risks.
Q. RBT
Question CO
No Level
OR
Explain the concept of Datafication. Discuss its implications in modern data
collection systems with suitable examples.
OR
Smart cities deploy IoT sensors, surveillance cameras, and mobile applications
to monitor citizens’ activities in real time.
a) Explain how this reflects the concept of Datafication.
b) Analyze its implications on data collection, personalization, and privacy.
c) Suggest safeguards for responsible data usage.
OR
Analyze the impact of datafication on modern decision-making systems. How
has datafication enabled AI-driven applications in smart cities?
OR
Explain the concept of Datafication. Analyze how datafication enables AI-
driven decision-making in smart cities, and discuss its implications on privacy
and personalization.
OR
Define Datafication and apply the concept to explain how traffic sensors
generate data used in an AI-based traffic signal system.
Describe the current landscape of perspectives on Data Science. Explain how
it integrates statistics, computer science, and domain expertise.
6. Or L3 1
Describe Data Science perspectives, integrating statistics, computer science,
and domain expertise. Outline the process lifecycle with feedback loops.
Distinguish between a population and a sample, with suitable examples.
7. Explain why sampling bias is critical in data science projects. For a 10,000- L2 1
student dataset, identify the population, draw a sample, and estimate the mean.
Explain statistical modeling with reference to populations, samples, and
overfitting. Illustrate the impact of overfitting on model performance.
Or
A predictive model performs exceptionally well on training data but fails when
8. deployed in real-world conditions. L3 1
a) Explain this behavior using concepts of population, sample, and
statistical modeling.
b) Analyze this behavior and explain overfitting and its mitigation
techniques.
9. Define probability distribution. Explain Normal and Poisson distributions with L3 1
suitable real-world examples.
Given a dataset following a normal distribution:
Mean = 50, SD = 10
Calculate and interpret the percentage of data lying between 40 and 60.
OR
Define probability distribution. Differentiate Normal and Poisson distributions
and apply them to identify which distribution suits:
 Number of calls per hour
 Heights of students
Given: Mean = 60, SD = 8
Q. RBT
Question CO
No Level
Calculate the percentage of values between 52 and 68.

A call-center manager wants to model the number of incoming calls per hour,
while an education researcher studies students’ exam scores.
10. L3 1
Identify suitable probability distributions for both scenarios and justify your
choices with reasoning.
Explain the Data Science process life cycle. Highlight the importance of
11. L2 1
feedback loops in improving models.
Explain the relational data model with a neat diagram. Describe tables, rows,
columns, and keys. Design a university database schema showing:
12.  Primary key L2 1
 Foreign key
 One-to-many relationship
Explain how SQL supports data preprocessing in data science workflows.
Write SQL queries to:
1. Create a student table
2. Insert records
13. 3. Retrieve students with CGPA > 8 L2 1
OR
Explain SQL for data preprocessing. Write queries: create student table, insert
records, retrieve CGPA >8. Evaluate RDBMS/SQL limits for modern Data
Science.
Describe R environment and basic data types. Explain
14. L2 1
variables/constants/naming rules with examples.
15. List and describe arithmetic, relational, and logical operators in R. L2 1
What are vectors in R? Show operations like subsetting and functions (mean,
16. L2 1
sd).

