Improvement of initial centroid selection of K-means clustering algorithm