Feature Selection for Improving the Usability of Classification Results of High-Dimensional Data