The prediction of vessel arrival time (VAT) is crucial for port operations, which depend on vessel-reported estimated arrival time (ETA) for scheduling. However, discrepancies between ETA and actual arrival time often undermine planning reliability, causing inefficiencies and economic losses. Unlike ocean shipping, inland waterway transportation (IWT) faces distinct challenges such as fixed routes and dense vessel traffic. Existing VAT studies mainly focus on maritime transport and typically use single data sources (e.g., AIS or port call data), overlooking key IWT characteristics like structured routes and high traffic intensity. To address this gap, we propose a novel framework that integrates multiple data sources, including port call records, AIS trajectories, vessel features for VAT prediction. A tailored feature engineering pipeline reconstructs vessel routes and estimates sailing distances and traffic flow. Using the Port of Rotterdam as a case study, advanced ensemble tree-based models are applied. Experimental results show a 79.55% reduction in mean absolute error (from 17.06 to 3.49 h) relative to vessel-reported ETAs. Feature importance analysis identifies voyage distance, reported ETA, and waterway traffic flow as key factors. This study demonstrates that data-driven VAT prediction can support proactive berth scheduling, traffic-aware coordination, and just-in-time vessel arrivals, thereby improving operational efficiency, reducing emissions in IWT, and supporting sustainable port management. ... Read more