Total No. of Questions: 6 Total No.

of Printed Pages:3

Enrollment No......................................

Faculty of Engineering
End Sem Examination Dec-2023
CA5EL52 Machine Learning
Programme: MCA / BCA- Branch/Specialisation: Computer
MCA (Integrated) Application
Duration: 3 Hrs. Maximum Marks: 60

Note: All questions are compulsory. Internal choices, if any, are indicated. Answers of
Q.1 (MCQs) should be written in full instead of only a, b, c or d. Assume suitable data if
necessary. Notations and symbols have their usual meaning.
Q.1 i. Among the following option identify the one which is not a type of 1
learning?
(a) Semi unsupervised learning (b) Supervised learning
(c) Reinforcement learning (d) Unsupervised learning
ii. What is the difference between AI and ML? 1
(a) AI is a broader field that encompasses ML and other subfields
(b) AI is a subset of ML that focuses on creating intelligent agents
(c) AI and ML are synonymous terms that refer to the same field
(d) AI and ML are unrelated fields that have different goals and
applications
iii. What is the cost function used in linear regression? 1
(a) Cross entropy (b) Mean squared error
(c) Hinge loss (d) Kullback-Leibler divergence
iv. What is the name of the algorithm that can be used to find the optimal 1
weights for linear regression?
(a) K-means clustering (b) Newton’s method
(c) Gradient descent (d) Genetic algorithm
v. What is the name of the algorithm that can be used to find frequent 1
itemsets in a large dataset?
(a) K-means (b) PCA (c) AdaBoost (d) Apriori
vi. What is the name of the technique that can be used to find the optimal 1
number of clusters in a dataset?
(a) Silhouette analysis (b) Elbow method
(c) Gap statistic (d) All of these

P.T.O.
 [2] [3]
vii. What is the name of the ensemble method that combines multiple weak 1 provide an example of a real-world application where reinforcement
classifiers into a strong classifier? learning is commonly used.
(a) Bagging (b) Boosting (c) Stacking (d) Voting
viii. What is the name of the algorithm that can be used to implement 1 Q.5 i. Explain the importance of evaluating machine learning algorithms and 4
bagging? the role of model selection in this process.
(a) Random forest (b) AdaBoost ii. Differentiate between bagging and boosting as ensemble methods. 6
(c) Gradient boosting (d) XGBoost Provide a practical example where bagging and boosting are
ix. What is the name of the basic building block of a neural network? 1 commonly employed. Explain why each method is effective in the
(a) Neuron (b) Node (c) Perceptron (d) All of these given scenario.
x. What is the name of the function that can be used to activate or 1 OR iii. Explain the AdaBoost algorithm and its main principles. Describe how 6
deactivate a neuron? AdaBoost combines multiple weak learners to create a strong
(a) Activation function (b) Transfer function classifier.
(c) Threshold function (d) All of these
Q.6 Attempt any two:
Q.2 i. List and briefly define the two main types of machine learning 2 i. Explain the fundamental concept of a neural network and its relevance 5
algorithms. in the field of deep learning.
ii. Explain the process of building a supervised machine learning model 8 ii. Define a perceptron and how it serves as a basic building block in 5
from data, step by step. Provide a detailed description of each stage feedforward neural networks. Explain the feedforward process and its
and its significance in the model-building process. role in making predictions using a neural network.
OR iii. Explain the key difference between supervised and unsupervised 8 iii. Introduce Convolutional Neural Networks (CNNs) and explain their 5
learning. Provide one real-world example for each. significance in image and pattern recognition.

Q.3 i. Discuss the fundamental differences between linear regression and 3 ******
logistic regression.
ii. Explain how decision trees are used in supervised learning for both 7
classification and regression tasks. Provide an example of a real-world
application where decision trees are commonly employed.
OR iii. Introduce the concept of Support Vector Machines (SVM) in 7
supervised learning. Describe the primary objective of SVM and
discuss an application where SVM is particularly effective.

Q.4 i. Explain the fundamental differences between K-means clustering and 3

hierarchical clustering.
ii. Define the Apriori algorithm for association analysis. How does it 7
work? What are the key measures used to identify significant
associations between items in a dataset?
OR iii. Introduce the concept of reinforcement learning in unsupervised 7
learning. Discuss the primary goal of reinforcement learning and
 [4]