Module 2 Questions
RBT levels: L1 – Remember, L2 – Understand, L3 – Apply, L4 – Analyze, L5 – Evaluate
RBT
[Link] Question CO
Level
1. Define Exploratory Data Analysis (EDA) and explain its philosophy as L2/L3 2
 RBT
[Link] Question CO
Level
proposed by John Tukey.
OR
Define Exploratory Data Analysis (EDA). Why is EDA considered a
mindset rather than merely a set of tools? Illustrate with suitable
examples.
OR
Define Exploratory Data Analysis (EDA). You are given a real-world
dataset containing missing values, outliers, and unknown distributions.
Explain how EDA helps in understanding data characteristics before
model building, and justify the use of appropriate EDA techniques.
You are given a real-world dataset containing missing values, outliers, and
unknown distributions. Explain how EDA helps in understanding data
2. L3 2
characteristics before model building, and justify the use of appropriate
EDA techniques.
Outline the complete Data Science process, from raw data acquisition to
the feedback loop, with a neat diagram and an illustrative example.
3. L4 2
Also, analyze the differences between EDA and data visualization in the
context of the data science lifecycle.
Explain how a data scientist contributes across different stages of the Data
Science process.
4. L3 2
Compare the Data Science process with the traditional scientific method,
highlighting similarities and differences.
Explain how data is used in the RealDirect case study for real estate
decision-making.
OR
In the RealDirect case study, property prices, listings, and historical trends
are analyzed to support buyer and seller decisions. Explain how data is
transformed into insights and identify challenges unique to real-estate data
analytics.
OR
5. Analyze the Real Direct (online real estate) case study. How did EDA L3/L4 2
contribute to better business decisions?
OR
Analyze the RealDirect (online real estate) case study.
Explain how EDA and data analysis transform property listings, prices,
and historical trends into actionable insights.
OR
Discuss key data challenges in real-estate analytics and propose suitable
data science solutions.
6. Explain the working of Linear Regression with a suitable example. L3 2
Given the dataset:
 X (size of house): [1200, 1800, 2400]
 Y (price): [72, 108, 144]
Tasks:
1. Fit a simple linear regression model for the dataset.
2. Find slope and intercept.
3. Write the regression equation.
 RBT
[Link] Question CO
Level
4. Discuss the limitations of Linear Regression when applied to
real-world datasets.
OR
Explain the working of Linear Regression with a suitable example.
Given the dataset:
 X (size of house): [1000, 1500, 2000]
 Y (price): [50, 75, 100]
Fit a simple linear regression model and interpret the result.
Also, discuss the limitations of Linear Regression in real-world datasets.
Explain the k-means clustering algorithm and its objective function.
7. Given customer purchase behavior data, design a k-means-based customer L3 2
segmentation model and justify your choice of k.
Explain k-NN algorithm and apply it for classification.
Training data:
Point Class
(2,3) A
(4,5) A
8. (7,8) B L3 2
Test point = (5,6), k = 3
Use Euclidean distance and find the predicted class.
OR
Discuss the working of the k-NN algorithm. Given a test point and k=3,
demonstrate how kNN classification works using Euclidean distance.

A company needs to predict house prices, classify customer segments, and

group similar users without labels. Select the appropriate algorithms from
9. L4 2
Linear Regression, k-NN, and k-means, and justify your answer for each
task.
Compare Linear Regression and k-NN in terms of bias, variance, and
10. scalability. Discuss how these factors influence algorithm selection in L3 2
large-scale data systems.
Compare statistical modeling and machine learning approaches in terms
11. L3 2
of assumptions, interpretability, and scalability, using suitable examples.
Explain the role of algorithms such as sorting, optimization, and machine
12. learning in the Data Science pipeline. Illustrate how algorithmic L3 2
efficiency impacts large-scale data systems.

Module3
RBT
[Link] Question CO
Level
1. Explain the spam filtering problem as a motivating application in Data L3 3
Science.
 RBT
[Link] Question CO
Level
OR
Describe the key challenges involved in detecting spam emails and the
types of data used in spam filtering systems.
Why are Linear Regression and k-Nearest Neighbors (k-NN) considered
poor choices for spam filtering?
Analyze their limitations with respect to:
 Nature of output
 High dimensionality
 Scalability
 Real-world text data characteristics
2. OR L3 3
State why Linear Regression and k-NN are not suitable for spam filtering
and apply this reasoning to a dataset containing 10,000-word features.

OR
Explain why Linear Regression and k-NN are unsuitable for spam
filtering. Support your justification using characteristics of high-
dimensional text datasets.
State Bayes’ theorem and apply Naïve Bayes classification to compute the
probability of spam.
Given: P(Spam)=0.4
P(Word|Spam)=0.7 P(Word|NotSpam)=0.2
P(NotSpam)=0.6
3. L3 3
Find P(Spam|Word).
OR
Explain the Naïve Bayes classification algorithm based on Bayes’
Theorem and discuss why Naïve Bayes works effectively for spam
filtering, despite its strong independence assumption.
Describe the step-by-step working of a Naïve Bayes spam filtering
system, including:
 Text preprocessing
4.  Feature representation L3 3
 Training phase
 Classification of a new email
Illustrate with a simple example.
5. Given a small dataset of emails labeled spam and not spam, compute the 3
probability of a new email being spam using Naïve Bayes.
Spam emails: 5
Not spam emails: 5

