TWO-STAGE VARIABLE CLUSTERING FOR LARGE DATA SETS