Nonlinear Time Series Prediction Model Based on Particle Swarm Optimization B-spline Network

Abstract

In order to improve the prediction accuracy of nonlinear time series, a prediction model based on particle swarm optimization B-spline network is proposed. In designing the structure of the network, the nodes of B-spline basis functions which are considered to be independent variables and every correlative weight parameter are to be optimized together in the network training process. And the forecasting error square sum is adopted to evaluate the training effect of the network. A particle swarm optimization algorithm with an appropriate search strategy is used as the training algorithm to search the distribution of optimal nodes of B-spline basis functions and find the optimal weight parameters, so that the structure of the network is optimized. Then, the nonlinear time series is predicted by the network. The simulation results indicate that the prediction model based on particle swarm optimization B-spline network has a fine generalization performance, and the algorithm optimizes the network effectively. The proposed prediction model is not only simple in structure, but also has higher prediction accuracy.