The IIT Madras BS Machine Learning Practice (MLP) End Term paper sat on 13 Sept 2026, in the May 2026 term: 30 questions for 100 marks in 180 minutes. Every question is below with its answer. Take it as a timed mock test to be marked, or read it through first.
Consider the following pandas DataFrame representing employee training records:
We want to retrieve all records of employees who belong to departments where the average training score is strictly greater than 75. Which of the following code snippets correctly achieves this?
df[df['Score'] > 75]
df.groupby('Department').filter(lambda x: x['Score'].mean() > 75)
df[df.groupby('Department')['Score'].transform('mean') >= 75]
df.groupby('Department').apply(lambda x: x[x['Score'] > 75])
Correct answer
df.groupby('Department').filter(lambda x: x['Score'].mean() > 75)
How many features remain in X_new after the transformation?
0
1
2
3
Correct answer
1
Determine the output of the following text preprocessing pipeline:
['red','blue','green','yellow']
['blue' 'red','yellow']
['blue' 'red']
['red','blue','green']
Correct answer
['blue' 'red']
Four data points are given: A=(1,1), B=(2,2), C=(8,8), D=(10,10) A hierarchical agglomerative clustering algorithm using single linkage and Euclidean distance is applied. Initially, each point is considered as an individual cluster. Which two clusters will be merged in the first step?
A and B
C and D
A and D
B and C
Correct answer
A and B
A genre-based content recommender uses the 19 one-hot genre columns of each movie:
Which of the following statements is TRUE?
Because the genre vectors are binary, two movies that share no genre still receive a positive cosine similarity.
Correct answer
What is the output of the following preprocessing pipeline code? (Round off to 2 decimal places if applicable)
Correct answer: 2.50
Suppose the orginal dataset has 10 rows (1 to 10) Which row/rows of the dataset will be printed if the following code is executed?
Enter the line number (1,2...,10) of the printed row.
If multiple lines are printed enter the sum of the row numbers (i.e. if line 1 and 2 are printed enter 3)
Correct answer: 7
Correct answer: -1.45 (accepted within ±0.15)
Consider a 3 class classification problem where the training dataset contains 10 features. The given model is trained on the dataset.
What is the output of the code given below?
Correct answer: 4
Assume both models in the code given below are fit on the same dataset (X_train) containing 10 features.
Enter the difference between the total number of learnable parameters in both the models. (Enter the absolute value of the difference)
Correct answer: 1
What is the output of the following code? (Round Upto one decmimal) X = [[0], [1], [2], [3]] y = [0, 0, 1, 1] from sklearn.neighbors import RadiusNeighborsRegressor neigh = RadiusNeighborsRegressor(radius=1.0) neigh.fit(X, y) print(neigh.predict([[1.5]]))
Correct answer: 0.5
Correct answer: 5.50
A binary classification model was evaluated on a test set with 500 samples. The model correctly identified 90 positive samples and 360 negative samples. Out of all the samples predicted as positive, 120 were predicted as positive. There were 100 actual positive samples in the dataset. Compute the F1-score of the model. (Round to 2 decimals)
Correct answer: 0.815 (accepted within ±0.015)
Select the false statements:
StandardScaler ensures the features correspond to a uniform distribution after scaling.
MinMaxScaler ensures that feature values follow a Gaussian Distribution
OridnalEncoder is best for data which contains values corresponding to a Normal Distribution.
Pipeline helps us apply separate transformations to numerical and categorical columns directly.
Correct answers
StandardScaler ensures the features correspond to a uniform distribution after scaling.
MinMaxScaler ensures that feature values follow a Gaussian Distribution
OridnalEncoder is best for data which contains values corresponding to a Normal Distribution.
Pipeline helps us apply separate transformations to numerical and categorical columns directly.
Which of the following metrics have their lower bound as 0?
Accuracy
F1-Score
R2-Score
Mean_Squared_Error
Precision
Correct answers
Accuracy
F1-Score
Mean_Squared_Error
Precision
Correct answers
Consider the following implementation of PCA:
Which of the following are correct?
X_pca.shape[1] is equal to the original number of features.
The sum of pca.explained_variance_ratio_ is at least 0.95.
len(pca.explained_variance_ratio_) is equal to X_pca.shape[1].
The principal components correspond to the original features with the highest variance.
Correct answers
The sum of pca.explained_variance_ratio_ is at least 0.95.
len(pca.explained_variance_ratio_) is equal to X_pca.shape[1].
