Method of Accelerating K-means by Directed Perturbation of the Codevectors

Abstract

K-Means clustering algorithm is a simple and yet very powerful technique of partitioning data sets. This paper presents a method of decreasing the total iterations needed to run K-Means. This is done by adding perturbations to the cluster centroids and using the perturbed centroids as the seed values to compute the next codevectors. The use of this method significantly improved the performance of K-Means while preserving the quality of final cluster centroids.