How does bagging reduce variance in machine learning models?
Updated May 15, 2026
Short answer
Bagging reduces variance by training multiple models on different data samples and averaging their predictions.
Deep explanation
Bootstrap Aggregation (bagging) creates multiple training datasets via random sampling with replacement. Each model learns slightly different patterns. When predictions are averaged, random errors cancel out, reducing variance while maintaining low bias. Random Forest is a bagging extension using decision trees.
Real-world example
Used in credit risk systems to stabilize predictions across noisy financial data.
Common mistakes
- Thinking bagging improves model bias significantly.
Follow-up questions
- What is bootstrap sampling?
- How is bagging different from boosting?