基于蚁群优化算法的支持向量机参数选择及仿真-中南大学学报.pdfVIP

第 39 卷第 6 期 中南大学学报( 自然科学版) Vol.39 No.6 2008 年 12 月 J. Cent. South Univ. (Science and Technology) Dec. 2008 基于蚁群优化算法的支持向量机参数选择及仿真 刘春波，王鲜芳，潘 丰 (江南大学 通信与控制工程学院，江苏 无锡，214122) 摘 要：基于支持向量回归机(SVR)模型的拟合精度和泛化能力取决于其相关参数的选取，以蚁群优化算法为基 础，给出支持向量回归机参数优化的一种新方法。该方法以最小化 k−fold 交叉验证误差为目标，对支持向量回归 机中的核参数 σ和惩罚系数 C 由蚁群系统中的节点值体现，数值的优选通过蚂蚁对最优路径的选择进行确定。计 算机仿真结果表明：与正交法、遗传算法等相比，该方法在参数优化方面有良好的鲁棒性能和较强的全局搜索能 力；该方法用于青霉素发酵过程的建模研究，建模精度较高。 关键词：蚁群算法；支持向量回归机；参数选择；优化 中图分类号：TP391.9 文献标识码：A 文章编号：1672−7207(2008)06−1309−05 Parameters selection and stimulation of support vector machines based on ant colony optimization algorithm LIU Chun-bo, WANG Xian-fang, PAN Feng (School of Communications and Control Engineering, Jiangnan University, Wuxi 214122, China) Abstract: Based on the fact that the regression accuracy and generalization performance of the support vector regression (SVR) models depend on a proper setting of its parameters, a new trial was applied in parameters selection. The new optimal selection approach of SVR parameters was put forward based on ant colony optimization (ACO) algorithm. The k-fold cross-validation error was used as the fitness function of ACO. The node values in ant system were reflected by the kernel parameter σ and regularization parameter C of SVR. The optimal values are equal to those of the best route which are decided by ants. Simulation results show that the optimal selection approach based on ACO has good robustness and strong global search capability. The method used for the research of modeling in the penicillin ferment process obtains higher accuracy. Key words: ant colony algorithm; support vecto