论文部分内容阅读
针对最小二乘支持向量机(LSSVM)在利用生产现场数据建模时易出现鲁棒性差的问题。提出了基于PSO的鲁棒最小二乘支持向量机建模方法,该方法利用一种改进的PSO方法确定LSSVM的惩罚参数C和核宽度?,增强了LSSVM对数据的适应性;通过给LSSVM优化问题中误差平方项赋予不同的权值,使得LSSVM在训练过程中克服了噪声的影响。最后将该方法应用于乙烯产品浓度预测,并与普通LSSVM进行了比较;仿真和实验结果表明,该算法建立的模型比普通LSSVM建立的模型具有更好的泛化能力和鲁棒性。
The least square support vector machine (LSSVM) is prone to poor robustness in the use of production site data modeling. This paper proposes a PSO-based robust least squares support vector machine modeling method, which uses an improved PSO method to determine the penalty parameter C and kernel width of LSSVM, and enhances the adaptability of LSSVM to data. By optimizing LSSVM, The square error terms in the problem are given different weights so that the LSSVM overcomes the noise impact during training. Finally, the method was applied to the prediction of ethylene concentration and compared with the general LSSVM. Simulation and experimental results show that the proposed algorithm has better generalization ability and robustness than the general LSSVM model.