Forecasting Port Container Throughput with Deep Learning Approach

Fuxin Jiang, Gang Xie, Shouyang Wang

Abstract: Due to the international transfer of manufacturing industry, the change of trade policy and frequent irregular events in the global trade, it becomes more difficult to predict port container throughput accurately. In order to improve the predictive accuracy, we develop a bidirectional long short-term memory network model to forecast the throughput. Using the data of port in Qingdao, this study investigates for the first time how to use the deep learning approach to predict port container throughput. The empirical results show that the proposed model can achieve highest average predictive accuracy, which indicates that the approach is effective in the increasingly complex trade situation.