关键词: 置信规则库, 基于证据推理的置信规则库推理方法（RIMER）, 参数学习, 分类方法

Abstract: This paper proposes a new classification approach based on improved belief rule-base reasoning by introducing linear combinational mode, setting the number of rules based on the classifications and improving the method of calculating individual matching degree. Compared with the traditional belief rule-base inference methodology, the number of rules in the proposed method does not depend on the number of antecedent attributes or its referential values, and it is only related to classification number. In this way, the new method can ensure the applicability for complex problems. In the experiments, the differential evolution algorithm is applied to train parameters, including rule weights, attribute weights, referential values of antecedent attributes and belief degrees. Three commonly public datasets have been employed to validate the proposed method. And the classification results are proved to be ideal, which shows that the proposed method is reasonable and effective.

Key words: belief rule-base, belief rule-base inference methodology using evidence reasoning (RIMER）, parameter learning, classification method