What is stacking in ensemble learning?
Updated May 16, 2026
Short answer
Stacking combines multiple models using a meta-model to learn optimal blending.
Deep explanation
Stacking trains multiple base learners and uses their outputs as inputs to a higher-level meta-learner. Unlike bagging or boosting, stacking learns how to combine predictions optimally rather than using simple averaging or weighting.
Real-world example
Used in Kaggle competitions to maximize predictive performance.
Common mistakes
- Training meta-model on same data without cross-validation.
Follow-up questions
- Why is stacking powerful?
- What is the risk in stacking?