不一致邻域决策信息系统的增量式属性约简研究  

Research on Incremental Attribute Reduction of Inconsistent Neighborhood Decision Information System

作　　者：王诚彪 王磊[1,2] 徐阳 张义宗 Wang Chengbiao;Wang Lei;Xu Yang;Zhang Yizong(School of Information Engineering,Nanchang Institute of Engineering,Nanchang 330099,China;Jiangxi Province Key Laboratory of Water Information Cooperative Sensing and Intelligent Processing,School of Information Engineering,Nanchang Institute of Engineering,Nanchang 330099,China)

机构地区：[1]南昌工程学院信息工程学院,南昌330099 [2]南昌工程学院信息工程学院江西省水信息协同感知与智能处理重点实验室,南昌330099

出　　处：《人工智能科学与工程》2023年第9期29-38,共10页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：江西省教育厅科技项目(GJJ219110);国家自然科学基金项目(61562061)。

摘　　要：属性约简是数据挖掘、机器学习等研究领域中的一个非常重要的预处理步骤,其效率的高低会直接影响到数据挖掘、机器学习等相关任务的性能。针对目前已有的非增量式属性约简方法在处理不一致邻域信息系统动态变化时无法高效更新属性约简的问题,提出一种在不一致邻域决策信息系统中对象集发生变化时的增量式属性约简方法。首先,该文以不一致邻域决策信息系统为研究对象,结合不一致邻域的特点给出了一种新的不一致度的表示方法。在此基础上用不一致度来表征属性重要度,以属性重要度为启发式信息研究不一致邻域信息系统对象集发生变化条件下邻域类以及不一致度的更新机理。随后,以不一致度为启发信息构建了增量式属性约简算法。进一步,在UCI上选取6个数据集,通过分类精度测试实验精选出各个数据集的最佳邻域半径δ,最后,利用最佳邻域半径δ在各个数据集上执行增量式属性约简算法实验,实验结果表明该文提出的增量式属性约简算法在保持分类精度不变的前提下较其他算法更加快速和有效。Attribute reduction is a very important preprocessing step in data mining,machine learning and other research fields.Its efficiency will directly affect the performance of data mining,machine learning and other related tasks.Aiming at the problem that the existing non-incremental attribute reduction methods cannot update attribute reduction efficiently when dealing with the dynamic changes of inconsistent neighborhood information systems,an incremental attribute reduction method is proposed when the object set changes in inconsistent neighborhood decision information systems.Firstly,this paper takes the inconsistent neighborhood decision information system as the research object,and gives a new representation method of inconsistency degree based on the characteristics of inconsistent neighborhood.On this basis,the inconsistency degree is used to characterize the attribute importance,and the attribute importance is used as heuristic information to study the update mechanism of neighborhood class and inconsistency degree under the condition that the object set of inconsistent neighborhood information system changes.Subsequently,an incremental attribute reduction algorithm is constructed with inconsistency as heuristic information.Further,six data sets are selected on the UCI,and the best neighborhood radiusδof each data set is selected by the classification accuracy test experiment.Finally,the incremental attribute reduction algorithm experiment is performed on each data set by using the best neighborhood radiusδ.The experimental results show that the incremental attribute reduction algorithm proposed in this paper is faster and more effective than other algorithms while maintaining the classification accuracy.

关 键 词：不一致邻域决策信息系统 不一致度 属性集 增量学习 粒计算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]