Managing uncertainty in ferry terminals: a machine learning approach

Abstract

Listen

Translate article

Ferry service across the Gibraltar Strait usually faces with the congestion problem at ferry terminals. Recognising the need to manage this problem, port managers must be prepared in advance to reduce waiting times, give space in the car park, coordinate ferry departures, etc. With this aim, we propose a machine learning methodology based on a classification and regression tree (CART) model. Thus, by means of the CART model, port managers can predict (with a certain error) the number of vehicles (or passengers) that will use the ferry terminal in the future. The accurate prediction that the model provides is crucial not only for port managers, but also for ferry operators. Our CART gives the predicted value and the measure of the expected error. Both are presented in sunburst graphs.