Abstract: Attribute reduction is one of the core issues in the rough set theory. This paper proposes a new uniform attribute reduction algorithm to solve the problems of existing attribute reduction algorithms in which the tradition difference matrix occupies too much storage space and the operation process requires too high memory. It uses segmentation to divide the original decision table into a number of new sub-decision tables. Moreover, it extracts the same attributes to construct feature matrix on which the data mining work is done. The analysis and the experimental results show this algorithm costs less time and reduces the number of condition attributes than existing algorithms.

Key words: segmentation, sub-decision table, uniform attribute reduction, feature matrix

摘要： 属性约简是粗糙集理论的核心内容之一，针对现有属性约简算法存在的差别矩阵占用存储空间过大，运算过程对内存要求过高等问题，提出了一种新的同属性约简算法。该算法采用分割技术将原始决策表分割为若干新的子决策表，对子决策表中的元素提取属性的共同特征组成特征矩阵，来替换传统的差别矩阵，并在特征矩阵上进行挖掘工作。理论分析和实验结果表明该算法具有较好的约简结果和更高的运算效率。

关键词: 分割技术, 子决策表, 同属性约简, 特征矩阵