基于试探折扣修正的证据组合方法

摘要

Dempster-Shafer (D-S) evidence theory plays a significant role in the related fields of uncertainty information processing .The combination of conflicting evidence using Dempster’s rule ,however ,often generates counterintuitive results .This paper presents a new quantitative standard which is constituted by the reliability of the evidence set and the convergence measure of the result can evaluate quality of the combination results ,and proposes a novel method ,which need do some tentative discount of ev-idences .Each tentative discount is constructed by the degree of not subordination .The reliability of the evidence set and the conver-gence measure of the result can increase together ,so the optimal convergence result can be obtained .By setting different goals ,cred-ibility of evidence set can be flexible controlled ,and the high-quality combination result can be obtained ,so it can meet the needs of all various types of decisions .The experimental results and the related analysis indicate that this method is rational and effective .%Dempster-Shafer（D-S）证据理论在不确定性信息处理相关领域具有十分重要的地位，然而对冲突证据进行 Dempster 规则组合时，常常会出现反直观结果的问题。本文提出了一种新的对证据组合结果质量评价的量化标准，该标准由证据集可信度与组合结果聚焦度构成，并在该标准的基础上提出了一种对证据进行多次试探折扣的修正方法，每次试探折扣由证据的不从属度来构造，实现了证据集可信度与组合结果聚焦度的共同提高，获得最佳聚焦结果，并且还可以通过设置优化目标，灵活控制证据集可信度，获得高质量证据组合结果，以满足各种类型决策的需要。实验结果和相关分析表明，本文方法是合理有效的。