ETD

• estimated time of arrival (ETA)
• machine learning
• maritime container shipping
• vessel delay prediction

This thesis presents a machine learning-based approach to predicting maritime container transit times within the ConTrack 4.0 - Smart Intermodal Container Tracking project. The research addresses the critical challenge of accurate arrival time estimation in global supply chains, where uncertainty in transit times significantly impacts logistics planning and operational efficiency.
A comprehensive pipeline was developed to process and integrate multiple data sources, including carrier provided data, vessel specifications, and meteorological information. The system handles thousands of container shipments across 18 major maritime routes, implementing data quality frameworks. The predictive models employ advanced feature engineering techniques with rigorous temporal validation. The final solution achieves an average improvement of ETA errors over carrier-provided ETAs. This work demonstrates how machine learning can enhance visibility and reliability in maritime logistics operations.