Demurrage pattern analysis using logical analysis of data: A case study of the Ulsan Port Authority

Abstract

Maritime logistics, which accounts for around 80% of international trade around the world, has been a driving force for economic growth. Increases in maritime traffic, however, lead to increased congestion in berths and terminals. This congestion in turn negatively affects the total ship turnaround time and leads to decreased efficiency in port operations and a higher demurrage rate, which refers to the number of vessels in queue for more than a fixed time period waiting to load/unload out of the total number of vessels entering a port. The demurrage rate directly affects a port’s operating profits; thus, this rate needs to stay as low as possible. In this study, we focus on developing a methodology to address the demurrage rate of a port. To this end, we first collect two sets of vessel data (2016 annual data for training and 2019 annual data for validating) for ships entering and leaving the Port of Ulsan in the Republic of Korea and integrate these datasets with berth data and weather data. We tailor the logical analysis of data (LAD) technique to derive the patterns from the training data that mitigate or aggravate the demurrage rate. We use these patterns to predict the demurrage rate for the validating set of data. The overall binary classification results demonstrate the proposed LAD technique’s competitive performance, compared with other state-of-the-art machine learning methods. We then analyze the patterns to derive policy suggestions that can lower the demurrage rate at the Port of Ulsan. Our computational experiments find that the availability of tugs or pilots and port arrival times mainly affect the demurrage rate at the Port of Ulsan. Finally, our study showcases new possibilities for using patterns of demurrage and non-demurrage vessels obtained by LAD to help policymakers and port operators address the growing demurrage problem.