BP Neural Network-Based Prediction For Tianjin Port Throughput

An important index in comprehensive evaluation of port development is port throughput, which gives representation of level of port development, and determines ranking in ports of the world, port layout, investment scale, operation strategy and development strategy. In the process of port planning, prediction of port throughput is a key step, which relates to rationality of port planning and high effectiveness of port management, and ensures the development of the national economic. The growth of port throughput relates to the transportation ability of railway, highway, and waterway, also relates to investment, freight types, and development of neighboring port. Therefore, prediction of port throughput is complicated.In this paper, the development, infrastructure construction and throughput trends in Tianjin port are analyzed, and the relationship between the development of hinterland area of Tianjin port, gross domestic product and port throughput are illustrated. Based on BP Neural Network, according to port throughput of Tianjin port in 1982-2010, with consideration of value added in secondary industry of Tianjin port hinterland and investment, a prediction of port throughput is given out.By using gradient descent algorithm and Newton method—trainlm function, number of intermediate implicit layer in neural network are determined. This paper makes an improvement of standard BP neural network, by using adaptive learning rate moment gradient decreased method, compared improved BP neural network with standard BP neural network, the former has the advantages of fast convergence and good accuracy. And based on improved BP neural network, this paper also gives out prediction of port throughput of Tianjin port in 2011.