基于NMI-SC的糖尿病混合数据特征选择  

Feature selection of diabetes mixed data based on NMI-SC

作　　者：朱潘蕾 容芷君[1] 但斌斌[1] 代超 吕生 ZHU Panlei;RONG Zhijun;DAN Binbin;DAI Chao;LV Sheng(Key Laboratory of Metallurgical Equipment and Control of Ministry of Education,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]武汉科技大学冶金装备及其控制教育部重点实验室,湖北武汉430081

出　　处：《电子设计工程》2024年第11期6-10,共5页Electronic Design Engineering

基　　金：国家自然科学基金资助项目(51475340);武汉市科技局企业技术创新项目(201901070211288)。

摘　　要：针对糖尿病预测精度受高维混合数据影响的问题,提出基于NMI-SC的糖尿病特征选择方法,通过邻域互信息(NMI)计算混合属性特征邻域半径内的联合概率密度,构建相似度矩阵,通过糖尿病特征之间的相似性构建无向图,基于谱聚类(SC)将糖尿病特征切分为多个特征相似组,实现非线性特征间的聚类,根据特征分类重要性选出相似组中的代表特征。并将其与原始特征集在支持向量机分类器上的准确率进行比较,该特征选择方法在删除46个冗余特征后,准确率提高了13.07%。实验结果表明,该方法能有效删除冗余特征,得到糖尿病分类性能优异的特征子集。Aiming at the problem that the prediction accuracy of diabetes is affected by high-dimensional mixed features,a diabetes feature selection method based on NMI-SC is proposed.The joint probability density within the neighborhood radius of mixed attribute features is calculated through Neighborhood Mutual Information(NMI),and a similarity matrix is constructed.An undirected graph is constructed through the similarity between diabetes features,and it is divided into multiple feature similarity groups based on Spectral Clustering(SC)to realize the clustering among nonlinear features.Representative features in similar groups are selected according to the importance of feature classification.Compared with the accuracy of the original feature set on the support vector machine classifier,the accuracy of this feature selection method is improved by 13.07%after deleting 46 redundant features.Experimental results show that this method can effectively delete redundant features and get a feature subset with excellent diabetes classification performance.

关 键 词：特征选择 混合数据降维 邻域互信息 谱聚类 

分 类 号：TN911[电子电信—通信与信息系统]