监督邻域粗糙集下基于标准差属性重要度的启发式算法  

Heuristic Algorithm Based on Standard Deviation Attribute Significance in Supervised Neighborhood Rough Set

作　　者：刘旭东 LIU Xudong(School of Computer,Jiangsu University of Science and Technology,Zhenjiang 212100)

机构地区：[1]江苏科技大学计算机学院,镇江212100

出　　处：《计算机与数字工程》2025年第1期15-20,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61906078,62006099,62076111,62006128);江苏省高等学校自然科学基金项目(编号:20KJB520010);浙江省海洋大学数据挖掘与以应用重点实验室开放课题(编号:OBDMA202104)资助。

摘　　要：作为粗糙集理论中的重要模型,监督邻域粗糙集因其良好的区分性能而备受关注。然而,当在监督邻域粗糙集上进行约简时,因为重复计算候选属性的重要度,大多数的搜索都会导致巨大的时间消耗。为解决这一问题,提出了基于标准差属性重要度的启发式算法。该算法通过样本的离散程度对属性重要度进行排序,以此来减少计算约简过程中候选属性的遍历次数。12组UCI数据集上的实验结果表明,与其余三种算法相比,基于标准差属性重要度的启发式算法所求约简的分类性能得到了提高,同时时间消耗有着显著减少。As a significant model in rough set theory,the supervised neighborhood rough set has been paid much attention be⁃cause of its better discriminating performance.However,when deriving reduct in supervised neighborhood rough set,most research⁃es will cause massive time consumption because of repeatedly computing the significance of candidate attributes.To fill such a gap,a heuristic algorithm based on the standard deviation attribute significance is proposed.The attribute significance is sorted according to the degree of the dispersion of the samples,so as to reduce the traversal scale of attributes in the process of deriving reduct.Exper⁃imental results over 12 UCI data sets show that the proposed algorithm can effectively decrease the elapsed time of deriving reduct and improve the classification performance simultaneously over the other three algorithms.

关 键 词：加速策略 属性约简 属性重要度 标准差 监督邻域粗糙集 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]