改进型BP神经网络的港口吞吐量预测

为了提高港口吞吐量预测模型的适用性，满足港口决策的需求，对传统时间序列BP神经网络预测模型进行改进，将未来三年的吞吐量作为输出层参数，以tansig函数和logsig函数为传递函数，建立了改进型时间序列BP神经网络预测模型，利用trainlm函数训练神经网络，预测未来三年的港口吞吐量。对深圳港集装箱吞吐量进行了预测，结果表明，改进型时间序列BP神经网络模型泛化能力更强，拟合精度更高，且避免了传统预测模型循环预测产生的误差叠加，具有较好的适用性。

Port Throughput Prediction Based on Improved BP Neural Network

In order to improve the applicability of the port throughput prediction model and meet the requirements of port decision-making the traditional prediction model of time series based on BP neural network is improved.An improved prediction model of time series based on BP neural network is established by taking the throughput of the next three years as the output layer parameter.Meanwhile,the functions of tansig and logsig are used as the transfer function.The function of trainlm is utilized to train the neural network to predict the port throughput of the next three years.The forecasting of container throughput of Shenzhen Port indicates that the improved prediction model of time series based on BP neural network has stronger ability of generalization and higher fitting degree.As a result,it avoids the error superposition generated by the loop prediction of traditional model and has good applicability.