K-CLUSTERING AND THE DETECTION OF INFLUENTIAL SUBSETS