Sorted K-Means Towards the Enhancement of K-Means to Form Stable Clusters

Abstract

Clustering is a mechanism of arranging data points of a data set in groups which are similar. The similarity is found on the basis of some metric like Euclidian distance, density, etc. One of the major data clustering methods is the K-Means in which initial centroids are selected randomly. Here, we have presented an effective and modified algorithm, the Sorted K-Means which determines initial centroids after sorting the data points and provides stable clusters using lesser number of iterations. The tool used for implementation is Matlab to find initial centroids so as to reduce number of iterations for K-Means cluster formation. As a final result, stable clusters are obtained with minimized complexity.