Databricks Machine Learning Associate Question 392
Select 2A data science team is working on a classification problem where the dataset suffers from high variance in predictions when using a single decision tree model. To improve the model's performance, the team decides to use an ensemble learning approach. Which of the following techniques are appropriate ensemble methods to reduce variance?
- A
Bagging
- B
Boosting
- C
Stacking
- D
Random Forest
- E
Principal Component Analysis (PCA)
Show answer and explanation
Correct answers: A, D
Explanation
Bagging and Random Forest are ensemble learning techniques specifically designed to address high variance by aggregating predictions from multiple models trained on different subsets of data. Boosting, while powerful, targets bias reduction, and PCA is unrelated to ensemble learning. Understanding the appropriate application of ensemble methods is crucial for improving model performance in scenarios like this.
- A. Correct.
Bagging (e.g., Bootstrap Aggregating) is an ensemble method designed to reduce variance by training multiple models on different subsets of data and averaging their predictions.
- B. Incorrect.
Boosting focuses on reducing bias rather than variance by sequentially training models to correct the errors made by previous models, so it is not primarily used to address high variance.
- C. Incorrect.
Stacking combines predictions from multiple models using another model (meta-learner) but is generally not aimed specifically at reducing variance.
- D. Correct.
Random Forest is an ensemble technique based on bagging that uses multiple decision trees and random feature selection, effectively reducing variance.
- E. Incorrect.
Principal Component Analysis (PCA) is a dimensionality reduction technique and not an ensemble learning method, so it is not applicable in this scenario.