基于动态聚类和合理性差别的证据理论改进算法

An Improved Algorithm of Evidence Based on Dynamic Clustering and Rationality Discriminant

针对证据高度冲突时利用DS证据理论直接进行融合会得出不合理结果的问题,文中提出了一种新的证据融合方法。首先,利用动态聚类对证据进行分类,将含证据最多的类称为主类,其余类称为旁类。当旁类的证据个数与总的证据个数的比值小于或等于设定的阈值时,认为上述证据不合理并删除;当比值大于这一阈值时,则保留该旁类中的证据。然后,提出基于重心距离的方法确定剩余证据的可信度,求出期望证据,最后对其迭代融合。计算机仿真结果验证了该方法的有效性。

According to the problem that using evidence theory to fuse highly conflict evidence will get unreasonable results,this paper proposes a new method of evidence fusion.The method firstly classifies the evidence with dynamic cluster,the class containing most evidence is called main class,and the rest of the classes are side classes.When the ratio of evidence number of side class to total number of evidence is less than or equal to the fixed threshold value,the evidence of side class is considered unreasonable and should be deleted.If the ratio is greater than the threshold value,then keep the evidence.Then this paper determins the credibility of the residual evidence based on the method of gravity distance,finally we calculate out the expectation evidence and iteratively fuse it.Computer simulation results verify the effectiveness of the method.