Multi Objective Genetic Algorithm Based Ensemble Classifier Using Classification Error, Sparsity, Diversity and Density Criterion

Ensemble classifier is an effective method in machine learning that attempted to provide a better approximation of an optimal classifier with combination of some classifiers results. To achieve better performance, the base classifiers should have acceptable efficiency and different classification error, also a suitable method used to combine their results. Various ensemble classification methods such as bagging, voting and strengthening methods have been presented. In this paper, we proposed the ensemble classifier based on weighted mean of the base classifiers output. The weights were estimated using a multi-objective genetic algorithm with taking classification error, sparsity, diversity and density criterion. The results of implementations on UCI datasets show that the proposed method causes more increasing classification accuracy related to other traditional ensemble classifiers.