A new framework of fuzzy clustering algorithm

Abstract

In this paper, a novel data clustering algorithm based on the subtractive clustering (SC) algorithm and a new validity index are proposed. The SC algorithm is a simple method for data clustering; however, it has two problems which must be overcome. The first problem is such that the cluster centers found by SC are taken from data with the highest potential values, but that this data may not be the real cluster centers. The second problem is such that the cluster number generated by the SC algorithm is influenced by a predefined parameter. The proposed algorithm is based on distance relations between data and centers and is designed to ascertain the real centers generated by the SC algorithm. In addition, a novel robust cluster index is proposed to identify the real cluster number generated by SC algorithm.