基于BP-SA混合优化策略的铁路货运量时间序列预测

铁路货运量的时间序列预测可以看为一个从输入到输出的非线性影射。神经网络尤其是BP神经网络,被广泛用于非线性逼近问题。但是,BP算法训练神经网络速度慢、易陷入局部极值。而模拟退火算法(SA)具有很好的全局寻优性。因而提出混合优化策略,即将反向传播算法(BP)和模拟退火算法(SA)结合起来训练神经网络,来实现铁路货运量的时间序列预测。与单纯的BP算法比较,数值计算结果表明BP-SA混合优化策略具有较高的速度和精度。

Forecast on Temporal Sequence of Railway Freight Transport Volume based on BP-SA Mixing and Optimizing Solution

The temporal sequence of railway freight transport volume can beregarded as a mapping of nonlinear approximation from input to output. TheNN (neural network), especially the BP NN is widely applied in solving nonlinearapproximation issues. However, the BP algorithm provides a slow training toNN and is easy to fall into local extremum while the Simulant Anneal Algorithm(SA) gives a good performance in overall optimization searching. Hence thepaper raises the mixing and optimizing solution, i.e. to train the NN by combiningthe Back Prevalence Algorithm (BP) with SA in order to realize the forecast ontemporal sequence of railway freight transport volume. Comparing with singleBP algorithm, the value calculation result shows that BP-SA mixing andoptimizing solution has a higher speed and higher accuracy.