基于类间半径的自适应邻域特征选择算法  

Adaptive Neighborhood Feature Selection Algorithm Based on Inter-Class Radius

作　　者：徐久成 马妙贤 张杉 白晴 XU Jiucheng;MA Miaoxian;ZHANG Shan;BAI Qing(College of Computer and Information Engineering,Henan Normal University,Xinxiang 453007,Henan,China;Engineering Lab of Intelligence Business&Internet of Things,Xinxiang 453007,Henan,China)

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007 [2]智慧商务与物联网技术河南省工程实验室,河南新乡453007

出　　处：《昆明理工大学学报(自然科学版)》2025年第1期85-95,共11页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(61976082,62076089,62002103).

摘　　要：邻域粗糙集模型被广泛应用于特征选择领域,然而传统邻域粗糙集模型受限于网格搜索法,且存在仅从特征角度确定邻域的粒度和特征评价函数构造视角单一等问题.针对上述问题,提出一种基于类间半径的自适应邻域特征选择方法.首先,提出类间半径的概念,从样本角度与特征角度出发,为不同类的样本生成相应的邻域半径,构造了基于类间半径的自适应邻域粗糙集模型,并基于此定义了自适应邻域互信息.其次,由类间边界引出类间系数,并将其与自适应邻域互信息结合,进而构造了类间互信息这一特征评价函数,该函数从代数和信息论视角评价特征.最后,设计一种基于类间半径的自适应邻域特征选择算法.通过在8个UCI数据集上与5种算法进行实验对比分析.实验结果表明,所提算法在选择的特征数量和分类精度上优于其他算法.The neighborhood rough set model has been widely used in the field of feature selection.However,the traditional neighborhood rough set model is limited by the grid search method and has issues such as determining the granularity of the neighborhood solely from the feature perspective and the singular construction perspective of the feature evaluation function.To address these issues,an adaptive neighborhood feature selection method based on inter-class radius is proposed.Firstly,the concept of inter-class radius is introduced.From the perspectives of samples and features,corresponding neighborhood radii are generated for samples of different classes,and an adaptive neighborhood rough set model based on inter-class radius is constructed.Based on this,adaptive neighborhood mutual information is defined.Secondly,the inter-class coefficient is derived from the inter-class boundary and combined with the adaptive neighborhood mutual information to construct the inter-class mutual information as a feature evaluation function,which evaluates features from algebraic and information-theoretic perspectives.Finally,an adaptive neighborhood feature selection algorithm based on inter-class radius is designed.Through experimental comparison and analysis with five algorithms on eight UCI datasets,the experimental results show that the proposed algorithm is superior to other algorithms in terms of the number of selected features and classification accuracy.

关 键 词：自适应邻域 类间半径 类间系数 特征选择 邻域粗糙集 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]