Freight Prediction Model Based on GABP Neural Network

Abstract

Back Propagation (BP) Neural Network has the ability of self-studying, self-adapting, fault tolerance and generalization. But there are some defaults in its basic application. Such as low convergence speed, local extremes and so on. So there are some limitations in practice. A quantitative forecast method based on the BP Neural Network improved by genetic algorithm (GA) is proposed in the paper. And the genetic algorithm is used to optimize the initial weights and threshold of BP network. The model is trained with the freight data of a city, and then it is used to forecast the freight. Form the comparison of simulated results of GABP network and these worked out by traditional BP network, it concludes that GABPNN has small error in forecasting. And it indicates that GA has the capability of fast learning the weight of network and globally search, in addition, the training speed of the improved BP network is greatly raised.