证据特征提取及其在证据理论改进中的应用

尽管DS证据理论可以有效地处理不确定性信息,但是在证据高度冲突情况下,经典DS证据组合规则因归一化过程往往会产生有悖常理的融合结果.为此,很多学者提出了许多不同的修改方法,可以归为两类修改组合规则和修改证据.在总结这两类修改思路的基础上,分析了几种代表性的改进方法,同时通过引入证据距离,对证据源进行特征提取,提出了相应的改进方案.理论分析和数值实验结果表明,改进后的3种方法不仅能充分体现相应的改进思想,而且可以更好地实现证据组合,较快地收敛真实目标,得到合理的融合结果.

Feature Extraction of Evidence and Its Application in Modification of Evidence Theory

The classical DS evidence theory involves counter-intuitive behaviours when the high conflict information exists. On the base of the two modification ideas, the modification of combination rules and the modification of evidence sources,this paper analyzed some typical existing modification methods and proposed to extract the intrinsic characteristics of the existing evidence sources by using evidence distance theory. With the application of the extracted feature, three modification methods were proposed. From the results of the numerical examples, the three proposed methods can not only realize the modification ideas, but also provide reasonable results with good convergence efficiency.