What is the advantage of hierarchical clustering over K-means clustering?