Freight Volume Forecast of Wuhan City Circle Based on Wavelet Neural Network

Freight system is a complex system with multi factors coupling. The relationship among the factors of freight is staggered, showing strong randomness, uncertainty and nonlinearity. The wavelet neural network combines the self-learning ability of the neural network with the powerful detail feature of the wavelet, which has adaptive resolution and good tolerance. The freight volume of Wuhan city circle and eight important parameters which affect the freight volume taken as the basic data, the annual comprehensive index of the impact factor of the freight volume is obtained through the significant analysis of F statistics. Then a wavelet neural network is constructed with the enhanced flexibility by the introduction of translation factors and of the expansion factors. The parameters are optimized continuously until the whole network reaches the minimum error by training. So the network can extrapolate the future freight volumes. Through the test and analysis of the forecast results, the validity of the wavelet neural network used in the prediction of freight volume is proved.