Empirical evaluation of K-Means, Bisecting K-Means, Fuzzy C-Means and Genetic K-Means clustering algorithms

Clustering is one of the most widely studied problem in machine learning and data mining. The algorithms for clustering depend on the application scenario and data domain. K-Means algorithm is one of the most popular clustering techniques that depend on distance measure. In this work, an extensive empirical evaluation of three significant variations of K-Means algorithm is carried out on the basis of six internal and external validity indices. It has been seen that performance of K-Means and Bisecting K-Means are similar, while Fuzzy C-Means gives better performance and Genetic K-Means performs the best. On the light of empirical result obtained in this paper, method for further improvement of the performance of Genetic K-Means is suggested.