基于粗集可辨识矩阵的属性约简算法  

An Approach for Attribute Reduction Based on Discernibility Matrix of Rough Set

作　　者：范敏[1] 刘文奇[1] 朱兴东[1] 

机构地区：[1]昆明理工大学系统理论与应用数学系,昆明650093

出　　处：《计算机工程与应用》2004年第13期79-80,130,共3页Computer Engineering and Applications

摘　　要：论文主要对RoughSet理论中的属性约简问题进行了研究。从Skowron可辨识矩阵出发,通过对属性重要性及属性间依赖性的讨论,利用条件熵构造了一种一致数据属性约简的启发式算法;通过实例和UCI数据库证明了该算法的有效性;并对文献犤4犦中的错误进行了修正。This paper mainly discusses the attribute reduction in Rough Set theory.Firstly,the paper researches the Skowron discernibility matrix,the significance and relevance of attributes.Meanwhile,we use the conditional entropy to propose a heuristic algorithm for attribute reduction;Then,the paper corrects the mistake in the reference.Finally,the experiment shows that it can get better effect and it also shows the reduction results of UCI database using this algorithm.

关 键 词：ROUGH SET理论 属性约简 可辨识矩阵 一致数据 条件熵 启发式算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]