Decision Trees DATA-SCIENCE-FOUNDATIONS Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A data scientist is building a decision tree for a medical diagnosis task. The dataset is small (500 samples) and has many features. The scientist is concerned about overfitting. They want to use pre-pruning to control tree complexity. Which pre-pruning strategy would be most effective in this scenario?

   • ASet a minimum impurity decrease of 0.0.
   • BUse no pre-pruning and rely on post-pruning.
   • CSet a maximum depth of 10.
   • DSet a minimum number of samples per leaf to 20.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which pruning technique stops the tree from growing further during the construction phase based on a predefined condition?

   • AReduced error pruning
   • BCost-complexity pruning
   • CPre-pruning
   • DPost-pruning
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   Which splitting criterion is specifically designed for regression trees to measure the reduction in target variable variability?

   • AVariance reduction
   • BEntropy
   • CInformation gain
   • DGini impurity
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   Which decision tree algorithm uses information gain based on entropy to select splits and handles categorical features natively?

   • AID3
   • BCART
   • CRandom Forest
   • DC4.5
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A data scientist is building a decision tree to predict customer churn. The dataset has 10,000 records and 15 features, including both categorical and numerical variables. The scientist notices that the tree is growing very deep and the training accuracy is 99%, but validation accuracy is only 72%. Which approach would most directly address the overfitting while preserving the tree's interpretability?

   • ARemove all categorical features and retrain the tree.
   • BIncrease the maximum depth of the tree to allow more splits.
   • CApply post-pruning to remove branches that do not improve validation performance.
   • DSwitch to a random forest with 500 trees.
   Reveal Solution Discussion Select an answer first