Weighted least squares support vector machine local region method for nonlinear time series prediction

For the prediction of nonlinear time series, weighted least squares support vector machine (WLS-SVM) local region method is proposed in this paper. The method has the following two advantages. First, the WLS-SVM can obtain robust estimates for regression through the limited observation, and in the WLS-SVM framework, there is a simple and efficient approach to model parameters selection based on leave-one-out cross-validation. Second, considering the estimate of the given point, using all samples is unnecessary. Training a segment of samples, which are familiar with the given point, can achieve high quality precise. Our method has been tried for prediction on two synthetic and the neuronal data sets. The results show that the method has more superior performance than other methods like LS-SVM.