MLS-C01 Question 211
Select 2You are training a Random Forest model in SageMaker to classify customer churn. During experimentation, you observe that the model is overfitting the training data. Which of the following changes are likely to reduce overfitting in this scenario?
- A
Reduce the maximum depth of the trees in the Random Forest.
- B
Increase the number of trees in the Random Forest.
- C
Reduce the number of features considered for splitting at each node.
- D
Increase the maximum depth of the trees in the Random Forest.
- E
Decrease the number of trees in the Random Forest.
Show answer and explanation
Correct answers: A, C
Explanation
Overfitting occurs when a model becomes too complex and learns patterns specific to the training dataset, failing to generalize to new data. In tree-based models like Random Forests, reducing the maximum depth of the trees and increasing the number of trees are common techniques to reduce overfitting. Reducing the number of features considered for splitting at each node also increases randomness, which can help prevent overfitting. Increasing the tree depth or decreasing the number of trees will likely exacerbate overfitting.
- A. Correct.
Reducing the maximum depth of the trees decreases model complexity, which helps prevent overfitting by avoiding overly specific splits.
- B. Incorrect.
Increasing the number of trees in a Random Forest generally reduces overfitting because it averages the predictions of more trees, leading to a more robust model.
- C. Correct.
Reducing the number of features considered for splitting at each node can decrease overfitting by introducing more randomness and preventing the model from relying on a small subset of features.
- D. Incorrect.
Increasing the maximum depth of the trees increases model complexity, which can lead to overfitting as the model becomes overly specific to the training data.
- E. Incorrect.
Decreasing the number of trees reduces the model's ability to generalize as it relies on fewer predictions to make decisions, which can increase the risk of overfitting.