Abstract: The attribute importance measure and attribute reduction is one of the core content of rough sets theory. The classical rough set model based on equivalence relation, is suitable for dealing with discrete attribute values. Fuzzy-rough set theory, combining fuzzy set and rough set theory, extending equivalence relation to fuzzy relation, can deal with fuzzy attribute values. By analyzing three problems of FRAR which is a fuzzy decision table attribute reduction algorithm having extensive use, this paper proposes a new reduction algorithm which has better overcome the problem, can handle larger fuzzy decision table.

Key words: rough sets, fuzzy-rough set, fuzzy relation, similarity degree, dependency degree

摘要： 粗糙集理论研究的核心内容之一是属性重要性的度量和属性约简。经典的粗糙集模型基于等价关系，适合于处理离散属性值。模糊粗糙集理论将模糊集和粗糙集理论结合起来，将等价关系扩展为模糊关系，可处理模糊属性值。分析了已有广泛运用的模糊决策表的属性约简算法FRAR存在的三个问题，提出了一种新的约简算法，较好地克服了原算法的问题，能处理规模较大的模糊决策表。

关键词: 粗糙集, 模糊粗糙集, 模糊关系, 贴近度, 依赖度