基于权重邻域熵的数值型信息系统属性约简算法  

Attribute reduction algorithm of numerical information system based on weighted neighborhood entropy

作　　者：陈宝国 邓明 陈金林 CHEN Baoguo;DENG Ming;CHEN Jinlin(School of Computer Science,Huainan Normal University,Huainan 232038,Anhui,China;College of Electronic and Information Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,Jiangsu,China)

机构地区：[1]淮南师范学院计算机学院,安徽淮南232038 [2]南京航空航天大学电子信息工程学院,江苏南京211106

出　　处：《山东大学学报（工学版）》2024年第1期33-44,共12页Journal of Shandong University（Engineering Science）

基　　金：安徽省高校自然科学研究重点项目(KJ2018A0469,KJ2021A0972)。

摘　　要：在邻域粗糙集的属性约简中,每个属性被赋予相同的权重而不能更好地进行属性选择,针对这一问题,提出一种属性权重的邻域条件熵属性约简算法。通过条件属性与决策属性之间的相关系数评估条件属性的权重,基于权重方法提出一种改进的邻域关系,称为权重邻域关系,并提出相应的权重邻域粗糙集模型。以权重邻域粗糙集模型为基础,进一步提出权重邻域熵模型,理论证明权重邻域条件熵的单调性。通过权重邻域条件熵作为启发式函数提出一种新的数值型信息系统属性约简算法。试验结果表明,提出的属性约简算法具有更好的属性约简性能。In the attribute reduction of neighborhood rough set,each attribute was given the same weight and could not make better attribute selection.To solve this problem,a neighborhood conditional entropy attribute reduction algorithm with attribute weight was proposed.The weight of conditional attributes was evaluated by the correlation coefficient between conditional attributes and decision attributes.Based on the weight method,an improved neighborhood relation was proposed,which was called weighted neighborhood relation.The corresponding weighted neighborhood rough set model was also proposed.Based on the weighted neighborhood rough set model,the weighted neighborhood entropy model was further proposed,and the monotonicity of the weighted neighborhood conditional entropy was proved theoretically.A new attribute reduction algorithm for numerical information system was proposed by using the weighted neighborhood conditional entropy as the heuristic function.The experimental results showed that the proposed attribute reduction algorithm had better attribute reduction performance.

关 键 词：数值型信息系统 邻域粗糙集 属性约简 属性权重 邻域熵 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]