One model makes mistakes. A hundred models that err in different ways together make far fewer mistakes. That’s the ensemble idea — combine many weak models into one strong one. Random Forest, XGBoost, LightGBM, CatBoost — Kaggle winners on tabular data in recent years all use ensemble methods. This chapter explains how two fundamental approaches work: bagging (parallel trees) and boosting (sequential trees), and how to tune hyperparameters with cross-validation.