DCA 71A2 MACHINE LEARNING JULY SEPTEMBER 2025

Description

SESSION	JULY/SEPTEMBER 2025
PROGRAM	BACHELOR OF COMPUTER APPLICATIONS (BCA)
SEMESTER	3
COURSE CODE & NAME	DCA 71A2  MACHINE LEARNING

 

 

SET 1

 

Q1. A) A dataset contains the following feature values for “Age”: [15, 20, 25, 30, 35, 40]. Normalize the data using Min-Max Normalization to the range [0,1]. Standardize the data using Z-score normalization. (Show all steps clearly.)

1. B) Differentiate between classification and regression algorithms in supervised learning. Explain how decision trees can handle both types of problems with suitable examples.

Ans 1.

1. Min–Max Normalization and Z-Score Standardization

Given feature Age:

1. Min–Max Normalization to [0,1]

Formula:

Here:
so .

Now compute for each value:

• For 15:

• For 20:

• MUJ
• Its Half solved only
• Buy Complete assignment from us
• Price – 190/ assignment
• MUJ Manipal University Complete SolvedAssignments JULY-AUGUST  2025
• buy cheap assignment help online from us easily
• we are here to help you with the best and cheap help
• Contact No – 8791514139 (WhatsApp)
• OR
• Mail us- [email protected]
• Our website – https://muj.assignmentsupport.in/

 

Q2. A) Given the following two-dimensional data points:

A(1,1), B(1.5,2), C(3,4), D(5,7), E(3.5,5), F(4.5,5), G(3.5,4.5). Perform one iteration of the K-Means clustering algorithm for k = 2.

(Assume initial centroids are A and D. Compute new centroids after one iteration.)

1. B) Explain the importance of model evaluation. Compare Hold-Out, k-Fold Cross-Validation, and Leave-One-Out Cross-Validation methods. Which one is preferred for small datasets and why?

Ans 2.

1. One Iteration of K-Means (k = 2)

Given Data Points

Initial centroids:

• Cluster 1 centroid
• Cluster 2 centroid

We use Euclidean distance:

 

 

 

Q3.A) Consider a single-layer neural network with inputs x₁ = 0.6, x₂ = 0.8, weights w₁ = 0.5, w₂ = 0.4, and bias b = 0.1. Compute the weighted sum.

Apply a sigmoid activation function to find the output.

(Show all calculations.)

1. B) Define the key elements of a reinforcement learning system — Agent, Environment, Reward, Policy, and Value Function.

How does Q-learning help an agent learn optimal behavior?

Ans 3.

1. Single-Layer Neural Network: Weighted Sum + Sigmoid Output

Given:

• Inputs:
• Weights:
• Bias:

1. Compute Weighted Sum (Net Input)

General formula for a single neuron:

Substitute values:

 

 

 

SET II

 

Q4. A) Given the following two sentences:

• S₁: “Machine learning is amazing.”
• S₂: “Learning machines are amazing.”

1. Create a Bag of Words (BoW) representation.
2. Calculate the cosine similarity between S₁ and S₂.(Show each step clearly.)

1. B) Explain the difference between content-based filtering and collaborative filtering in recommender systems. Give one example of how each method is used in real-world applications (e.g., Netflix, Amazon).

Ans 4.

1. Bag of Words and Cosine Similarity

Given:

•  “Machine learning is amazing.”
•  “Learning machines are amazing.”

1. Create Bag of Words (BoW)

Step 1 – Tokenize (lowercase, remove punctuation):

• : machine, learning, is, amazing
• : learning, machines, are, amazing

Step 2 – Build vocabulary (unique words from both):

Let the order be:

machine

learning

is

amazing

machines

are

 

 

Q5. A) You have a dataset with values: [10, 12, 11, 13, 50, 12, 11].

1. Compute the mean and standard deviation.
2. Identify whether 50 is an anomaly using the z-score method (threshold = ±2).

1. B) Discuss how Apache Spark supports large-scale machine learning. Explain the role of MLlib and how it handles data processing differently from traditional ML libraries.

Ans 5.

Mean, Standard Deviation, and Anomaly Detection with Z-Score

Dataset:

Let’s treat it as a population for simplicity (denominator = N).

1. Compute Mean

Sum:

Number of values :

So, mean = 17.

 

 

 

 

Q6.A) List and explain key ethical issues in machine learning, such as bias, privacy, and transparency.

How can organizations ensure responsible AI development?

1. B) A company wants to predict customer churn using a logistic regression model.

If the model outputs a probability of 0.82 for a given customer, and the threshold is 0.6:

1. Will the model predict “Churn” or “No Churn”?
2. Explain briefly how changing the threshold might affect precision and recall. 5

Ans 6.

1. Ethical Issues in Machine Learning + Responsible AI

Key Ethical Issues

Bias and Fairness

• If training data is biased (e.g., under-represents certain groups), the model may discriminate.
• Example: Loan approval model unfairly rejecting applications from a particular community.

Privacy

• ML models often use sensitive data: health, finance, location.