Abstract: Through the real rough set model based on consistent relation， the conceptions of the lower and upper approximation，boundary，B －positive region，B －negative，and B －borderline region were analyzed．Then attribute reduction
in real value rough set was presented，and the steps of attribute reduction were given．Successive search generalized important degree of subset，to see whether the generalized important degree is equal to that of the set．The feasibility is proved by the following example．

Key words: Real rough set , Interval number , Information decision system ,  Generalized importance degree , Attribute reduction

摘要： 通过对相容关系的实数粗糙集模型研究，分析了实数粗糙集的上近似，下近似，边界以及实数粗糙集的正域，负域，边界域的概念，并在此基础上得出了实数粗糙集的属性约简的方法，并给出步骤：逐次检索属性子集的广义重要度，看其是否与属性集合的广义重要度相等，当相等时即为其一个约简集。并通过算例证明该方法的可行性。

关键词: 实数粗糙集 , 区间数 , 信息决策系统 , 广义重要度 , 属性约简