基于偏差熵的专家聚类赋权方法  

Expert Cluster Weighting Method Based on Deviation Entropy

作　　者：王泽洲[1] 陈云翔[1] 蔡忠义[1] 林思铭[1] 

机构地区：[1]空军工程大学装备管理与安全工程学院,西安710051

出　　处：《火力与指挥控制》2016年第9期61-65,共5页Fire Control & Command Control

基　　金：总装"十二五"国防基金资助项目(51327020104)

摘　　要：在群决策专家聚类赋权过程中,可能出现专家给出的判断矩阵一致性比率与排序向量信息熵都相等但专家意见不同,却被赋予了相同权重的情况。针对上述问题,提出一种基于偏差熵的专家聚类赋权方法。该方法采用聚类分析的思想,基于比例构建相似系数,实现对专家群的分类;引入专家判断矩阵的一致性权重,并综合类容量构建权重指标来反映类别间的差异,确定专家类间权重;最后,在各专家类中建立偏差熵模型,依据类中专家达成一致性意见的贡献程度确定专家的类内权重,并得到专家的总体权重。具体算例表明,该方法可行有效。In order to settle the matter that the experts are given the same weight, which is caused by the case where the consistency ration of judgment matrix and information entropy of priority vector are equal although the expert＇s opinion is different in the process of group decision-making, an expert cluster weighting method based on deviation entropy is proposed. This method uses the clustering analysis principle to classify experts by constructing the similarity coefficient proportionally; then introduces the consistency weight of expert judgment matrix and synthesizes the class capacity which can define the weights of indexes to illustrate the difference between classes, and the experts＇ betweenclass weights are defined; finally ,the deviation entropy models are built for every experts classes to define the experts＇ within-class weights according to the experts＇ contribution level of consensus opinion, and the expert＇s overall weight is determined. The numerical examples show the effectiveness and feasibility of the proposed method.

关 键 词：群决策 聚类分析 偏差熵 专家权重 判断矩阵 

分 类 号：TB114.3[理学—概率论与数理统计]