神经网络模型在港口吞吐量预测中的应用与误差分析

应用神经网络BP算法对杭州港的吞吐量预测实例进行了详细分析。通过对网络各种参数的调试与组合得出,当隐含层节点数为15,训练控制误差为0.035,分级迭代级数为4级,平滑因子参数为0.2,学习速率参数为1.5时,网络性能最佳。将网络预测结果与时间序列和回归分析2种方法进行了比较,得出神经网络方法在短期预测中要优于传统方法。通过对模型预测误差产生原因的简要分析,得出神经网络方法并不适用于吞吐量长期预测。最后对其应用过程中可能存在的一些问题提出了建议。

Neural network model applying in port throughput forecasting and error analysis

The BP arithmetic of neural network model is used to forecast the throughput of Hangzhou Port.After debugging and combination of several parameters in the network,the best model can be set up.When the number of hidden layer nodes is 15,the training control error is 0.035,the classification iteration series is 4,the smooth factor parameter is 0.2 and the learning rate parameter is 1.5,the performance of the network is optimal.Comparing the forecasting result with time series and regression analysis method,the advantages of neural network method in short-term forecasting can be elicited.After analyzing the production reasons of forecasting error,the conclusion that the neural network is not suitable for long-term forecasting can be educed.Finally the advice to the problems which the neural network method may encounter in application is given.