Tree-Based Learning DY0-001 Practice Questions (Page 7)

Questions 31–35

1. 31application · medium

   A data scientist is training a gradient boosted tree model and observes that the training loss keeps decreasing but the validation loss starts increasing after a certain number of boosting rounds. Which strategy is most appropriate to address this?

   • AStop training early when validation loss begins to increase.
   • BIncrease the number of boosting rounds to allow the model to fit the training data better.
   • CRemove the validation set and use the training loss to determine the optimal number of rounds.
   • DIncrease the learning rate to speed up convergence and reduce overfitting.
   Reveal Solution Discussion Select an answer first
2. 32foundation · easy

   How do boosting algorithms like AdaBoost and Gradient Boosting improve model performance?

   • ABy training all weak learners independently in parallel
   • BBy training weak learners sequentially, each focusing on the errors of the previous ones
   • CBy averaging the predictions of many deep trees
   • DBy using a single complex model with all features
   Reveal Solution Discussion Select an answer first
4. 33application · medium

   A data scientist is building a decision tree to predict customer churn. The dataset has two candidate features for the root split: 'contract_type' (categorical) and 'monthly_charges' (continuous). The scientist computes the Gini impurity for both potential splits and finds that 'contract_type' yields a lower Gini impurity than 'monthly_charges'. What should the scientist do?

   • AChoose 'monthly_charges' because Gini impurity is only applicable to categorical features.
   • BChoose 'contract_type' for the root split because lower Gini impurity indicates a purer split.
   • CChoose 'monthly_charges' because continuous features are always more informative than categorical ones.
   • DChoose 'contract_type' because it has fewer categories, which always leads to better generalization.
   Reveal Solution Discussion Select an answer first
5. 34foundation · easy

   What is the role of randomness in a random forest?

   • ATo create diversity among trees by sampling data and features randomly
   • BTo ensure each tree is identical
   • CTo make the model deterministic
   • DTo eliminate the need for training data
   Reveal Solution Discussion Select an answer first
6. 35expert · hard

   A data scientist is tuning a random forest model and wants to reduce overfitting. The scientist is considering adjusting the following hyperparameters: max_depth, min_samples_split, and max_features. Which combination of changes is most likely to reduce overfitting?

   • AIncrease max_depth, increase min_samples_split, and decrease max_features.
   • BKeep max_depth the same, decrease min_samples_split, and keep max_features the same.
   • CDecrease max_depth, increase min_samples_split, and decrease max_features.
   • DIncrease max_depth, decrease min_samples_split, and increase max_features.
   Reveal Solution Discussion Select an answer first