AUTOMATIC SCALE SELECTION FOR CLUSTERING OF CORRELATED HIGH-DIMENSIONAL DATASETS