Ensemble Learning with Local Experts

Ensemble learning methods have received considerable attention in the past few years. Various methods for combining several learning experts have been developed and used in different domains of machine learning. Many works have focused on decision fusion of different exports. Some methods try to train all the experts on the same training data and then use statistical techniques to combine the results so that the overall decision is of better accuracy. This paper presents a method in which the experts are not trained on the same data set, but rather they are trained locally with a subset of the training data. Behavioral partitioning is used here as the means to divide the problem space. Different methods are discussed for combining the results. Simple implementation of the method shows results comparable to those of similar methods.