基于组合相似度的混合多指标信息聚类分析方法

A Cluster Analysis Based on Combinational Similarity for Hybrid Multi-index Information

针对传统聚类分析只采用单一相似测度进行刻画的不足,提出了一种组合相似度的聚类分析思路.同时在实际问题中,由于聚类问题的复杂性、模糊性和不确定性,聚类信息常常是包含精确实数形式、区间数形式、模糊数形式等多种形式的混合型多指标信息,为此利用组合相似度的聚类思想对这种混合多指标信息进行了聚类分析.该聚类分析方法相对于传统单一方法更加全面,并考虑了混合指标信息.最后给出了一个算例,证明了所提方法的有效性.

Conventionally,a cluster is depicted with the measure of a single similarity among the objects in cluster analysis. The idea of combinational similarity is therefore proposed to get rid of the shortcoming. Because of the complexity,fuzziness and uncertainty of actual clustering problems,the clustering information is often a hybrid multi-index on which involves various forms of number,sucha as the precise real number,interval number and fuzzy number. A cluster analysis was made for the hybrid multi-index information via the idea of combinational similarity,and the results showed that the cluster analysis via the combinational similarity is more comprehensive than conventional ones because the hybrid index information is involved in analysis. A numerical example was given to illustrate the effectiveness of the proposed method.