Rapid Initial Clustering of Large Data Sets