和声搜索最小二乘支持向量机预测模型及其应用

为了改进目前最小二乘支持向量机(LSSVM)参数选择的盲目性,将和声搜索(Harmony Search)算法引入到最小二乘支持向量机中来.利用具有全局优化功能的和声搜索算法对LSSVM中正则化参数γ和核函数参数σ的进行自动优选,提出了和声搜索最小二乘支持向量机(Harmony Search Least Squares Support Vector Machine,HS-LSSVM)算法.通过对丰满大坝位移的建模预测并和BP神经网络模型及传统统计回归模型的分析比较,表明HS-LSSVM模型具有更小的预测误差和更高的预测精度.

Prediction model based on least squares support vector machine with harmony search and its application

Aimed at the blindness of parameter selection and kernel function of LSSVM,a new heuristic global optimization algorithm: harmony search algorithm was employed to realize the adaptive selections of regularization parameter γ and kernel function parameterσ,and a novel fusion algorithm: Harmony Search Least Squares Support Vector Machine (HS-LSSVM) was brought forward. According to the measured field data of dam,the HS-LSSVM-based prediction model was applied to forecast the deformation of dam. Compared with BP neural network model and statistical regression model,we can conclude that HS-LSSVM based prediction model can possess smaller prediction error and higher prediction precision. Therefore,the HS-LSSVM prediction model can realize the automatic selection of model parameters and improve the training efficiency and the prediction precision.