Testing of Diversity Strategy and Ensemble Strategy in SVM-Based Multiagent Ensemble Learning

In this study, a four-stage SVM-based multiagent ensemble learning approach is proposed for group decision making problem. In the first stage, the initial dataset is divided into training subset and testing subset for training and testing purpose. In the second stage, different SVM learning paradigms with much dissimilarity are constructed as diverse agents for group decision making. In the third stage, multiple single SVM agents are trained using training subset and the corresponding decision resulcts are also obtained. In the final stage, all individual results produced by multiple single SVM agents are aggregated into a group decision result. Particularly, the effects of different diversity strategies and different ensemble strategies on multiagent ensemble learning system are tested. For illustration, one credit application approval dataset is used and empirical results demonstrated the impacts of different diversity strategies and ensemble strategies.