Deep learning models for vessel’s ETA prediction: bulk ports perspective

Abstract

Accurate vessels’ estimated time of arrival to ports is an important information to ensure efficient port operations management. At all stages of ports and vessels operations planning, arrival times are key milestones. Therefore, the variation of vessels’ arrival times affects port operations and causes disruptions along the global port chain. For this reason, intelligent systems are needed to predict vessels’ estimated time of arrival to speed up rescheduling operations in case of perturbation. This study addresses the problem of predicting bulk vessels’ estimated time of arrival to the destination port. For that, we propose an approach based on Deep Learning sequence models and using different data sources including the Automatic Identification System historical traffic data. This study shows how both recurrent and convolutional neural networks can leverage vessel historical voyage data to predict travel time to the destination.