Stacking, or stacked generalization, involves training a new model to aggregate the predictions of several base models. The base models are typically different machine learning algorithms, and their predictions are used as inputs for the final model to make the final prediction. This approach leverages the strengths of each base model and mitigates their weaknesses, aiming to achieve better performance than any single model could on its own.
Historically, stacking was introduced by David Wolpert in 1992. It gained popularity as a powerful method to improve predictive performance by combining the strengths of multiple models.
David Wolpert is a key contributor to the development of stacking. His work laid the foundation for using this technique in various machine learning competitions and real-world applications, demonstrating its effectiveness in enhancing model accuracy by intelligently integrating diverse models.