I want to use rough set theory for features reduction, sometimes the features after reduction give me the same accuracy before reduction with some databases and lower accuracy with other databases where the size of origin features equal 256 or more. but when I testing data by rough set function to find upper and lower approximation and boundary it always give me the upper approximation equal lower approximation (Crisp situation) with bound equal zero, so can apply the rough set on data for reduction if i get the these results.