Abstract: Attributes value reduction is one of the important parts researched in rough set theory.In order to correct the errors of computing reduction of the values in a decision table based on discernibility matrix,this paper puts forward an improved discernibility matrix definition together with a method for values reduction.It takes the new inconsistency into consideration caused in the values reduction.Therefore,the matrix discerns the rules that have the same values of decisive attributes with the inconsistent rules in order to avoid the possibility of including wrong rules after values reduction.

摘要： 属性值约简是粗糙集理论的重要研究内容之一,针对利用分辨矩阵求值约简的错误,提出了一种改进的规则分辨矩阵和值约简方法,主要考虑属性值约简可能导致新的不一致性问题。该矩阵区分与不一致规则的决策值相同的规则,从而避免了值约简出现错误规则的可能。