Hands-On Exercise: Demonstrate Overfitting and Underfitting CT-AI Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A junior analyst trains a linear regression model to predict energy consumption from temperature and humidity. The model achieves an R² of 0.30 on both the training set and the test set. The residual plot shows a clear curved pattern. Which action is most appropriate?

   • AAdd polynomial features to capture the nonlinear relationship.
   • BApply L2 regularization to the existing linear model.
   • CCollect more training data and retrain the same linear model.
   • DIncrease the learning rate of the gradient descent optimizer.
   Reveal Solution Discussion Select an answer first
2. 17expert · hard

   A team is training a gradient-boosted tree model on a dataset with 10,000 samples and 500 features. The training AUC is 0.98, but the test AUC is 0.74. The team has limited compute budget and cannot add more data. Which mitigation strategy is most likely to improve test AUC while keeping compute costs reasonable?

   • AReduce the maximum tree depth, increase the minimum samples per leaf, and apply feature subsampling.
   • BIncrease the number of trees and reduce the learning rate.
   • CSwitch to a deep neural network with more layers.
   • DRemove all features except the top 5 by correlation with the target.
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   Which situation is most likely to lead to underfitting?

   • AUsing a decision tree with a very high maximum depth on a small dataset.
   • BApplying a linear regression model to data that has a non-linear relationship.
   • CTraining a model with a very low learning rate for a sufficient number of epochs.
   • DTraining a deep neural network on a small dataset without regularization.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   In a hands-on exercise, a student trains a k-nearest neighbors (k-NN) classifier on a dataset of 200 samples. The student wants to demonstrate overfitting. Which value of k is most likely to produce overfitting?

   • Ak = 10
   • Bk = 1
   • Ck = 50
   • Dk = 200
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   Which action is most likely to help reduce underfitting?

   • AReducing the model's capacity by using fewer parameters.
   • BAdding more relevant features through feature engineering.
   • CUsing a smaller training dataset.
   • DIncreasing the strength of L2 regularization.
   Reveal Solution Discussion Select an answer first