Time series forecasting for port throughput using recurrent neural network algorithm

Abstract

ABSTRACT Container throughput is a critical factor to appraise a seaport performance and predicting this measure has played a vital role in port operations. Within the scope of effective decision making, predictive techniques have been presented for forecasting port throughput. Based on observing throughput variations in seasonal patterns and business cycles, the obtained data might help port authority to make better and more accurate decisions and improve seaport productivity. By applying predictive methods to the throughput data of Singapore and Busan port, the decision-makers can assess forecasting accuracy by measuring prediction errors. The numerical tests show that the echo state network (ESN) provides a high level of accuracy for predicting container throughput. As a result, the port managers could make use of this decision support strategy to foresee short-term plans for improving facilities, establishing effective cargo loading and unloading plans, consequently ensuring port productivity and profitability.