Port Throughput Forecast Based on Nonlinear Combination Method

Port throughput forecast plays a great role in the development of port logistics economy. This paper works over that significant subject based on a new kind of nonlinear combination method, which is composed of linear exponential smoothing model, simple moving average method and Elman network. The forecasted values through the first two methods are taken as the inputs of Elman network, which are shown as medium and long-term effect, what's more, some recent periods of port throughputs are chosen as the inputs too, which are shown as short-term effect, thus different influences are considered comprehensively. Otherwise, the actual port throughputs in specific time periods are taken as the outputs. According to the historical samples, Elman network is trained in supervised manner for the future forecast. In this way, the different singular methods are combined together well and that new kind of nonlinear method shows higher forecasting accuracy.