结合一致性准则的粗糙集属性约简算法  

Algorithm of rough set attributes reduction integrated with the consistency criterion

作　　者：吴敏[1] 张崇巍[1] 

机构地区：[1]合肥工业大学电气与自动化工程学院,安徽合肥230009

出　　处：《合肥工业大学学报（自然科学版）》2006年第7期851-855,887,共6页Journal of Hefei University of Technology：Natural Science

摘　　要：文章从属性选择角度对粗糙集属性约简概念进行扩展,在属性约简综合多重属性选择标准方面做了初步的研究。将属性类内一致性选择标准与粗糙集属性约简结合,以获得抗噪的约简;设计了一种全局最优算法和一种次优filter算法,从分类正确率、最简性及类内类间距离方面对一致性判据JC进行了测试,实验结果显示一致性高的属性约简集优于其他约简集。This paper extends the notion of attribute reduction in the context of feature selection and suggests integrating feature selection criteria into the algorithm of attributes reduction so as to obtain high quality attribute subsets. Two algorithms for attributes reduction integrated with the consistency criterion are presented. Experimental results shows that consistent attribute subsets outdo other reducts.

关 键 词：粗糙集 属性约简 分支定界 filter算法 一致性 

分 类 号：O235[理学—运筹学与控制论]