Tree-Based Learning DY0-001 Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   How does a random forest typically compute feature importance?

   • ABy measuring the correlation between features
   • BBy counting the number of times a feature appears in the root node
   • CBy averaging the decrease in impurity (e.g., Gini impurity) across all trees for each feature
   • DBy the order in which features were loaded into the model
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   In decision tree splitting, what does information gain measure?

   • AThe total number of leaves in the tree
   • BThe reduction in entropy before and after a split
   • CThe increase in tree depth after a split
   • DThe number of misclassified samples after a split
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A data scientist is explaining a decision tree to a colleague. The tree has a root node that splits on 'age > 30', and the left child is a leaf node predicting 'No', while the right child is an internal node that splits on 'income'. What does the leaf node 'No' represent?

   • AAn internal node that further splits the data based on another feature.
   • BA branch that connects the root to the next internal node.
   • CA terminal node that outputs a predicted class for records that follow that path.
   • DThe root node of the tree, which is the first split.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A data scientist is training a random forest on a dataset with high-dimensional features and notices that individual decision trees overfit, but the random forest performs well on validation data. What is the primary reason the random forest reduces overfitting compared to a single decision tree?

   • AIt uses a single deep tree but applies strong pruning to all branches.
   • BIt averages predictions from many trees trained on different bootstrap samples and random feature subsets.
   • CIt selects the best tree from a set of candidates based on validation performance.
   • DIt trains each tree on the same data but with different hyperparameters to create diversity.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   What is a 'weak learner' in the context of boosting?

   • AA model that performs slightly better than random guessing
   • BA model with high variance and low bias
   • CA model that always predicts the majority class
   • DA model that is trained on the entire dataset with all features
   Reveal Solution Discussion Select an answer first