Feature Selection for High-Dimensional Data