Clustering is an unsupervised classification method, used in various disciplines, with a goal of organizing objects into different groups. Various methods and algorithms exist for clustering which attempt to find better clustering results. In clustering, choosing the centroids is a very sensitive concern as it is the essential element to do the clustering. This paper presents a novel partitioning clustering method that generates better clusters which is efficient in terms of Rand Index and Purity compared to traditional K-means algorithms. Experiments are performed on various datasets from UCI machine learning repository to obtain clusters and results show that the proposed method generates better clusters in terms of Purity and Rand Index.

Published version