AN IMPROVED K-MEANS ADAPTIVE CLUSTERING METHOD ON LARGE DATASETS