支持向量机与RBF神经网络回归性能比较研究.pdfVIP

计算机工程与设计 Computer Engineering and Design 4202 2011, Vol.32, No.12 计算机工程与设计 Computer Engineering and Design 支持向量机与RBF 神经网络回归性能比较研究 + 刘苏苏， 孙立民 (烟台大学计算机学院，山东烟台264005) 摘 要：支持向量机与RBF 神经网络相比各有优缺点，通过对支持向量机与RBF 神经网络的研究，从理论上分析了这两种 学习机在回归预测原理上的异同，通过仿真实验对比了两者在测试集上的逼近能力及泛化能力。仿真结果表明，对于小样 本集，支持向量机的逼近能力及泛化能力要优于RBF 神经网络。对实际应用中回归模型的选择问题提出了建议。 关键词：支持向量机; RBF 神经网络; 逼近能力; 泛化能力; 回归 中图法分类号：TP301 文献标识码：A 文章编号：1000-7024 (2011) 12-4202-04 Performance comparison of regression prediction on support vector machine and RBF neural network LIU Su-su, SUN Li-min+ (School of Computer Science, Yantai University, Yantai 264005, China) Abstract ：The support vector machine (SVM) and RBF neural network have their respective advantages and defects while using in re- gression prediction. By study of SVM and RBF natural network, the paper firstly analyzes the two kinds of learning machines used in regression theoretically. And secondly, the experiments are done to compare their approximation and generalization abilities. The ex- periment results show that the approximation and generalization abilities of SVM are better than RBF neural network for the small dataset. Then the selection recommendation of the regression model for the practical application is proposed. Key words ：SVM; RBF neural network; approximation; generalization; regression 理出发，分析支持向量机和RBF 神经网络在回归预测原理的 0 引 言 异同。然后分别采用支持向量机和RBF 神经网络进行回归预