Toward an Improved Clustering of Large Data Sets Using Maximum Common Substructures and Topological Fingerprints