基于RBF核的SVM分类应用研究  

The Classification and Application of SVM Based on the RBF Kernel

作　　者：朱芳[1] 赵庆平[1] 王江涛[1] 

机构地区：[1]淮北师范大学物理与电子信息学院,安徽淮北235000

出　　处：《唐山学院学报》2017年第3期13-17,39,共6页Journal of Tangshan University

基　　金：安徽省教育厅高校自然科学研究一般项目(KJ2016B007);安徽省高校自然科学研究重点项目(KJ2016A650)

摘　　要：支持向量机(Support Vector Machine,SVM)在解决小样本、非线性及高维模式识别中具有优势,但核函数的选取没有定论,且其参数对SVM模型的性能起重要作用。针对这些问题,文章建立了基于SVM的分类模型,并通过UCI数据集验证了径向基核函数(Radial Basis Function,RBF)较其他核函数的有效性,其中核参数的选取采用改进的网格搜索法进行寻优。分类实验结果表明,选择RBF核函数的分类准确度较其他核函数提高了2.5%到35%。The support vector machine （ SVM） has some advantages in the small-sample, nonlinear and high dimensional pat tern recognition, but the selection of kernel function is not conclusive, and its parameters have an important influence on the performance of the SVM model. To solve these problems, the authors of this paper established a classification model based on SVM, and verified the greater effectiveness of the radial basis function （Ra-dial Basis Function, RBF） than th a t of other nuclear functions through the analysis of UCI data sets and kernel parameters were determined with the improved method of grid search. The experiment results show th at the classification accuracy of RBF kernel function has been improved by 2. 5 % to 35 % in comparison with other kernel functions.

关 键 词：支持向量机 径向基核函数 网格法 分类 

分 类 号：O234[理学—运筹学与控制论] TP273.22[理学—数学]