一种基于决策表的快速属性约简矩阵算法  

Matrix Algorithm for Quick Attribute Reduction Based on Decision Table

作　　者：钱文彬[1] 杨炳儒[1] 徐章艳[1,2] 

机构地区：[1]北京科技大学计算机与通信工程学院,北京100083 [2]广西师范大学计算机科学与信息工程学院,广西桂林541004

出　　处：《小型微型计算机系统》2012年第9期1954-1958,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60963008;60875029)资助;广西自然科学基金项目(2011GXNSFA018163)资助

摘　　要：针对目前基于决策表的属性约简矩阵算法效率不理想的情况,首先引入了简化决策表的概念,有效剔除了决策表中大量重复冗余的对象,并给出了正区域模型下简化矩阵的构造,从而有效地缩小了求解属性约简的搜索空间;然后从理论上详细分析了基于简化矩阵的属性约简与基于正区域的属性约简是一致的,并给出了一种有效的属性重要性度量方法.为此,在简化矩阵的基础上设出了一种快速的属性约简矩阵算法,最后通过实例分析和实验对比说明了算法的有效性和可行性.At present, the efficiency of matrix algorithm for attribute reduction is not ideal. Aiming at this situation, in this paper, firstly the concept of the simplified decision table is introduced ; a large number of repeated objects in decision table are deleted effec- tively. At the same time, the construction of the simplified matrix based on the model of positive region is provided, which can re- duce the searching space of attribute reduction. Furthermore, it is analyzed theoretically that the attribute reduction definition based on simplified matrix is equal to the attribute reduction definition based on positive region. And an efficient measurement method for at- tribute importance is presented. On this condition, a matrix algorithm for quick attribute reduction is designed. Finally, an example and experimental comparisons are employed to illustrate the efficiency and feasibility of the algorithm.

关 键 词：决策表 正区域 属性约简 矩阵算法 算法复杂度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]