Predicting Vessel Arrival Time in Inland Waterways: Integrating Eta, Ais, and Traffic Flow Data

Abstract

Accurate vessel arrival time (VAT) prediction is critical for port operational efficiency, as discrepancies between captains' estimated time of arrival (ETA) and actual arrival time (ATA) in inland waterway transportation (IWT) cause economic losses. Unlike ocean shipping, IWT faces unique challenges including fixed routes, high traffic density, and weather/lock delays. Existing research predominantly uses single data sources (e.g., AIS or port data) and focuses on maritime transport, leaving IWT understudied. To address this gap, we propose a novel data processing and prediction framework that integrates multisource data, including port call records, AIS trajectories, vessel physical dimensions, weather conditions, and waterway traffic flow, to improve VAT prediction accuracy. Using the Port of Rotterdam as a case study, we apply ensemble tree models including random forest, eXtreme Gradient Boosting (XGBoost), and light gradient boosting machine (LightGBM), with tailored data pre-processing and feature engineering. The results show that our model significantly improves prediction accuracy. When the VAT dataset is split based on time order, the model reduces the mean absolute error (MAE) by 79.55% (from 17.06 to 3.49 hours) and the root mean squared error (RMSE) by 63.22% (from 29.99 to 11.03 hours), compared to the ETAs reported by the vessels, demonstrating its effectiveness in handling real-world sequential data. Feature importance analysis reveals the estimated remaining sailing distance, reported ETA, and vessel dimensions as key predictors. This approach demonstrates significant potential for optimizing berth allocation, resource scheduling, and emission reduction in IWT systems, offering a data-driven solution to mitigate scheduling inefficiencies while supporting sustainable port management strategies.

Keywords: Maritime transportation, Vessel arrival time prediction, Inland waterway, Machine Learning, Maritime traffic flow