Tree-Based Learning DY0-001 Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A data scientist is building a model for a small dataset and is considering using bagging with decision trees. The scientist is concerned about the variance of the model. Which statement correctly describes the effect of bagging on bias and variance?

   • ABagging reduces both bias and variance by training on multiple samples.
   • BBagging increases bias but reduces variance, leading to a trade-off.
   • CBagging reduces variance by averaging predictions from models trained on bootstrap samples, but it does not significantly reduce bias.
   • DBagging has no effect on bias or variance because it uses the same algorithm.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A data scientist is using a random forest model and wants to explain feature importance to a non-technical audience. Which statement accurately describes how feature importance is typically calculated in a random forest?

   • AIt is the average reduction in impurity (e.g., Gini impurity) attributed to each feature across all trees.
   • BIt is the correlation between the feature and the target variable computed on the training data.
   • CIt is the number of times a feature appears in the root node of any tree.
   • DIt is the coefficient of the feature in a linear regression fitted to the forest's predictions.
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   Why is hyperparameter tuning important for tree-based models?

   • ATo make the model completely deterministic
   • BTo find the best combination of settings that optimizes model performance on unseen data
   • CTo increase the training time of the model
   • DTo eliminate the need for feature engineering
   Reveal Solution Discussion Select an answer first
5. 9application · hard

   A data scientist is building a decision tree for a binary classification problem. At a particular node, the current Gini impurity is 0.48. The scientist evaluates two possible splits: Split A results in child nodes with Gini impurities of 0.2 and 0.3, while Split B results in child nodes with Gini impurities of 0.1 and 0.4. Which split should the scientist choose?

   • AEither split, because both have the same average Gini impurity.
   • BSplit A, because it has a lower average Gini impurity.
   • CThe split that results in the greatest reduction in weighted Gini impurity.
   • DSplit B, because it creates a pure node with impurity 0.1.
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A data scientist is comparing AdaBoost and Gradient Boosting for a dataset with many outliers. The scientist notices that AdaBoost performs poorly because it heavily weights misclassified samples, which may be outliers. Which approach is most appropriate to improve robustness?

   • AIncrease the learning rate in AdaBoost to make it less sensitive to outliers.
   • BUse AdaBoost with a deeper tree base learner to better handle outliers.
   • CRemove all outliers from the dataset before training any boosting model.
   • DUse Gradient Boosting with a robust loss function, such as Huber loss, instead of squared error.
   Reveal Solution Discussion Select an answer first