Word frequencies:
Word Spam NotSpam
free 4 1
offer 3 0
hello 1 4
New email contains: free, offer
 RBT
[Link] Question CO
Level
OR
State Bayes’ theorem and apply Naïve Bayes classification to compute the
probability of spam.
Given:
P(Spam)=0.4
P(Word | Spam) =0.7
P(Word | NotSpam) =0.2
P(NotSpam)=0.6
Find P(Spam | Word)
Evaluate the performance of Naïve Bayes in comparison with kNN for
6. L3 3
text classification problems.
Compare Naïve Bayes, Linear Regression, and k-NN for spam filtering in
terms of:
 Accuracy
7.  Computational efficiency L3 3
 Interpretability
 Suitability for high-dimensional data
Justify why Naïve Bayes is preferred.
Explain how probabilistic outputs improve decision-making in spam
8. filtering systems. Why is this aspect important when compared to L2 3
distance-based or regression-based classifiers?
Define Data Wrangling and apply preprocessing steps to clean a dataset
containing missing values and inconsistent date formats.
OR
9. L2 3
Define Data Wrangling. Explain why data wrangling is a critical step
before applying machine learning algorithms, especially in real-world
applications such as spam filtering.
Describe the role of APIs in data collection with a real-world example.
OR
10. Explain how APIs are used for data collection in Data Science. L3 3
Describe the steps involved in collecting data from a REST API, and
discuss challenges such as rate limits, authentication, and data quality.
Explain the concept of web scraping. Discuss commonly used tools and
techniques for web scraping and analyze the ethical and legal
considerations involved.
11. L3 3
OR
Explain web scraping and apply it to extract product prices from an e-
commerce page using any tool or method.
Differentiate between API-based data collection and web scraping.
Compare them in terms of:
 Reliability
12. L3 3
 Scalability
 Data structure
 Legal compliance
Design a data wrangling pipeline to collect and prepare email text data for
13. spam filtering. Include steps such as data acquisition, cleaning, L3 3
transformation, and feature extraction.
 RBT
[Link] Question CO
Level
Explain how unstructured web data is transformed into structured data
14. L2 3
suitable for machine learning.
Propose a complete spam detection system using Naïve Bayes, from data
collection to deployment.
OR
A company wants to monitor fake reviews on its platform. Design an end-
15. to-end spam filtering system using Naïve Bayes, starting from raw email L3 3
data collection to final classification.
OR
Evaluate a spam filtering system deployed in an email service. What
factors affect its accuracy?

Module 4:
RBT
[Link] Question CO
Level
1. Explain the concept of feature generation. L3 4
Discuss the role of brainstorming, domain expertise, and imagination in
creating meaningful features with a suitable real-world example.
OR
 RBT
[Link] Question CO
Level
Explain feature generation and its role in machine learning. Discuss
brainstorming, domain expertise, imagination, and a real-world example.
OR
Define feature generation and apply it to create three new features from a
dataset containing age, income, and purchase count.
Describe user (customer) retention as a motivating application for feature
selection. Discuss how features influence retention prediction models.
2. OR L3 4
Describe user retention as an application for feature selection. Explain how
features influence retention prediction models.
What is feature selection? Explain why feature selection is crucial in
building interpretable and high-performance machine learning models
3. Or L3 4
What is feature selection? Explain its importance for interpretable, high-
performance models.
Explain Filter-based feature selection methods. Discuss commonly used
4. L3 4
statistical measures and their advantages and limitations.
Explain Wrapper-based feature selection methods.
Describe the working principle, computational complexity, and scenarios
where wrappers outperform filters.
OR
Explain wrapper-based feature selection methods, their working principle,
computational complexity, and scenarios where they outperform filters.
5. L3 4
Compare filter and wrapper methods on performance, scalability, and
overfitting risks.

OR
Explain wrapper feature selection and demonstrate how Recursive Feature
Elimination would select features from a dataset with 5 features.
Compare filter and wrapper methods for feature selection in terms of
6. L3 4
performance, scalability, and overfitting.
Explain embedded feature selection methods with Decision Trees.
Describe how entropy and information gain are used to select features.
7. OR L3 4
Explain embedded feature selection with Decision Trees. Describe entropy,
information gain, handling of continuous variables, and stopping criteria.
Explain the Decision Tree algorithm in detail. Discuss handling of
8. L3 4
continuous variables and stopping criteria.
9. Given a dataset with 4 features:
Feature Gini Importance from Trees
Age 0.10
Income 0.35
Credit Score 0.40
Loan Amount 0.15
Tasks:
1. Construct a Random Forest model conceptually using 3 decision
 RBT
[Link] Question CO
Level
trees and explain how bagging is applied.
2. Calculate the overall feature importance ranking.
3. Identify which feature the model considers most influential.
4. Use the feature importance values to decide which two features
should be selected for a reduced model.

Describe Random Forests as an ensemble method.

Explain how randomness improves predictive power and reduces overfitting
Or
Explain the concept of Random Forests. How do Random Forests help in
feature importance estimation and improve model performance?
OR
Describe Random Forests as an ensemble method. Explain randomness for
10. L3 4
reducing overfitting, feature importance estimation, and improved
performance over single Decision Trees (including interpretability vs.
predictive power trade-offs).
OR
Explain Random Forest and apply it to determine feature importance for a
dataset predicting loan default.

Compare Decision Trees and Random Forests.

