top of page
Image by SJ Objio
Title: Vessel Movement Tracking and Vessel Destination Prediction for Supply and Demand Forecasting
Organization

Vale - IBM

Period

January 2020 - December 2022

Status

Completed

Est. Budget

USD 3.0M

Est. Business Benefit

USD 24.0M

Indicative Value-to-cost

8.0x

Keywords

AIS analytics, destination prediction, Random Forest, shipping optimisation, supply-demand forecasting, bulk vessels

Benefit basis

Estimated value from improved chartering decisions, reduced demurrage and waiting time, better vessel-positioning awareness, and improved customer delivery reliability.

Business Challenges 
Large-scale bulk-material producers depend on a limited global pool of Capesize and other bulk vessels. Chartering decisions are highly sensitive to vessel availability, vessel positioning, port congestion, cargo demand, competitor activity, and the expected destination of vessels already in motion. The business required a reliable method to estimate vessel supply and demand by region, reduce uncertainty in chartering decisions, and improve confidence that material would reach customer ports on time.
Expanded Technical Solution
The project delivered a data-driven vessel intelligence capability that automatically collected and harmonised internal and external shipping datasets, including vessel movement feeds, port calls, berth activity, cargo schedules, vessel characteristics, historical voyages, trade routes, weather constraints, and commercial demand signals. A machine-learning prediction layer estimated vessel destination, likely availability windows, regional supply, and demand pressure. The original Random Forest concept was expanded into an ensemble architecture suitable for operational decision support, with classification models for destination prediction and regression/time-series models for vessel availability and demand forecasting.
Technical Work Packages/Methods
  • Built automated data pipelines for AIS-style movement data, port-call data, voyage history, vessel master data, customer shipment schedules, and commercial demand indicators.
  • Engineered route, speed, heading, draught, port-stay, vessel-class, historical-lane, and seasonal features to improve destination and availability prediction.
  • Implemented Random Forest and complementary gradient-boosting models for destination classification and regional supply forecasts.
  • Created probability scoring for expected destination, estimated time of arrival, vessel release date, and availability confidence.
  • Developed exception-handling logic for missing signals, route deviations, port congestion, and abnormal vessel behaviour.
Key Deliverables
  • End-to-end vessel tracking and destination prediction data pipeline.
  • Machine-learning models for vessel destination, supply availability, and demand forecasting.
  • Commercial dashboard for supply-demand balance, vessel availability, regional heat maps, and confidence scoring.
  • Operational playbook for model monitoring, model retraining, and user adoption in shipping/chartering workflows.
Results and Value Created
The project improved visibility over global vessel availability and reduced reliance on manual interpretation of fragmented shipping information. It enabled commercial teams to evaluate chartering options earlier, compare regional supply-demand pressure, and reduce exposure to last-minute vessel shortages or high-cost chartering decisions. The estimated benefit reflects lower freight/chartering cost, reduced demurrage exposure, improved cargo delivery reliability, and better commercial timing.
bottom of page