南京理工大学学报（自然科学版）2018，Vol.42Issue(1)：68-75,8.DOI:10.14177/j.cnki.32-1397n.2018.42.01.010
模糊粗糙集的稳定约简方法
Stable attribute reduction approach for fuzzy rough set
摘要
Abstract
Attribute reduction plays a core role in rough set theory.Presently,most of the results of such topic are based on the measurements such that classification performances,costs,uncertainties and so on.Those do not carefully take the fluctuations of reducts into account if data perturbations happen.To fill this gap,a heuristic framework for generating stable reduct is proposed.Firstly,multiple boundary sample sets are induced by multiple clusterings' technique.Secondly,the fused significance for each attribute can be computed using the multiple significances of such attribute obtained in all boundary sample sets.Finally,the attribute with greatest fused significance is selected and then added into the pool set.The proposed algorithm is tested on several UCI data sets and the experimental results indicate that by comparing with traditional heuristic algorithms,this approach can not only effectively improve the time efficiency for computing reduct and the stability of the re-duct,but also advance the classification stability based on the reduct.关键词
属性约简/数据扰动/模糊粗糙集/稳定性Key words
attribute reduction/data perturbation/fuzzy rough sets/stability分类
信息技术与安全科学引用本文复制引用
李京政,杨习贝,王平心,陈向坚..模糊粗糙集的稳定约简方法[J].南京理工大学学报（自然科学版）,2018,42(1):68-75,8.基金项目
国家自然科学基金(61572242,61503160,61502211) （61572242,61503160,61502211）
江苏省高校哲学社会科学基金(2015SJD769) （2015SJD769）
中国博士后科学基金(2014M550293) （2014M550293）
江苏省青蓝工程人才项目 （）
