基于灰色-反向传播神经网络的江西省公路货运量预测

鉴于反向传播（Back Propagation，BP）神经网络算法收敛速度慢、局部极小化、结构选择不一的问题,提出结合灰色关联度分析的BP神经网络方法进行公路货运量预测,提高了模型的非线性学习和泛化能力及预测精度。该预测模型以江西省为例,首先利用灰色关联度分析确定预测目标的影响因子；然后,将关联度强的第一产业、第二产业和人均GDP作为公路货运预测模型的自变量,公路货运量和自变量作为训练样本,BP神经网络模型通过正向计算传播,误差反向传播,训练神经网络；最后,该方法应用于江西省公路货运量实际预测中进行有效性验证,结果表明该方法非线性拟合效果较好,具有较高的预测精度。

Forecast of Highway Freight Volume in Jiangxi Province Based on Grey-BP Neural Network

In view of the problems of slow convergence speed, local minimization, and different structure choices of back propagation(BP) neural network algorithm, a BP neural network method combined with gray relational analysis was proposed to predict road freight volume, which improved the model''s nonlinear learning ability, generalization ability and prediction accuracy. The data from Jiangxi Provincial Statistical Yearbook was used in this forecasting model. First, the gray correlation analysis was carried out to determine the impact factors of the forecast target. Then, the first industry, the second industry and the per capita GDP of strong correlation were used as the independent variables of the road freight forecast model. The freight volume and independent variables were used as training samples. The BP neural network model was propagated through forward calculation, and the error was backpropagated to train the neural network. Finally, the method is applied to the actual forecast of highway freight volume in Jiangxi Province to verify the validity. The results show that the method has better nonlinear fitting effect and higher prediction accuracy.