(SEM V) THEORY EXAMINATION 2024-25 DATA ANALYTICS

Subject Code: BCS052
Maximum Marks: 70
Time: 3 Hours
Paper ID: 310908

Question Paper Overview

SECTION A (2 × 7 = 14 Marks)

(Short Answer / Conceptual Questions)

a. Differentiate between Predictive and Prescriptive Data Analytics.
b. Define the term Data Lake, Database, and Data Warehouse.
c. Explain the concept of Outliers.
d. Describe the concept of Lasso Regression.
e. Differentiate between Stream Processing and Traditional Data Processing.
f. Write the two limitations of K-Means.
g. Discuss the various categories of clustering techniques.

SECTION B (Attempt any three × 7 = 21 Marks)

a. Explain the different categories of data analytics with examples.
b. Explore PCA (Principal Component Analysis).

Given data = {4, 8, 13, 7; 11, 4, 5, 14}.

Compute the principal components and reduce dimension from 2D to 1D.
c. Explain Market Basket Analysis.

Is it supervised or unsupervised?

How can a company use it to improve marketing strategies?
d. Differentiate between CLIQUE and ProCLUS clustering algorithms.
e. Differentiate between NoSQL and Relational Databases.

Identify when to use NoSQL instead of a Relational Database, with an example.

SECTION C (Attempt one part from each question × 7 = 35 Marks)

Q3

(a) Differentiate between Structured, Semi-Structured, and Unstructured Data.
OR
(b) Describe Big Data and its characteristics.

Q4

(a) Differentiate between Neural Network and Artificial Neural Network.
OR
(b) Given two fuzzy sets:

A = {(10, 0.2), (20, 0.4), (25, 0.7), (30, 0.9), (40, 1), (50, 0.4)}

B = {(10, 0.4), (20, 0.1), (25, 0.9), (30, 0.2), (40, 0.6), (50, 0.6)}

Apply Union, Intersection, Complement, Bold Union, and Bold Intersection operations.

Q5

(a) Apply the Flajolet-Martin Algorithm on the data stream:

S = 1, 3, 2, 1, 2, 3, 4, 3, 1, 2, 3, 1

Given: h(x) = (6x + 1) mod 5

Identify unique elements in the stream.
OR
(b) Discuss the concept of filtering in Data Stream Processing and explain Bloom Filtering in detail.

Q6

(a) Cluster the following eight points into three clusters using K-Means Algorithm:

A₁(2,10), A₂(2,5), A₃(8,4), A₄(5,8), A₅(7,5), A₆(6,4), A₇(1,2), A₈(4,9)

Initial centers: A₁(2,10), A₄(5,8), A₇(1,2)

Distance function:

• P(a,b)=∣x2−x1∣+∣y2−y1∣P(a,b) = |x₂ - x₁| + |y₂ - y₁|P(a,b)=∣x2​−x1​∣+∣y2​−y1​∣

Find the final cluster centers.
OR
(b) A transaction database has 6 transactions with Support = 50%, Confidence = 60%:

i) Use Apriori Algorithm to find frequent itemsets.
ii) Show all strong association rules (with support & confidence).

Q7

(a) Brief about the main components of MapReduce.
OR
(b) Draw and explain the architecture of HIVE with its features.

Key Topics for Revision

1. Categories of Data Analytics

2. Data Storage Concepts

Database: Structured, transactional data (SQL). Data Warehouse: Historical, analytical storage (OLAP).

Data Lake: Raw, unstructured storage (Hadoop, AWS S3).

3. Outliers

Data points that deviate significantly from others.       Detected using:

Z-score,                                                                          IQR (Interquartile Range),

Visualization (Box Plot).

4. Lasso Regression

Regularized regression using L1 penalty.

Shrinks coefficients to zero → performs feature selection.

5. Stream Processing vs Traditional Processing

6. PCA (Principal Component Analysis)

Used for dimensionality reduction.

Steps:

Standardize data.                                                  Compute covariance matrix.

Calculate eigenvalues & eigenvectors.                 Project data onto principal components.

7. Market Basket Analysis

Unsupervised learning (association rule mining).

Uses Apriori Algorithm:               Finds frequent itemsets, e.g., “Beer → Diaper.”

Applications:                                  Retail recommendations, cross-selling, layout optimization.

8. Clustering

Partitioning methods: K-Means, K-Medoids.          Hierarchical methods: Agglomerative, Divisive.

Density-based: DBSCAN, OPTICS.                            Grid-based: CLIQUE, STING.

9. NoSQL vs Relational Database

10. Big Data Characteristics (5Vs)

Volume: Massive data size.                                 Velocity: Fast data generation.

Variety: Structured, semi/unstructured.              Veracity: Data accuracy.

Value: Extracting useful insights.

11. Flajolet–Martin Algorithm

Estimates number of distinct elements in data streams using hash functions.

Efficient for large-scale streaming data.

12. Bloom Filtering

Probabilistic data structure for membership testing.

Space-efficient but allows false positives.

Used in caching, networking, and databases.

13. Apriori Algorithm

Step 1: Generate frequent itemsets using support.   Step 2: Generate strong association rules using confidence.

Example:

Support(A→B) = freq(A∪B) / total transactions         Confidence(A→B) = freq(A∪B) / freq(A)

14. K-Means Clustering

Iterative algorithm that partitions data into k clusters.

Limitations:

Sensitive to initial centroids.             Assumes spherical clusters.

15. MapReduce Components

Map Phase: Input split → key-value pairs.

Shuffle & Sort: Group similar keys.

Reduce Phase: Aggregate output.

16. HIVE Architecture

Built on top of Hadoop for data querying (SQL-like interface).

Components:

Driver: Compiles queries.

Metastore: Stores schema.

Execution Engine: Converts queries to MapReduce.

HiveQL: SQL-based query language.