Predicting vessel arrival times on inland waterways: A tree-based stacking approach

Abstract

Estimating vessel arrival times is crucial for maintaining efficient transportation operations and ensuring the stability of the entire supply chain in maritime transportation. Vessel arrival uncertainty can lead to a domino effect of delays and inefficiencies throughout the supply chain. To address this issue and enhance the efficiency of maritime logistics, in this study, we introduce a novel methodology for predicting vessel arrival times in inland waterways. This methodology employs a tree-based stacking machine learning algorithm and considers factors such as vessel generic characteristics and inland waterway conditions, such as water depth and maritime traffic flow features. The model is validated using automatic identification system (AIS) data from the Dongliu section of the Yangtze River. The results reveal that the proposed model can achieve a prediction time accuracy of 90.5% for downstream vessels and 88.6% for upstream vessels, compared to their original arrival times. The model's outcomes also demonstrate the significant impact of water level and traffic volume on prediction accuracy, underscoring the potential of prediction models in enhancing the management of inland waterway transportation systems.