Cluster Analysis as a Tool of Interpretation of Complex Systems
Satoshi Miyamoto · IIASA PURE (International Institute of Applied Systems Analysis) · 1987
This paper deals with several problems in cluster analysis. It appears that the suggested solutions have not been considered in current literature. First, the author proposes the use of a permuted matrix as a tool for interpretation of clusters generated by hierarchical agglomerative clustering algorithms. Second, a new method of defining similarity between a pair of clusters is shown. This method leads to a new class of hierarchical agglomerative clustering. Third, two criteria are defined to optimize dendrograms that are outputs of hierarchical clustering. This paper has been presented at the Task Force Seminar Session on New Advances in Decision Support Systems, Laxenburg, Austria, November 3-5, 1986.