New algorithm for cluster analysis --Reduce the number of features to 1/100th to reduce computational cost--