Clustering Variation looks for a good subset of attributes in order to improve the classification accuracy of supervised learning techniques in classification problems with a huge number of attributes involved. It first creates a ranking of attributes based on the Variation value, then divide into two groups, last using Verification method to select the best group.