基于测试代价的三支邻域属性约简算法  

Three-way neighborhood attribute reduction algorithm based on test cost

作　　者：张欣蕊 万仁霞 岳晓冬[2] 陈瑞典 Zhang Xinrui;Wan Renxia;Yue Xiaodong;Chen Ruidian(College of Mathematics&Information Science,North Minzu University,Yinchuan 750021,China;School of Computer Engineering&Science,Shanghai University,Shanghai 200444,China;Institute for Big Data in Health Fujian Hongyang Software Co.,Ltd.,Fuzhou 350002,China)

机构地区：[1]北方民族大学数学与信息科学学院,银川750021 [2]上海大学计算机工程与科学学院,上海200444 [3]福建弘扬软件股份有限公司健康大数据研究院,福州350002

出　　处：《计算机应用研究》2024年第3期836-841,共6页Application Research of Computers

基　　金：国家自然科学基金资助项目(62066001,61662001);宁夏自然科学基金资助项目(2021AAC03203);中央高校基本科研业务费专项资金资助项目(FWNX04);北方民族大学研究生创新项目(YCX22088)。

摘　　要：针对粗糙集属性约简时很少考虑属性自身的测试代价等问题,提出了一种基于测试代价的三支邻域属性约简算法。算法根据各属性在邻域分辨矩阵中出现的频次和比例来计算属性重要性,并结合属性自身的测试代价来构造性价比指标,以此指导属性的甄选。三支决策方法被用于划分属性集,为属性的约简处理提供数据支撑。在7个UCI公共数据集上进行对比实验,结果表明,该算法可得到比对比算法更小的属性约简集合,在分类精度不降低的情况下,该算法具有更少的运行时间和更小的测试代价。基于财政收入的预测应用实例进一步证明了所提算法的有效性和实用性。In order to address the issue of test cost being rarely considered in rough set attribute reduction,this paper proposed a three-way neighborhood attribute reduction algorithm based on test cost.The proposed algorithm calculated the attri-bute importance according to the frequency and proportion of each attribute in the neighborhood resolution matrix,and combined the test cost of the attributes to construct the the cost performance index to guide the selection of attributes.Three-way decision-making method was employed to partition attribute sets,which provided data support for the attribute reduction process.Comparative experiments were conducted on seven UCI public datasets,which demonstrate that the proposed algorithm yields a smaller attribute reduction set compared to the comparison algorithm.Moreover,the proposed algorithm exhibited a shorter running time and lower test cost without compromising classification accuracy.Furthermore,it provided an application example based on fiscal revenue prediction to further validate the effectiveness and practicality of the proposed algorithm.

关 键 词：邻域粗糙集 邻域分辨矩阵 属性约简 测试代价 三支决策 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]