证据冲突下D-S融合算法的改进  

Improvement of D-S Fusion Algorithm under Evidence Conflict

作　　者：吕雪婷[1] 贾瑞生[1,2] 孙惠惠[1] 

机构地区：[1]山东科技大学信息科学与工程学院,青岛266510 [2]山东科技大学矿山灾害预防控制国家重点实验室,青岛266510

出　　处：《系统仿真学报》2013年第3期571-574,共4页Journal of System Simulation

基　　金：国家863计划重点项目(2009AA062704);山东省科技攻关项目(2009GG10001004);矿山灾害预防控制教育部重点实验室开放基金项目(MDPC0808)

摘　　要：针对D-S证据理论在处理高冲突证据时会出现悖论这一情况,在D-S理论框架下,给出冲突程度的判别方法,并提出一种改进的相对加权融合算法。该算法根据证据距离来确定证据的相互支持度,并以证据支持度矩阵模最大特征值对应的特征向量为证据的权重向量,并求得各证据的相对折扣因子,对各证据信息进行折扣处理后用D-S规则融合,消除了信息冲突环境下D-S合成法则悖论。为验证改进算法的有效性,给出了仿真分析,验证了该算法在证据冲突的情况下能有效地减小冲突对组合结果的影响,降低了决策风险。Aiming at the situation that appeared paradox when using D-S evidence theory to deal with the high conflict evidences, under the framework of the D-S evidence theory, the deciding method of conflicts was given, and an improved mutual weight combination algorithm was proposed. This approach made use of the evidence distance to determine the mutual support degree, and eigenvector for the maximal eigenvalue of evidence support degree matrix was considered to be weight vector. Then evidence discount coefficient could be gained, and it was used to every evidence, which was combined by D-S rule, and this avoided the paradox of D-S combination rule under the information conflict. In order to test the validity of this algorithm, simulation was given. And the algorithm validates that it can effectively decrease the effects that the conflict makes to combination results. And this algorithm reduces the decision-making risk.

关 键 词：信息冲突 D-S证据理论 证据距离 相对折扣因子 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]