What is the difference between bagging, boosting, and stacking in ensemble learning?

Bagging reduces variance via parallel training, boosting reduces bias via sequential learning, and stacking learns how to combine models using a meta-learner.

Bagging trains multiple independent models on bootstrapped samples and aggregates predictions (e.g., Random Forest). Boosting trains models sequentially where each model corrects previous errors (e.g., XGBoost). Stacking trains multiple diverse base models and uses their predictions as inputs to a meta-model that learns optimal combination weights. The key difference lies in training strategy (parallel vs sequential vs hierarchical) and objective (variance vs bias vs optimal blending).