Soft Clustering for Very Large Data Sets