Practice Questions
Practice Questions
Practice Questions
1. Given a classification problem with input features as shown in the graph. Can we use logistic regression to
classify this dataset? Explain why/why not. If yes, draw the decision boundary.
3. Consider the below ROC curves for datasets with different numbers of negative examples. Which model would
you prefer and why, considering the points on the curves with the same value of specificity 0.84 and sensitivity
0.5?
10. Shyam read about Generalized linear model (GLM) and made some assumptions, help shyam by commenting
True/False on each statement. Justify your answers. 2 marks
(a) Target variable are independently distributed.
(b) Target Variable must normally distributed
(c) Non linear relationship between link function and explanatory variables
(d) Errors must be normally distributed
11. Given the prediction data of a machine learning model, find (i) Precision (ii) Recall (iii) True Positive Rate (iv)
F1 Score 2 marks
Predicted Original
True True
False True
False False
True True
False True
True False
False True
12. Shyam and Raju are working on some linear model, Shyam reported Pearson correlation coefficient value of
+1.73 for its model, Raju is lazy and provided this data 2 marks
(x − x̄) (y − ȳ)
-3.2 -3.84
-1.2 -1.44
-0.2 -0.04
1.8 2.16
2.8 3.36
Calculate Pearson correlation coefficient for Raju and tell in which model x and y has strong positive linear
relationship.
13. Derive weight update equation for linear regression using mean square error (MSE) loss function.
14. Plot an approximate transformation how the given data (Inner circle is class 1) is transformed in higher domain
using gaussian kernel. Justify your answer. 2 marks
17. Select which method has best Equal error rate. Justify your answer. 2 marks
18. A student has access to some data where features and targets are given, he applied principal component analysis
(PCA) and claims he is using supervised learning. Is he correct, justify your answer?