中文摘要：

针对基于Pawlak和基于条件熵的属性重要性约简算法存在的局限性,提出了一种基于分辨矩阵的属性重要性约简算法。详细分析了这两类属性约简算法产生局限性的原因,据根属性在分辨矩阵中区分对象时所起的作用的情况,给出了一种基于分辨矩阵的属性重要性定义方法,并且通过该方法计算分辨矩阵中属性的重要性;最后按照属性重要性大小的顺序来依次添加属性到核集中,直至获取决策表的一个最小约简。实例分析表明,该算法能够有效找到最小约简,与其他法相比,当决策表中条件属性较多时,该算法能够大幅减少计算工作量。

英文摘要：

Aiming at the limitation of the importance of attribute reduction algorithm based on Pawlak and conditional entropy, Put forward a kind of importance of attribute reduction algorithm based on the distinguish matrix. Firstly, this paper has analyzed the causes of these two kinds of attribute reduction algorithm. Then, according to the attributes in differentiate matrix to distinguish the object, a definition of attribute importance method based on determine matrix was given, and used this method to calculate the importance of attribute in differentiate matrix. Finally, according to the size of the order of importance of attribute to add attributes to the nuclear concentration, and get a minimum of reduction. Example analysis shows that the algorithm can find the minimal reduction, Compared with other method, when more conditional attribute in the decision table, the algorithm can greatly reduce the amount of calculation.