Department of Physics, Nankai University, Tianjin 300071, China
Abstract:
The least squares support vector machine (LS-SVM) isused to study the nonlinear time series prediction. First, theparameter γ and multi-step prediction capabilities of theLS-SVM network are discussed. Then we employ clustering methodin the model to prune the number of the support values. Thelearning rate and the capabilities of filtering noise for LS-SVMare all greatly improved.