Local reduction of decision system with fuzzy rough sets

One important and valuable topic in fuzzy rough sets is attribute reduction of decision system. The existing attribute reductions with fuzzy rough sets consider all decision classes together and cannot identify key conditional attributes explicitly for special decision class. In this paper we introduce the concept of local reduction with fuzzy rough sets for decision system. The local reduction can identify key conditional attribute and offer a minimal description for every single decision class. Approach of discernibility matrix is employed to investigate the structure of local reduction. At last, several experiments are performed to show that the idea of local reduction is feasible and valid.