Consider the following two ensemble regression architectures trained on the same dataset:
ensemble1 computes a weighted sum of predictions where later trees have higher weights; ensemble2 takes a simple arithmetic mean of all trees.
ensemble1 reduces variance by averaging independent predictions from 100 trees, meaning increasing n_estimators to 500 will ideally not cause overfitting.
ensemble2 reduces bias sequentially, meaning increasing n_estimators to 500 carries no risk of overfitting.
Both models output predictions by calculating the majority vote across the terminal leaf nodes of all individual base estimators.
Correct answer
ensemble1 reduces variance by averaging independent predictions from 100 trees, meaning increasing n_estimators to 500 will ideally not cause overfitting.
Consider the following code snippet:
Which of the following holds true?
The model performs no regularization.
The model performs regularization and the regularization can also perfrom some feature selection.
The model performs regularization and the regularization cant perfrom some feature selection.
The model is very likely to be prone to overfitting
Correct answer
The model performs regularization and the regularization can also perfrom some feature selection.
Consider the following assertion and reason. Select the correct option.
Assertion (A): Increasing the number of estimators in AdaBoost can sometimes reduce the test accuracy, even though the training error continues to decrease.
Reason (R): In AdaBoost, after each iteration, the weights of correctly classified training samples are increased so that future weak learners focus more on these correct samples. Choose the correct option:
Both A and R are true, and R is the correct explanation of A.
Both A and R are true, but R is not the correct explanation of A.
A is true, but R is false.
A is false, but R is true.
Correct answer
A is true, but R is false.
An email service develops a machine learning model to automatically identify phishing emails.
Any email predicted as phishing is immediately moved to the user's spam folder without notification. A legitimate email incorrectly marked as phishing could cause users to miss important messages, while an actual phishing email that is not detected can still be reported manually by the user.
The company wants to minimize the number of legitimate emails incorrectly classified as phishing.
Which evaluation metric should be given the highest priority while selecting the model?
Accuracy
Precision
Recall
F1-score
Correct answer
Precision
Which data augumentation operation is being perfomed on the input image in the given code?
Rotation
Cropping
Magnification
Flipping
Correct answer
Cropping
Consider the following code:
Assuming the all the models are trained on the same dataset, which of the models is likely to converge the slowest?
model1
model2
model3
code sinppet for model3 is incorrect
Correct answer
model2
A monthly airline-passenger series shows seasonal swings that grow larger as the overall level of the series rises over the years. It is decomposed as:
Which value should fill the blank, and for which reason?
Correct answer
Identify the lines with errors in the given code snippet:
There are no errors in the code
Line1
Line2
Line3
Correct answers
Line1
Line3
Select the correct statements regarding the following code snippet:
model2 is more likely to have a lower bias and higher variance compared to model1.
Increasing the value of C allowed model2 to penalize misclassification errors more strongly.
model1 must always perform better on the test set because it has lower training accuracy.
The gamma parameter has no effect because the kernel is already set to 'rbf'.
Correct answers
model2 is more likely to have a lower bias and higher variance compared to model1.
Increasing the value of C allowed model2 to penalize misclassification errors more strongly.
A company uses K-Means clustering to divide its customers into 4 groups based on their purchasing behavior. After training the model, the data science team observes the following:
Two different runs of K-Means on the same dataset produce different customer groups. One cluster contains customers that are much farther away from their assigned cluster center compared to other clusters.
Which of the following statements are most likely to explain these observations?
K-Means can converge to different solutions because the initial placement of cluster centroids can affect the final clusters.
The cluster with customers far from its centroid may indicate that the chosen value of K does not represent the natural grouping structure of the data.
K-Means guarantees finding the globally optimal clustering solution regardless of initialization.
K-Means assigns clusters by maximizing the distance between points within the same cluster.
Correct answers
K-Means can converge to different solutions because the initial placement of cluster centroids can affect the final clusters.
The cluster with customers far from its centroid may indicate that the chosen value of K does not represent the natural grouping structure of the data.
Identify which of the following lines (marked in comments) contain errors in the given code snippet:
Line1
Line2
Line3
Line4
Correct answers
Line2
Line4
Using user–user collaborative filtering, the predicted rating is
Correct answer: 3.85 (accepted within ±0.05)
A forecasting model achieves an R2 score of 0.98 on the test set. However, it was trained using a feature that contains the sales value of the next day. Which of the following is the most appropriate conclusion?
The model has low bias.
The model has successfully captured seasonality.
The evaluation is unreliable due to data leakage.
The model has underfit the data.
Correct answer
The evaluation is unreliable due to data leakage.
