一种快速属性约简算法  

Speeded attribute reduce algorithm

作　　者：丁军[1] 李凡[2] 冯嘉礼[3] 

机构地区：[1]北京科技大学管理学院,北京海淀100083 [2]华中科技大学计算机科学与技术学院,湖北武汉430074 [3]上海海运学院计算机系,上海200135

出　　处：《华中科技大学学报（自然科学版）》2006年第8期40-42,共3页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60075016)

摘　　要：基于差别矩阵的属性约简算法,先要求出差别矩阵,随着问题的规模增大,存放差别矩阵的空间和算法的执行时间过大和过长,让计算机难以承受.针对这一问题,提出了差别对象对集的定义,给出了基于差别对象对的属性约简定义,并证明了该定义等价于基于差别矩阵的属性约简定义.利用差别对象对,设计了一个新的属性约简算法.由于这一算法在求属性约简的过程中不用生成差别矩阵和大量的无用元素,因而大大减少了存储量和计算量,从而提高了算法的效率.计算实例说明了新算法的高效性.In the attribute reduction algorithm based on discernibility matrix, the discernibility matrix must be acquired firstly. But the space of storing the discernibility matrix in computer is very difficulty when the scale of the problem is very large. On the other hand, the computing cost with the algorithm is not good. The definitions of discernibility object pair set were provided. And the attribute reduction based on discernibility object pair was also provided. At the same time, it is proved that this definition is equal to the definition based on discernibility matrix. A new algorithm of attribution reduction based on the discernibility object pair was designed, and it need not create the discernibility matrix and not making those unused elements in the attribution reduction progress. So it cari cut down the computing and storing capacity greatly and the new algorithm efficiency is improved remarkably. The algorithm efficiency was illustrated with an example.

关 键 词：粗糙集 差别矩阵 属性约简 差别对象对 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]