基于带RBF核的SVM模型对红酒品质的精准分类  

Accurate classification of red wine quality based onSVM model with RBF kernel

作　　者：李琴[1] 朱家明[1] 郎红 宋国锋 LI Qin;ZHU Jiaming;LANG Hong;SONG Guofeng(College of Statistics and Applied Mathematics, Anhui University of Finance and Economics, Bengbu 233030, China)

机构地区：[1]安徽财经大学统计与应用数学学院,安徽蚌埠233030

出　　处：《湖北大学学报（自然科学版）》2021年第4期417-422,共6页Journal of Hubei University：Natural Science

基　　金：国家自然科学基金(71974001);安徽财经大学教学研究项目(acjyyb2020011)资助。

摘　　要：针对红酒品质的“精准分类”和影响红酒品质的重要指标问题,综合带RBF(径向基函数)核的SVM(支持向量机)模型、递归特征删除、深层次回归分析及单因素方差分析等方法,构建支持向量机模型和作用效力挖掘模型,使用R和SPSS进行编程求解.研究结果表明:该模型的精确率高达74.01%,且其误差绝对值最大不超过3,并找出影响程度排名前三的分别为:酒精、密度、挥发性酸含量,影响程度分别为43%,20%和13%.并通过改变SVM模型函数的gamma参数数值以提高SVM模型准确率,再对模型精确率和不同误差值的数据数量进行赋权,构造线性择优模型,得分最高所对应的参数值即为最佳参数值,以此对SVM模型进行改进.Aiming at the“precise classification”of red wine quality and exploring important indicators that affect red wine quality,it integrates SVM(support vector machine)model with RBF(radial basis function)kernel,recursive feature deletion,deep regression analysis and single factor Variance analysis and other methods have constructed a support vector machine model and an effect mining model,and used R and SPSS for programming.The study concluded that the accuracy of the model is as high as 74.01%,and its error value is no more than 3,and the top three influence levels are found:alcohol,density,volatile acid content,and the influence degree is 43%,20%and 13%.In this paper,we constantly change the gamma parameter value of the SVM model function to improve the accuracy of the SVM model,weight the model accuracy and the number of data with different error values,and construct a linear optimal model.The parameter value corresponding to the highest score is the best parameter value to improve the SVM model.

关 键 词：精确分类 支持向量机 径向基函数 深层次回归分析 递归特征消除 单因素方差分析 R软件 SPSS 

分 类 号：TS262.6[轻工技术与工程—发酵工程]