基于信息熵的群组聚类组合赋权法

在多属性群决策方法的研究中,为了科学地确定专家的权重,提出一种基于信息熵的群组聚类组合赋权法。依据各个专家的判断矩阵归一化得到的排序向量,利用相关系数法构造相关矩阵。通过分析阀值变化率选取最优聚类阀值,对相似程度较高的排序向量给出合理的聚类。运用信息熵为类内专家赋权,综合聚类结果和排序向量的信息熵,确定专家的总权重。算例表明该方法可以对较为相近的专家评价结果进行有效分类,并准确衡量每位专家评价信息量的大小,能够有效提高专家赋权的合理性和群组决策的科学性。

Method for Combination Weighting Experts Based on Information Entropy and Cluster Analysis

In terms of the research of multi-attribute group decision-making, a method for combination weighting experts is put forward based on information entropy and cluster analysis so as to scientifically determine the weight of every expert. According to the experts' collating vectors obtained by normalization of corresponding judgment matrixes, correlation matrix is constructed by the correlation coefficient. Through the analysis of change rate of threshold, the optimal clustering threshold is selected and the higher priority vector similarity obtained the reasonable clustering. The experts' weight of within-class can be ascertained by the theory of information entropy weight. The experts' weights are determined according to the result of classification and information entropy of collating vectors. Finally, a numerical example shows that the method is effective for the higher priority vector similarity classification and can accurately weighing every experts' information. The method will effectively improve the rationality of determining experts' weight and contribute to scientific group decision-making.