An Efficient Centroid Selection Algorithm for K-means Clustering

This paper, we proposes an algorithm for performing data partitioning along the data axis with the highest variance to improve the accuracy. The data partitioning tries to divide data space into small cells or clusters where inter cluster distance are large and intra cluster distance are small as possible. Cells are partitioned one at a time until the number of cells equals to the predefined number of clusters, K. The centers of the K cells become the initial cluster centers for K-means. The experimental results shows that the proposed algorithm will be more effective and efficient converge to better clustering results than the existing clustering.