Marking Scheme
Problem Definition 1 Mark
CA5EL52 (T)-Machine Learning (T) Data Collection 1 Mark
Data Preprocessing 1 Mark
Data Splitting 1 Mark
Q.1 i) Among the following option identify the one which is not a type 1 Model Selection 1 Mark
of learning? Model Training 1 Mark
a) Semi unsupervised learning Model Evaluation 1 Mark
ii) What is the difference between AI and ML? 1 Conclusion and Model Deployment 1 Mark
a) AI is a broader field that encompasses ML and other OR iii. Explain the key difference between supervised and unsupervised 8
subfields learning. Provide one real-world example for each.?
iii) What is the cost function used in linear regression? 1
b) Mean squared error Marking:
iv) What is the name of the algorithm that can be used to find the 1 for correctly explaining the difference between supervised and
optimal weights for linear regression? unsupervised learning. 2 Marks
c) Gradient descent for providing a real-world example of supervised learning.
v) What is the name of the algorithm that can be used to find 1 2 Marks
frequent itemsets in a large dataset? for providing a real-world example of unsupervised learning.
d) Apriori 2 Marks
vi) What is the name of the technique that can be used to find the 1 for clarity and accuracy in the responses. 2 Marks
optimal number of clusters in a dataset?
d) elbow method Q.3 i. Discuss the fundamental differences between linear regression and 3
vii) What is the name of the ensemble method that combines multiple 1 logistic regression.
weak classifiers into a strong classifier? ii. Explain how decision trees are used in supervised learning for 7
b) Boosting both classification and regression tasks. Provide an example of a
viii) What is the name of the algorithm that can be used to implement 1 real-world application where decision trees are commonly
bagging? employed?
a) Random Forest
ix) What is the name of the basic building block of a neural network? 1 explaining how decision trees are used in supervised learning.
d) All of the above 3 Marks
x) What is the name of the function that can be used to activate or 1 providing a real-world application and its relevance 4 Marks.
deactivate a neuron? OR iii. Introduce the concept of Support Vector Machines (SVM) in 7
a) Activation function supervised learning. Describe the primary objective of SVM and
discuss an application where SVM is particularly effective?
Q.2 i. List and briefly define the two main types of Machine Learning 2
algorithms.? Introduction to Support Vector Machines (SVM) - 3 Marks
Explaining the primary objective of SVM and providing an
type of ML Algorithms. One Marks for application example where SVM excels - 4 Marks
each
ii. Explain the process of building a supervised machine learning 8 Q.4 i. Explain the fundamental differences between K-means clustering 3
model from data, step by step. Provide a detailed description of and hierarchical clustering?
each stage and its significance in the model-building process.
3 Marks
Marks Scheme ii. Define the Apriori algorithm for association analysis. How does it 7

P.T.O.
 [2] [3]
work, and what are the key measures used to identify significant
associations between items in a dataset? Explaining the concept of a neural network (2 Marks)
Its relevance in deep learning. (3 Marks)
Defining the Apriori algorithm 3 Marks. ii. Define a perceptron and how it serves as a basic building block in 5
Explaining how it works and mentioning key measures used for feedforward neural networks. Explain the feedforward process
identifying significant associations 4 Marks. and its role in making predictions using a neural network.
OR iii. Introduce the concept of reinforcement learning in unsupervised 7
learning. Discuss the primary goal of reinforcement learning and Defining a perceptron and its role in feedforward neural networks.
provide an example of a real-world application where (2 Marks)
reinforcement learning is commonly used? Explaining the feedforward process and its significance in making
predictions with neural networks. (3 Marks)
Introducing reinforcement learning and explaining its primary iii. Introduce Convolutional Neural Networks (CNNs) and explain 5
goal. 3 Marks their significance in image and pattern recognition?
Providing a real-world application example and its relevance to
reinforcement learning. 4 Marks Introducing CNNs (2 Marks),
Explaining their importance in image and pattern recognition.
Q.5 i. Explain the importance of evaluating machine learning algorithms 4 (3 Marks)
and the role of model selection in this process. ******

Explaining the importance of evaluating machine learning

algorithms 2 Marks
For explaining role of model selection. 2 Marks
ii. Differentiate between Bagging and Boosting as ensemble 6
methods. Provide a practical example where Bagging and
Boosting are commonly employed and explain why each method
is effective in the given scenario.

Accurately differentiating between Bagging and Boosting.

2 Marks
Providing a real-world example (2 Marks)
explaining the effectiveness of Bagging and Boosting in that
scenario. (2 Marks)
OR iii. Explain the AdaBoost algorithm and its main principles. Describe 6
how AdaBoost combines multiple weak learners to create a strong
classifier.

Explaining the AdaBoost algorithm and its main principles.

(2 Marks),
Describing how AdaBoost combines multiple weak learners and
providing a real-world example. (4 Marks)

Q.6 Attempt any two:

i. Explain the fundamental concept of a neural network and its 5
relevance in the field of deep learning.

P.T.O.