一种有效的证据理论融合方法

An effective fusion approach to evidence theory

鉴于Dempster组合规则在处理高冲突证据时所存在的缺陷,提出一种新的基于余弦相似度和证据距离的融合方法.首先引入余弦相似度函数和距离函数,而后求出系统中各个证据间的联合相似度,并以此获得各个证据的加权系数,最后对系统中的证据加权平均后利用Dempster组合规则进行组合.与其他方法相比较,该方法不仅考虑了系统中证据在距离上的相似性,而且考虑了证据在方向上的相似性,融合后的结果更为理想.

In view of the defects of Dempster＇s combination rule available in treatment of highly conflict evidence,a new method was proposed based on the cosine similarity function combined with evidence distance.The cosine similarity function and distance function were introduced first.Then the joint similarity of all evidences in the system was found and the weighing coefficient of every evidence was obtained.Finally,the Dempster combination rule was adopted to realize information fusion after the evidence weighting coefficients were compared with other methods,averaged this method not only considered the distance similarity but also considered the direction similarity,so that the result from the fusion was more ideal.