Random Forests DATA-SCIENCE-OPTIMIZE Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   How does increasing the number of trees in a random forest help control overfitting?

   • AIt averages out the overfitting of individual trees
   • BIt makes the forest more complex
   • CIt forces trees to be shallower
   • DIt reduces the number of features used
   Reveal Solution Discussion Select an answer first
2. 2expert · hard

   A data scientist is training a random forest and wants to use OOB error to compare two models: one with 50 trees and one with 500 trees. They notice that the OOB error for the 500-tree model is slightly higher than for the 50-tree model. What is the most likely explanation?

   • AThe 50-tree model is underfitting.
   • BThe 500-tree model is overfitting.
   • CThe difference is within random variation, and more trees generally do not increase OOB error significantly.
   • DOOB error is not a reliable metric for comparing models.
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   What does impurity-based feature importance measure in a random forest?

   • AThe number of times a feature appears in the forest
   • BThe correlation between the feature and the target
   • CThe feature's coefficient in a linear model
   • DThe total reduction in impurity (e.g., Gini) attributed to a feature across all trees
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A company wants to build a model to predict customer churn. They have a large dataset with many categorical and numerical features, and they need a model that handles non-linear relationships well. Which approach is most suitable?

   • AUse a random forest, which handles non-linear relationships and mixed feature types well.
   • BUse a single decision tree, which is easy to interpret.
   • CUse k-nearest neighbors, which is simple.
   • DUse linear regression, which is simple and interpretable.
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A data scientist is training a random forest on a highly imbalanced dataset. They are using bootstrap sampling, which may result in some trees seeing very few samples from the minority class. They want to improve the model's recall for the minority class without a separate validation set. Which approach is most appropriate?

   • AUse class weights to penalize misclassifications of the minority class and evaluate using OOB error.
   • BIncrease the number of trees to ensure the minority class is well represented.
   • CRemove the random feature subsampling to let trees focus on minority class features.
   • DUse a smaller bootstrap sample size to include more minority class samples.
   Reveal Solution Discussion Select an answer first