一个基于差别矩阵的属性约简改进算法  

An Updated Algorithm for Attribute Reduction Based on Discernibility Matrix

作　　者：程京[1] 朱靖[1] 张帆[2] 

机构地区：[1]湖南大学软件学院,湖南长沙410082 [2]湖南大学电气与信息工程学院,湖南长沙420082

出　　处：《湖南大学学报（自然科学版）》2009年第4期85-88,共4页Journal of Hunan University:Natural Sciences

基　　金：湖南省科技计划资助项目(2007JT1024)

摘　　要：介绍了决策表的基于差别矩阵的属性约简方法,针对AM-RASR约简算法的不足,将叶东毅的对不相容决策表的求核思想融合进来,提出一个改进的算法,新算法通过在扫描数据过程中用冲突对象对差别集进行修正,可以有效地减少数据的存储量和计算量,并能适用于不相容决策表.最后通过一个UCI数据集的实验说明改进的算法是有效的.Attribute reduction is one of the most important content of the Rough Set Theory. This paper analyzes attribute reduction for decision table based on discernibility matrix and YE＇s thought for calculating the core of inconsistent decision tables. As AM-RASR algorithm needs too much storing capacity and doesn＇ t support inconsistent decision tables, and YE＇s method needs much more computing, this paper presents an improved algorithm. Through revising discernibility collection with conflicting element when scanning data, which can cut down the original algorithm＇s computing and storing capacity, and can be applied in inconsistent decision tables. Fianlly, through an experiment with UCI data sets, the paper explains the effectiveness of the improved algorithm.

关 键 词：约简 差别矩阵 粗糙集 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]