A new fuzzy clustering menthod with adjustable membership characteristics

Abstract

Clustering is an unsupervised procedure to group ob- jects in accordance with their similarities. For non- separable clusters, the concept of fuzziness is incorpo- rated. Among other approaches, the fuzzy c-means al- gorithm is the most well-known fuzzy clustering method. In this work, we present a modi¯ed form of the fuzzy c-means based on a new de¯nition of distance measure which can be considered as an extension of the conven- tional one. The key advantage of this new fuzzy cluster- ing schemem is its ability to °exibly control the mem- bership function curves. Analytical formulae have been derived for both cluster centers and the fuzzy partition matrix. Parameter e®ects related to the membership function curves have also been analyzed. Examples are given to demonstrate the clustering results of the newly presented scheme.