Reconsideration about clustering analysis

Abstract

Clustering analysis is a crucial part of pattern recognition, which refers to the procedure of pattern labeling, and this paper would like to give a survey on this technology. We notice that existing clustering methods are not of goodness when the clustering rules and constricts are not concerned comprehensively. The understanding of clustering analysis concept is another critical issue while designing a clustering method. Though various clustering techniques are presented literally in decades, aiming to solve different kinds of problems, clustering results are still far from that expected, and we should demonstrate the reason. Components of clustering methods are already introduced literally such as concept, clustering algorithms, data type, similarity measure, clustering tendency and validity.