A new approach to the problem of clustering using the fuzzy joint points method

A new hierarchical approach to the problem of clustering, called the Fuzzy Joint Point, FJP) method is proposed. In the FJP method each element of the clusterized set is considered a fuzzy point of a multidimensional space. The concepts of a fuzzy conical point, fuzzy ?-neighborhood, and fuzzy ?-joint points are introduced and studies of certain properties relative to these concepts are carried out. The proposed algorithm of the FJP method may be used as a preparatory stage of the Fuzzy c-Means (FCM) algorithm for determining the initial classes and their dimensions and also as an independent clustering algorithm. © 2006 by Allerton Press, Inc.