Abstract: It is meaningful to approach new ways for achieving knowledge reduction in information systems with the huge volume of data. This paper gives distinguishable relation based on rough set theory. Against the new concepts, distinguishable matrix, distinguishable reduction and distinguishable core are presented. The main objective of this paper is to find and prove the relationship between the classical types of knowledge reduction, such as partition reduction and distinguishable reduction. The result shows, that distinguishable reduction is equivalent to partition reduction under all conditions, is easy to conclude. The relationships among other conceptions proposed are discussed. The judgment theorems and equivalent definitions with respect to these new concepts are obtained. New concepts and ways are introduced to figure out knowledge reduction in information systems, and the knowledge reductions based distinguishable relation is meaningful both in the theory and in applications.

Key words: data mining, rough set, knowledge reduction, distinguishable relation

摘要： 为寻求求解约简的有效方法，从而有效处理大规模数据，并减少后续挖掘算法在时间和空间上的压力，基于粗糙集理论提出可分辨关系的概念，并在此基础上定义对象差异矩阵、分辨约简集和分辨核心集等概念，证明划分约简这一传统知识约简与分辨约简的一致性，讨论其他概念间的关系，并给出相关的定理和等价命题。通过理论论证和示例分析，可以获知基于可分辨关系的属性约简的有效性和可 行性。

关键词: 数据挖掘, 粗糙集, 知识约简, 可分辨关系

CLC Number: 

• TP311.13