Grouping objects by similarity using k-means