Blending

Stacked Cross-Validation

Alternative to blending where predictions for the meta-model are generated via cross-validation on the training set, reducing overfitting risk but increasing complexity.

← 뒤로