Model predicts arrival time of ships

Date: 24 April 2014

Introduction: Predicting the arrival time of a ship is not self-evident. Because of unforeseen circumstances a ship often arrives later than the estimated time of arrival (ETA). This impedes the organisation of activities at seaport terminals and for hinterland transport. Claudia Pani (Faculty of Applied Economics) searched for a method to predict the ETA more accurately.

“The arrival time of ships is one of the main restrictions for the planning efficiency in a container terminal”, says Pani, who recently received her doctor’s degree from the University of Antwerp and the University of Cagliari (Italy). “Ships are under contract to report the estimated time of arrival (ETA) in advance, but unforeseen circumstances often make it impossible to respect the ETA. This leads to adverse consequences at the quay, amongst others for the dockworkers’ schedule, terminal planning and hinterland transport organisation.”

“I wanted to provide an instrument which would make it possible to make reliable predictions regarding the arrival times of ships in seaports. I therefore studied the arrival data of container ships in two European harbours, among which the Port of Antwerp. I compared different algorithmic models”, Pani explains. “The research put forward the Random Forest algorithms, which have a high probability rate for predicting the delay of a ship.”

The results of this research have a huge social added value. Pani: “By better predicting the arrival time, all operators in the maritime-logistic chain are able to better organise their workflow. This enhances the efficiency, which in turn leads to cost savings: a win-win situation for the entire sector.”