摘要
支持向量机是一类全新的小样本统计学习方法,它通过支持向量对样本进行分类或统计回归.将其应用于对非晶态聚氯乙烯的耐有机溶剂性能进行分类研究.选择74种溶剂(73种有机溶剂和水)的溶解度参数分量,即色散参数(δds)、偶极参数(δps)、氢键参数(δhs)为描述变量,采用径向基核函数,以留一法交互检验的识别率为目标函数进行支持向量分类.当选择SVM参数C=512及径向基核函数参数γ=0 5×10-3时,SVM对PVC耐蚀性能分类的模型识别率为94 59%,LOO识别率为91 89%.
Support Vector Machine (SVM)for classification and regression is a powerful technique based on statistical learning in small samples. In this study, a classification problem about the corrosion resistances of non-crystalline polyvinyl chloride (PVC) to organic solvents is investigated by the SVM technique. Selecting the solubility parameters of 74(73 organic solvent and water), namely dispersion component δ_(ds), polar component δ_(ps), and hydrogen bonding component δ_(hs) as independent variables and the corrosion-resistances of PVC as dependent one, the modes of corrosion-resistances are classified by the SVM with radial basis function (RBF). When the parameter C=512 in SVM and γ=0.000 5 in RBF kernel are used, the ratio of the samples correctly identified by the SVM to total samples is 94.59% and the ratio for LOO SVM is 91.89%.
出处
《桂林工学院学报》
2004年第4期474-479,共6页
Journal of Guilin University of Technology
基金
广西自然科学基金资助项目(桂科自0236063)
广西高校百名中青年学科带头人资助计划项目(桂教人[2003]97)
关键词
支持向量机
SVM
耐蚀性
聚氯乙烯
溶解度参数
support vector machine(SVM)
solubility parameter
corrosion-resistances
polyvinyl chloride(PVC)
