基于最小熵值相似矩阵构造方法及其聚类过程  

The method of construction similar matrices and its clustering process based on the minimum entropy

作　　者：魏书堤[1] 

机构地区：[1]衡阳师范学院计算机科学系

出　　处：《湖南科技大学学报（自然科学版）》2012年第3期72-75,共4页Journal of Hunan University of Science And Technology：Natural Science Edition

基　　金：国家自然科学基金资助项目(61070061)

摘　　要：针对利用相似矩阵进行聚类分析的分类问题,定义了相似矩阵及其性质,并给出相似矩阵的一些常见构造方法.针对现有构造方法缺少含义的问题,尤其是针对模糊问题,提出了一种基于最小信息熵值聚类构造相似矩阵的方法.该方法首先利用最小信息熵值原理获得多准则决策中不同属性的权重信息,然后求解不同方案集间的加权距离.通过构造正负不同的理想解,求解方案集与正负理想解直接的加权距离,并利用方案间加权距离与正负理想解整体距离之间的比例构造相似矩阵.算例表明该方法切实可行.For clustering problem based on similarity matrix, similarity matrix and its properties were defined and some common construction method were introduced. For existing similar matrix structure lack of practical implications, a method based on minimum information entropy was put forward. In this method, the weights of different attributes were attained by using the minimum information entropy principle and then the weighted distances among different scheme sets were solved. By constructing different positive and negative ideal Solution, the weighted distances between scheme and positive ideal or negative ideal were solved, then the similarity matrices were constructed with the proportion the weighted distances among different scheme sets to the weighted distances between scheme and positive ideal or negative idealThe example shows the effectiveness of this method.

关 键 词：信息熵 聚类 距离 相似矩阵 

分 类 号：C934[经济管理—管理学]