Forecasting Cargo Throughput with Modified Seasonal ARIMA Models

The importing-and-exporting goods normally requires organizational cooperation among different international sectors ranging from purchasing, manufacturing, transporting, inventory, to distribution centers. For having their smooth coordination in the international trade, accurately forecasting the volume of imported-exported goods has become the core issue. This paper aims to obtain efficient forecasting models for estimating the cargo throughput at sea ports. We comprehensively compare the predicting performance of traditional autoregressive integrated moving average (ARIMA) models, Grey models, and their residual Fourier-modified models. The forecasting accuracy of the conventional models combined with the Fourier modification in residual is discovered to be significantly boosted. Two empirical studies of cargo and container throughputs collected from the Hong Kong and Kaohsiung ports were investigated, where two Fourier-modified seasonal ARIMA models, FSARIMA(4, 1, 4)(1, 1, 0)4 and FSARIMA(4, 1, 4)(0, 1, 0)12, are strongly suggested due to their superior forecasting powers. Finally, both models are further employed to provide the valid forecasts of two-port cargo and container throughputs in 2013. Key–Words: ARIMA model, Grey Model, Cargo throughput, Container throughput, Fourier modification, Hong Kong port, Kaohsiung port