A differential evolution algorithm based automatic determination of optimal number of clusters validated by fuzzy intercluster hostility index
Sourav De, Siddhartha Bhattacharyya, Paramartha Dutta · 2009
Automatic data clustering through determination of optimal number of clusters from the data content, is a challenging proposition. Lack of knowledge regarding the underlying data distribution poses constraints in proper determination of the inherent number of clusters. A differential evolution (DE) algorithm based approach for the determination of the optimal number of clusters from the data under consideration, is presented in this article. The optimum number of clusters obtained by the algorithm is further validated by means of a proposed fuzzy intercluster hostility index between the different clusters thus obtained. Applications of the proposed approach on clustering of real life gray level images indicate encouraging results. The proposed method is also compared with the classical DE (which operates with a known number of classes) and the automatic clustering DE (ACDE) algorithms.