基于改进分辨矩阵的属性约简方法  

Attribute Reduction Based on Improved Discernibility Matrix

作　　者：黄治国[1] 杨晓骥 

机构地区：[1]河南工程学院软件学院,河南郑州451191 [2]武汉市教育科学研究院,湖北武汉430068

出　　处：《计算机仿真》2014年第9期305-309,共5页Computer Simulation

基　　金：国家自然科学基金资助项目(61300127);河南工程学院博士基金资助项目(D2013003)

摘　　要：在分辨矩阵的属性约简算法的研究中,需比较决策系统中各对象生成矩阵元素,导致所得分辨矩阵过于庞大,且造成较大的时间开销。为降低利用分辨矩阵求取属性约简算法的复杂度,依据条件等价类将原决策系统分解为一相容对象集与一非相容对象集,给出条件相对于决策的可辨识关系定义与改进的分辨矩阵定义,将条件相对于决策的可辨识关系变化作为属性约简的判定标准,结果证明改进分辨矩阵的属性约简与保持正域不变的属性约简等价。推理证明与仿真实例说明,改进方法的高效性与完备性。The objects in decision system are compared for generating the elements of discernibility matrix in the existing attribute reduction algorithms based on discernibility matrix. To reduce the time complexity of attribute reduction algorithms based on discernibility matrix, the original decision system is divided into a consistent object set and an inconsistent object set according to condition equivalence class, then an improved discernibility matrix is defined on this basis, and the fact is proved that attribute reduction based on this improved discernibility matrix is equivalent to the reduction which remains positive region. The reasoning proof and simulation example show the effectiveness and completeness of the proposed method.

关 键 词：粗糙集 决策系统 分辨矩阵 属性约简 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]