階層クラスター分析における最適なクラスター数の決定問題(一般講演J,数理2(応用))

抄録

We often use fuzzy graph to analyze inexact information such as sociogram structure ([1]and[2]). Concerning the hierarchical cluster analysis of a fuzzy graph ([3], [4] ,[5] and [11]), the number of clusters may have to be decided in the actual cluster analysis. In other word, we woud like to decide the optimal level with a partition tree. Concerning this problem, the steepest decent method in the multivariate analysis and the evaluation method based on Fuzzy Decision have been designed by us ([^*]). But, the steepest decent method could produce local limited solution problem. In this paper, we improved our methods by use of AIC to show that the new statistical scheme could find a statistically equilibrium optimal level of fuzzy clustering.