一种基于ISODATA聚类和改进相似度的证据推理方法

针对智能信息处理中Dempster组合规则不能处理高度冲突的问题, 从内、外证据不确定性分析的角度深入揭示了证据冲突产生的原因, 即证据的冲突性不仅仅根源于证据间的矛盾, 也与证据自身的不确定性密切相关, 提出了一种同时考虑证据自冲突和外部冲突的相似性测度, 然后利用新测度计算证据的众信度, 对证据源进行修正;与此同时, 根据原始证据间的聚类特性, 利用迭代自组织数据分析技术(Iterative selforganizing data analysis techniques algorithm, ISODATA)聚类方法进行聚类, 然后利用Dempster组合规则合成每一聚类中所有证据为证据代表, 并综合众信度和证据在该聚类的频度计算可靠度, 最后, 利用统一组合规则合成证据代表.并通过大量的算例, 同其他方法和自身改进前后进行深入比较, 优势比较明显, 有效地解决了冲突证据合成出现的问题.

A Method of Evidence Reasoning Based on ISODATA Clustering and Improved Similarity Measure

For the problem that the highly conflictive evidences can not be processed by Dempster rule in intelligent information processing, the reason of conflict occurrence is further revealed by analysis of uncertainty among evidences and evidences themselves. A similarity measure by considering self-conflict and external conflict of evidence together is proposed and used to compute the commonality of evidence. All initial evidences can be revised according to the commonality. At the same time, according to the clustering characteristic of initial evidences, they are clustered by applying ISODATA (Iterative selforganizing data analysis techniques algorithm), and then, all evidences in each cluster are combined as a representative of evidences by applying Dempster rule. The reliability of the representative of evidence is computed by considering the credibility and frequency of occurrence of the evidence in a cluster together. Finally, these representatives of evidences are combined by using the unified combination rule. Several numerical examples are given to deeply compare this method with others and itself before improving its similarity measure. The result shows that the new method has a distinct advantage over others and can effectively solve the problem of combination of conflictive evidences.