Abstract: Attribute reduction is a kernel part of rough set theory. The existing attribute union definition and the relative attribute reduction algorithm can not obtain the minimum reduction over inconsistent decision table. Aiming at this problem, the improved attribute union and the attribute union reduction based on combination knowledge partition are defined, and their several useful properties are derived. Some relationships among attribute union reduction, positive region and negative region are studied. Theoretical analysis and example show that the minimum attribute reduction algorithm based on the improved attribute union is practicable and can get all the minimum attribute reduction for the consistent or inconsistent decision table.

Key words: rough set, inconsistent decision table, attribute union, minimum attribute reduction

摘要： 属性约简是粗糙集理论的核心内容之一。针对现有关系积理论不能求解不一致决策表中最小属性约简的缺陷，提出一种基于知识联合划分的改进关系积和关系积约简概念，分析了关系积约简、正区域和负区域之间的性质，提出一种基于改进关系积的最小属性约筒算法。理论分析和实例计算结果表明，该算法具有可行性，能求取一致或不一致决策表中所有最小属性约简。

关键词: 粗糙集, 不一致决策表, 关系积, 最小属性约简

CLC Number: 

• TP18