What is diversity generation strategy in ensemble learning?

Diversity generation strategies create differences among base models to improve ensemble performance.

Ensemble performance depends heavily on diversity among base learners. Diversity generation strategies intentionally introduce variation through data sampling (bagging), feature sampling (random subspace), algorithm heterogeneity, hyperparameter variation, and input perturbations. The goal is to ensure that models make uncorrelated errors so aggregation reduces overall error. Without diversity, even highly accurate models may provide little ensemble gain.