基于进化策略调节参数的回归支持向量机及其在铁水脱硫中的应用  

Parameter tuning for Support Vector Regression with evolutionary strategy and its application in hot metal desulphurization

作　　者：张美荣[1] 吕勇哉[1] 

机构地区：[1]上海交通大学自动化系,上海200030

出　　处：《冶金自动化》2006年第2期5-9,共5页Metallurgical Industry Automation

摘　　要：在回归支持向量机的建模中,参数调节问题一直是影响模型性能的重要因素之一。本文提出了一种基于进化策略的参数选择的新方法,并将它应用于铁水脱硫过程的建模上以预测铁水中的最终含硫量,其预测结果比神经网络的预测结果有一定改进。理论分析和应用结果表明,该方法是一种快速、简单、有效的调参方法。Parameter tuning of Support Vector Regression （SVR） has been a critical task to develop a SVR model with good generalization performance. A solution of SVR parameter tuning is presented with Evolutionary Strategy （ES）. It is called SVR-ES. The SVR-ES solution has also been successfully tested with a hot metal dcsulphufization process for the prediction of final sulphur content. The generalization performance generated by the proposed SVR-ES method is much better than the published results given by neural networks, it shows that the proposed SVR-ES solution is an effective and simple learning method to select the fittest tuning parameters for SVR with much less computation time.

关 键 词：回归支持向量机 进化策略 参数调节 神经网络 预测模型 含硫量预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] X758[自动化与计算机技术—控制科学与工程]