Robust Ensemble Based Algorithms For Multi-Source Data Classification

Abstract

In many classification problems, data are generated from different sources and views. Taking advantage of all the data available is important for intelligent decision making. Fusion of heterogeneous data sources underlying the same problem presents a natural fit for ensemble systems since different classifiers could be generated using data obtained from different sources and then combined to achieve the desired data fusion. Robust methods are proposed for combining classifiers, aimed at reducing the effect of outlier classifiers in the ensemble. The proposed methods are shown to have better performance leading to significantly better classification results than existing ones.