Scribd
Upload
33 views4 pages

Machine Learning PYQ 2021

Uploaded by

nitob90303
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
33 views4 pages

Machine Learning PYQ 2021

Uploaded by

nitob90303
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Unique Paper Code : 32347607

Name of the Course : B. Sc. (H) Computer Science DSE-3


Name of the Paper : Machine Learning
Semester : VI
Year of Admission : 2015, 2016, 2017, 2018
Duration : 3 Hours Maximum Marks: 75

Instructions for candidates:


Attempt any FOUR questions.
All questions carry equal marks.
The complete answer to a question MUST be uploaded as a single PDF file.

1.
● A classification model performs with a high accuracy on training data but
generalizes poorly to new instances. Identify the problem and illustrate it with the
help of a suitable figure. Enumerate three possible solutions to this problem.

● Indicate whether each scenario given below is a classification or regression


problem. Justify your answer in each case.
o A Company wants to launch a new product and wishes to know whether
the product will be successful in the market. For this purpose, data of 20
similar products is collected that were previously launched. For each
product marketing budget, price of other products in competition, whether
that product was successful, its price, and five other variables were
collected.
o Government is interested in predicting the % change in the USD/INR.
Weekly data for the whole year of 2017 was collected. For each week, the
% change happened in the USD/INR, the % change in the US market, the
% change in the Indian market, and two other variables were studied.
o A data set relates to top 500 companies in the world. The dataset consists
of profit, number of employees, industry and salary of the CEO. The
problem is to identify and understand factors which affect the salary of the
CEO of the company.
o A music company wants to predict the number of copies of a music album
to be sold depending on the popularity rating of the singer, success index
of his previous albums and two other variables.
o A model to identify animal type (mammal vs non-mammal) in images from
the animal encyclopaedia.
2.
● Design a neural network of one variable for the boolean function NOT. Show that
the function is responding correctly with appropriate weight and bias.

● Give an example of a function, say f, which cannot be implemented using a single


perceptron. Suggest a suitable neural network model that can implement the
function f.

● Calculate the output y of a three inputs neuron with bias. The input feature vector
is (𝑥1 , 𝑥2 , 𝑥3 ) = (0.8, 0.6, 0.4) and weight values are [𝑤1 , 𝑤2 , 𝑤3 , 𝑏] =
[0.2, 0.1, −0.3, 0.35]. Use binary Sigmoid function as activation function.

3.
Given below is the data of five students who took a proficiency test as well as language
course.

S. No. Marks in proficiency test Marks in language course


1 95 85
2 85 95
3 80 70
4 70 65
5 60 70

● Use the least square approximation to estimate the linear equation that best predicts
language course performance, based on proficiency test scores?

● Compute the sum of squared error (SSE) using the estimated model.

● If a student scored 80 on the proficiency test, what marks would we expect her to
obtain in the language course?

4.
Consider the dataset given below, which categorizes an article either as Technical
(Class 1) or Non-Technical (Class 0) based on the time spent in reading (in Hours) and
the number of sentences (in multiples of 1000) in that article.

Time (Hours) Sentences (In multiples of 1000) Article Type


2.7 2.5 0
3 3 0
5.9 2.2 1
7.7 3.5 1
● Using the above data, build a logistic regression model to predict the class of an
article using gradient descent method. Assume 𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔 𝑟𝑎𝑡𝑒 = 0.3. Further, in
the first iteration the value of the coefficients is 0, and the bias is 1. Use two
iterations of the gradient descent process to learn the model parameters.

● Compute the error in prediction.

● Use the above model to predict the article type of an article which requires 6.2
hours of reading time and contains 3100 sentences.

5.
What is the assumption of the Naïve Bayes classifier? What are the advantages of the
assumption? Consider the following table of ten conditions, with the attribute purchase
home as the class label. Assume that the loan amount to be paid follows a normal
distribution.

S. No. PhD Student Marital status Loan amount to be Output class


paid (purchase home)

1 Yes Single 125K No

2 No Married 100K No

3 No Single 70K No

4 Yes Married 120K No

5 No Divorced 95K Yes

6 No Married 60K No

7 Yes Divorced 220K No

8 No Single 85K Yes

9 No Married 75K No

10 No Single 90k yes

● Obtain a Naïve Bayes classifier using the above table. Show all the prior and the
conditional probabilities required to compute the posterior probabilities.

● Use the Naïve Bayes classifier obtained above to predict the class label purchase
home for a given new instance: (PhD student = No, marital status =
Married, Loan to be paid = 120K).
6.
● Suppose 10000 patients get tested for flu; out of them, 9000 are actually healthy
and 1000 are actually sick. For the sick people, a test was positive for 620 and
negative for 380. For healthy people, the same test was positive for 180 and negative
for 8820. Construct a confusion matrix for the data and compute the True Positivity
Rate (TPR), False Positivity Rate (FPR), Specificity, Sensitivity and Accuracy of
the test.

● In each case, indicate whether the root mean squared error is a good performance
measure and justify your answer:
○ binary classification problems
○ multiclass classification problems

● List out and briefly explain the representation power of feedforward networks.

You might also like