基于多核LS-SVM的河涌水质预测模型

研究了多变量非线性河涌水质预测问题,提出了多核最小二乘支持向量机河涌水质预测模型。模型采用协同结构的非线性函数将水质时序样本映射到高维特征空间,进行多元线性回归。然后将该回归问题转化成半无限线性规划问题,运用交换集法求解。文章利用东江流域河涌水质数据进行了拟合预测实验,结果表明,与单核最小二乘支持向量机河涌水质预测模型相比,多核模型的预测误差减小了23%以上,它较单核模型具有更高的预测精度和更好的泛化推广性能。

Prediction of River Water Quality by Multiple Kernel Least Squares Support Vector Machines

For the purpose of predicting water quality of river system in the Pearl River Delta,this paper introduces a model called Multiple Kernel Least Squares Support Vector Machines,which is based on the support vector machines(SVM) and least squares support vector machines(LS-SVM).Multivariable and nonlinear problems in predicting water quality in the river system can be tackled by a multivariate linear regression model,and the problems are solved as a semi-infinite linear program problem.The case study in a river of the Dongjiang River basin showed that compared to the single kernel LS-SVM,the proposed model could reduce prediction errors by 23%,with better accuracy and generalization performance.