ENSEMBLE-BASED DECISION MAKING FOR FORECASTING

Ensemble learning is an approach that combines the results of several base models into a single final decision. In general, an ensemble is defined through an aggregation function that combines the outputs of base models. In the literature, three main characteristics are distinguished to explain ensembles: the dependency in training of base models, the combination method—voting or meta-learning, and the diversity of base models—homogeneous or heterogeneous. In this mini review, the differences between bagging, boosting, and stacking are summarized in simple terms based on these three characteristics.