关键词: 铁路货运量, 粒子群优化算法, RBF神经网络, 预测

Abstract: Railway freight volume demand prediction plays a significant role in national and regional economic development planning and transportation management decision making. In order to improve the prediction accuracy and convergence rate, a kind of prediction model which is based on RBF neural network is established. This kind of model has optimum approximation of function performance and global optimal characteristics, which is suitable for forecasting computation, meanwhile has problems about determination and optimization of parameters. A type of particle swarm optimization based on nonlinear learning factors adjusting（NLA-PSO） is introduced and applied to the parameters optimization of RBF neural network, which improves accuracy and efficiency of railway freight volume prediction. Comparison between simulation result and other algorithms is made by using examples of railway freight volume prediction during 1992 to 2011, which proves that the applied method has better prediction accuracy and convergence rate than other algorithms and has effective and feasible function on railway freight volume forecasting computation.

Key words: railway freight volume, Particle Swarm Optimization（PSO） optimize algorithm, Radial Basis Function（RBF） neural network, prediction