基于粗糙集的数据约简方法研究  

Study on data reduction algorithm based on rough set

作　　者：黄治国[1] 王端[1] 

机构地区：[1]黄淮学院,河南驻马店463000

出　　处：《计算机工程与设计》2009年第18期4284-4286,4289,共4页Computer Engineering and Design

基　　金：河南省教育科学十一五规划基金项目(2008-JKGHAGH-413);湖南省教育厅科研基金项目(08C015);河南省科技攻关基金项目(0624220043);河南省自然科学基金项目(092300410217)

摘　　要：属性约简是粗糙理论的基础与核心研究问题。针对求取属性最小约简的NP问题,设计一种基于粗糙集的数据约简算法,寻求决策系统的近似最小约简。首先改进差别矩阵快速获取优化差别集,然后利用优化差别集中属性频度作为启发式信息,递归地调用算法寻找当前最重要属性,搜索过程中所有重要属性构成的集合即为最终所求约简。将该算法应用于电子商务用户访问模式数据浓缩,理论分析与仿真实例表明了该算法的可行性和有效性。The attribute reduction is fundamental and core issue of rough set. Aiming at the NP problem of acquiring the minimal reduction, a data reduction algorithm is designed to find out approximate minimal reduction： First the optimization discernibility set is acquired by improving discernibility matrix, then the algorithm is called recursively to seek the most important attribute currently by using attribute I^equency as heuristic information, finally all of the most important attribute form the reduction that we find out. The algorithm is applied to data enriching of user access pattern in E-commerce, the theoretic analysis and simulation instance shows that this algorithm is feasible and effective in practice.

关 键 词：粗糙集 属性约简 分辨矩阵 差别集 近似最小约简 

分 类 号：TP311.131[自动化与计算机技术—计算机软件与理论]