Learning-Driven Berth Allocation Optimization With Port Authority Behavior

Abstract

This study presents a two-phase framework that integrates machine learning with optimization modeling considering port authority behavior. First, a machine learning model predicts port authority allocation behavior based on historical data. These predictions capture the port authority’s scheduling behavior which reflects implicit operational patterns by predicting vessel waiting times. Second, these predicted behaviors are embedded into an optimization model through explicit constraints that ensure berth assignments closely align with berth authority tendencies. Using real operational data, we demonstrate the effectiveness of the proposed approach against a conventional optimization approach that does not take port authority behavior into account. Sensitivity analysis indicates that the proposed model effectively incorporates port authority behavior by varying the port authority behavior rate parameter.