K-means clustering algorithm based on coefficient of variation