Attribute　reduction　is　an　important　issue　in　data　mining　and　knowledge　acquisition.　It　has　been　proven　that　computing　all　reductions　and　optimal　(minimal)　reduction　is　a　NP-hard　problem.　This　paper　proposed　a　hybrid　approach　using　the　rough　set　theory　and　neighborhood　systems　for　feature　selection.　Two　neighborhood　approximation　operators　are　defined　based　on　rough　set.　A　neighborhood　rough　model　is　constructed　subsequently　and　the　heuristic　information　is　introduced　according　to　the　significance　of　attributes　respectively.　Experimental　results　indicate　that　the　proposed　method　can　reduce　attributes　effectively.