Abstract: In the research of knowledge acquisition based on rough set theory，attribute reduction is a key problem.In this paper，combining with the idea of divide and conquer method，an intact decision table is decomposed into two sub-decision table with same condition and decision attributes，according to partition of its universe.Then，attribute reduction results of the two sub-decision tables are computed in terms of classic attribute reduction algorithms.Finally，the two sub-decision tables are merged into the intact decision table by merging reduction algorithm presented，and the reduction process of original decision table is implemented.Experimental results demonstrate the validity of algorithms.

Key words: rough set, attribute reduction, condition class set, divided and conquer

摘要： 在基于粗集理论的知识获取研究中，属性约简是最核心的工作之一。结合分治法的思想，从论域划分的角度将一个大的决策表分解成两个子决策表，并利用经典的属性约简算法计算两个子决策表的约简，在此基础上利用合并约简算法将这两个子决策表合并，并求出原问题的解。该方法为解决大数据集的属性约简提供了一个新的途径。实验说明了算法的有效性。

关键词: 粗集, 属性约简, 条件类集, 分治