A semi-fuzzy collaborative algorithm for cluster seeking

Abstract

In this paper, we present a semi-fuzzy collaborative algorithm for detecting the optimal number of clusters in a given data set of unlabeled objects. This algorithm is based on a measure of inter-points similarity that allows the detection and creation of clusters, plus a measure of ambiguity that allows collaboration between clusters during their formation. The algorithm also provides a matrix of optimized prototypes representing all the detected clusters. The performance of the proposed method is demonstrated through three examples of test data.