Self -adaptive parameter optimization approach for least squares support vector machines

Based on radial basis function (RBF) kernel, a new self-adaptive method to optimize the least squares support vector machines (LS-SVM) parameters, the width of kernel parameter sigma and the LS-SVM regularization parameter gamma are proposed. Detailed methodology steps of this algorithm method are presented. Compared with back propagation neural networks (BPNN), various simulation experiments for nonlinear function estimation are carried out. The results show that this prediction model can achieve higher identification precision with a reasonably small size of training sample sets and has high generalization performance.