An Ensemble Approach for Extended Belief Rule-Based Systems with Parameter Optimization

The reasoning ability of the belief rule-based system is easy to be weakened by the quality of training instances, the inconsistency of rules and the values of parameters. This paper proposes an ensemble approach for extended belief rule-based systems to address this issue. The approach is based on the AdaBoost algorithm and the differential evolution (DE) algorithm. In the AdaBoost algorithm, the weights of samples are updated to allow the new subsequent subsystem to pay more attention to those samples misclassified by pervious system. And the DE algorithm is used as the parameter optimization engine to ensure the reasoning ability of the learned extended belief rule-based sub-systems. Since the learned sub-systems are complementary, the reasoning ability of the belief rule-based system can be boosted by combing these sub-systems. Some case studies about many classification test datasets are provided in this paper in the last. The feasibility and efficiency of the proposed approach has been proven by the experimental results.