证据推理冲突问题研究

 证据推理是一种重要的人工智能方法,广泛应用于数据融合、数据挖掘和智能决策。但 Dempster组合规则对于高冲突情况的处理不是十分合理的。首先研究了几种组合规则的改进方法和造成其缺陷的机理,给出一类组合规则的改进方法,基于有无先验给出两类算法。仿真结果表明新的证据推理组合规则在处理冲突方面的性能得到了明显的改进。

CONFLICT PROBLEM OF DEMPSTER-SHAFER EVIDENCE THEORY

Evidence reasoning has good performance in dealing w ith uncertain information in intelligent decision making systems. But, Dempster's rule is unsuitable to solve the combination with high conflict. In this paper, the authors firstly review many improved combination rules and analyze the reasons of drawbacks of Dempster's rule mentioned above. In addition, a new kind of method is proposed to solve the problem of high conflict. Two concrete algorithms are introduced with consideration of with a prior and without any prior. As a result, the new methods have better performance in dealing with conflict than other rules.