改进分辨矩阵下的增量式条件属性约简算法  

An incremental conditional attribute reduction algorithm based on improved discernibility matrix

作　　者：谭旭[1,2] 

机构地区：[1]深圳信息职业技术学院计算机应用系,深圳518029 [2]国防科技大学信息系统与管理学院,长沙410073

出　　处：《系统工程理论与实践》2010年第9期1684-1694,共11页Systems Engineering-Theory & Practice

基　　金：广东省自然科学基金(10451802904005327);国家自然科学基金(60904065)

摘　　要：通过分析现有分辨矩阵的不足,给出了基于条件熵的分辨矩阵定义,以得到不一致决策信息系统下的合理条件属性约简结果.基于改进分辨矩阵的定义,提出了相应的条件属性约简算法,并证明了其约简为不同于代数观点下的约简,而与信息观点下的约简相同.进一步,对该改进分辨矩阵做进一步修正,提出了一个增量式条件属性约简算法.实验结果表明:所提出的改进分辨矩阵下的条件约简算法总体上优于同类求解算法,而其增量条件约简算法能够获得稳定可靠的约简结果,同时约简效率得到较大的提高.In order to get reasonable conditional attribute reduction results from inconsistent decision information system, a new discernibility matrix based on conditional entropy is put forward based on shortcomings analysis of current discernibility matrix definitions. Then, an attribute reduction algorithm based on improved discernibility matrix is presented in great detail, and the reduction result is proved to be different from algebra one, but close to the information view one. Additionally, the above discernibility matrix is modified to present an incremental conditional attribute reduction algorithm. Experimental results prove that the proposed attribute reduction algorithm based on modified discernibility matrix has advantages over the previously proposed algorithms, and the corresponding incremental reduction algorithm can get reliable reduction results with higher efficiency.

关 键 词：分辨矩阵 条件熵 增量式 条件属性约简 粗糙集 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]