基于可辨识矩阵的快速粗糙集属性约简算法  

Fast attribute reduction algorithm of rough set based on discernibility matrix

作　　者：薛安荣[1] 韩红霞[1] 潘雨青[1] 

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013

出　　处：《计算机工程与设计》2007年第20期4987-4989,4993,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(60373069);江苏省高校自然科学基金项目(05KJB520017)

摘　　要：Karno Bozi提出的CoreSearching算法在向约简中插入候选属性的时候,根据属性出现次数需要循环查找可辨识矩阵中的所有剩余项,直至矩阵为空,导致计算量较大和结果中冗余属性存在的可能。基于Core Searching算法提出通过给属性设立计数器的基于可辨识矩阵的快速属性约简算法,实例分析表明,该算法与CoreSearching算法相比,在计算量减少和循环次数减少的同时能得到更简约的结果,是一种快速、高效的属性约简算法。When core searching algorithm introduced by Karno Bozi extended the reduction, it needed to circularly search all the remanent items in discernibility matrix based on the appearance times of candidate attributes, until the matrix became empty, this induced the existence of redundant attributes in the result. The improved fast algorithm based on the discernibility matrix, it set up counters for different attributes. Theoretical analysis and instance results show that the algorithm, obtaining the more reductive result, requiring less computional effort than core searching algorithm, is a simple and effective attribute reduction algorithm.

关 键 词：可辨识矩阵 不可分辨关系 属性约简 核 粗糙集 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]