Ensembles are machine learning methods that combine several models/predictors to generate more accurate predictions. This reduces the variance of the predictions. Examples for ensemble methods are boosting, stacking, voting, bagging, and random forests, but also simple linear combinations of the predictor outputs. Ensembles are used for recommender systems to improve the quality of recommendations, and have been particularly popular in competitions like the Netflix Prize, where raw prediction accuracy (as opposed to user satisfaction, diversity, serendipity) was used to determine the winner.
- Dennis DeCoste: Collaborative prediction using ensembles of Maximum Margin Matrix Factorizations, ICML 2006