A MODIFIED K-MEANS CLUSTERING METHOD FOR EFFECTIVE UPDATING OF CLUSTER CENTROID

Abstract

Clustering is an unsupervised classification technique where a set of data, usually multidimensional is classified into clusters (groups) such that members of one cluster are similar to one another with respect to some predefined criterion. The k-means clustering methods are well known widely used partitioning-based clustering algorithms which minimize a given criterion by relocating points between clusters until a locally optimal partition is attained. We proposed a new k-means clustering method that is called modified k-means method which updates centroids (cluster centers) depending on if a point is added to a cluster or a point is removed from a cluster. The k-means clustering methods discussed in this research are the Forgys’ method, Lloyd’s method, MacQueen’s method, Hartigan and Wong’s method, Likas’ method, and Faber’s method. It was observed that the modified k-means method performed relatively better in terms of minimizing the total intra-cluster variance and their respective relative accuracy using simulated data and real-life data sets.
Keywords: Clustering; K-means methods; Cluster centroid; Euclidean distance; Minimum distance rule.