Radial Basis Function Neural Network-based Port Throughput Prediction

In the progress of the economic globalization, the coastal ports are the very important and strategic resources in our country's economic cooperation and competition. The port industry produced more and more transport demand in recent years, which implied by the rapidly rising throughput, and the problem will also deteriorate. The capacity of our coastal ports and the berth construction are facing a very big challenge. The forecast of the port throughput that measures a port's position in the international trade is important and indispensable in many aspects, such as the port development, the investment of Infrastructure, the choice of berth address and the management strategy etc.The article analyzed the regulation of the transport demand because of the close relationship between the throughput and the demand. Combining the characteristics of the port production, it researched the demanding appearance of the port industry and the main factors that influence throughput. Then it respectively applied three models of the classic forecast method to predict the throughput of Da Lian port and compare the results.The Artificial Neural Network has very strong adapting ability and learning ability. Because of its good association, Generalize, analogy and reasoning, it is used in various fields including forecast. Many factors in port system are nonlinear, classic forecast method is difficult using nonlinear system, while ANN has the opposite character. For this reason, we can use intelligence method based on ANN into port industry. This article established two RBF networks based on the time series theory and the related factors. Given some non-quantity factors, it improved them with code, and expand the model. After the improvement of the RBF ANN, the accuracy of forecast results is raised, and this prediction is more reasonable.