Abstract: Attribute reduction is very important to knowledge acquisition in rough set theory. Considering the reality that many attributes values are uncertain in real information system, ordered concentration relation combined with gray system theory is expanded. A kind of preference relation that attribute values tend to the standard values  is established in gray information system. The significance of attribute is defined and the degree of dominance is used  as  heuristic information. On this basis, the algorithm of attributes reduction is proposed to make it suitable for the information system which attributes valued continuous gray number. A practical operation method for computing reduction is given and the feasibility of algorithm is verified via case.

Key words: gray system, ordered concentration, dominance relation, knowledge reduction, standard value

摘要： 属性约简是粗糙集理论进行知识获取的核心问题之一。针对现实信息系统中属性值取值不确定的情况，结合灰色系统理论对集中有序关系进行扩展，建立了灰色信息系统中趋于某个标准值的一种偏好关系，并以集中有序关系下的优势度为启发式信息，给出了属性的重要性度量，在此基础上提出了适合于属性值为连续灰数的信息系统的属性约简算法，给出了约简的实际操作方法，并通过实例验证了算法的可行性。

关键词: 灰色系统, 集中有序, 优势关系, 知识约简, 标准值