MODIFIED SELECTION OF INITIAL CENTROIDS FOR K- MEANS ALGORITHM

Abstract

— Clustering is one of the important data mining techniques. k-Means [1] is one of the most important algorithm for Clustering. Traditional k-Means algorithm selects initial centroids randomly and in k-Means algorithm result of clustering highly depends on selection of initial centroids. k-Means algorithm is sensitive to initial centroids so proper selection of initial centroids is necessary. This paper introduces an efficient method to start the k-Means with good initial centroids. Good initial centroids are useful for better clustering.