11. Discuss interpretability versus predictive performance in the context of user L3 4
retention problems.
Discuss the trade-off between interpretability and predictive power in
feature selection and model building, with reference to decision trees and
ensemble models.
12. OR L3 4
Discuss the interpretability-predictive power trade-off in feature selection
and modeling, referencing Decision Trees, Random Forests, and user
retention problems.
Explain the architecture of a real-world recommendation engine.
Describe the process of building a user-facing data product using
recommendation engines.
13. L2 4
OR
Explain the architecture of a real-world recommendation engine and the
process of building a user-facing data product with it.
Describe the algorithmic ingredients of a recommendation system and
14. L2 4
discuss similarity measures, ranking, and feedback mechanisms.
Explain the Nearest-Neighbor approach for recommender systems.
Discuss its limitations in large-scale systems.
OR
15. L2 4
Explain the Nearest-Neighbor approach for recommender systems and its
limitations in large-scale systems (including the dimensionality problem in
high-dimensional user-item matrices).
16. Explain Singular Value Decomposition (SVD) in detail. L3 4
Discuss its role in collaborative filtering and matrix factorization.
OR
 RBT
[Link] Question CO
Level
Explain Singular Value Decomposition (SVD) and its role in collaborative
filtering/matrix factorization for recommendation systems.
OR
State the SVD formula and apply it to reduce a 3×3 matrix to a rank-2
approximation.
Given matrix: A= [ ]
3 0
4 5
Tasks:
1. Apply the Singular Value Decomposition steps to express the matrix A
in the form
T
A=U Σ V (Conceptual step outline allowed).
2. Identify singular values.
3. Explain how SVD can be used to approximate this matrix using rank-1
approximation.
4. State how such an approximation is useful in recommendation systems.
Explain Principal Component Analysis (PCA).
Compare PCA with SVD in the context of dimensionality reduction.
17. OR L3 4
Explain Principal Component Analysis (PCA) for dimensionality reduction.
Compare it with SVD.
Describe Alternating Least Squares (ALS) for recommendation systems.
18. L3 4
Explain the iterative optimization of user and item matrices.
Describe the steps involved in building your own recommendation system.
Explain data preparation, dimensionality reduction, model building, and
evaluation.
19. L4 4
OR
Describe steps to build a recommendation system: data preparation,
dimensionality reduction, model building, and evaluation.

Module 5:
RBT
Q. No Question CO
Level
Define social networks and explain how they are modeled as graphs (nodes,
edges, direction, weights). Discuss varieties and multi-node-type graphs
with examples.
1. L2 5
OR
Define social network. Explain how social networks are modeled as graphs,
discussing nodes, edges, direction, and weights with suitable examples.
2. Explain the different varieties of social networks. Discuss their structural L2 5
 RBT
Q. No Question CO
Level
characteristics with examples.
OR
Explain clustering in social-network graphs. Why are traditional techniques
insufficient? Describe key distance measures for node grouping.
Explain betweenness centrality in social-network graphs and its role in
3. identifying key nodes/edges. Describe its use in the Girvan-Newman L2 5
algorithm for community detection, with an example and limitations.
Explain clustering of social-network graphs.
4. L2 5
Why are traditional clustering techniques insufficient for graph data?
What is direct community discovery? Explain cliques and complete
5. bipartite graphs in social networks, including why they exist in large L2 5
networks.
Explain graph partitioning. What makes a partition "good"? Describe
6. L2 5
normalized cuts and why they outperform simple cuts.
Describe graph representation matrices (adjacency, degree, Laplacian).
7. L3 5
Explain the role of eigenvalues of the Laplacian in partitioning.
Explain neighbourhood properties, directed graphs, graph diameter,
8. L2 5
transitive closure, and reachability (with one method for large graphs).
Explain transitive closure and reachability in graphs.
9. L3 5
Discuss any one method used to compute transitive closure in large graphs.
Explain the basic principles of effective data visualization.
Illustrate how these principles help in selecting appropriate charts for
different types of data.
10. Or L3 5
Explain principles of effective data visualization. Illustrate chart selection
for data types and their role in exploratory data analysis (EDA) for
patterns/trends/anomalies.
Discuss Data Science visualization tools/ideas. Compare two tools and their
use cases.
11. Or L3 5
Discuss the ideas and tools used for data visualization in Data Science.
Compare at least two visualization tools and explain their use cases.
Analyze the ethical issues in Data Science with respect to data collection,
12. data usage, and algorithmic bias. L3 5
Support your answer with suitable examples.
Evaluate privacy/security challenges in large-scale analytics. Suggest
13. L3 5
strategies for ethical data handling.
Critically evaluate the impact of data-driven decision-making on society,
14. L3 5
focusing on privacy, transparency, and accountability.
Discuss the roles/responsibilities of data scientists. Propose essential
15. L3 5
technical skills and ethical competencies for future challenges.