基于可辨识矩阵的属性约简算法

属性约简是Rough集理论研究中的一个关键问题，已有的算法大致可以分为增加策略和删除策略2类，都是采用不同的启发式或适应值函数来选择属性。该文提出一种基于属性在可辨识矩阵中出现频率的新算法，以核为基础，不断从可辨识矩阵中选入出现频率最高的属性，直到可辨识矩阵元素集为空。为了得到Pawlak约简，算法增加了反向删除操作。实验分析表明该方法比其他方法快且有效。

Algorithm for Attribute Reduction Based on Discernibility Matrix

Attribute reduction is one of key problems in the theoretical research of Rough sets, and many algorithms have been proposed to study it. These methods may be divided into two classes (addition strategy and deletion strategy). This paper proposes a new algorithm based on attribute frequency in the discernibility matrix. In order to find optimum Pawlak reduction of decision tables, it adds the converse eliminate action until cannot delete. Tests indicate that the method performs faster than others.