Predictive tanker port-congestion and ETA dataset built on public AIS

PORTFLOWbuyer room · privateManage buyer room

Predictive tanker port-congestion and ETA dataset built on public AIS. Concretely: per-voyage predicted ETAs continuously refreshed for every trackable vessel (tanker, container, bulk, general, RoRo) approaching or anchoring at any of 51 digitised ports (including ARA, Hamburg, Algeciras, Fujairah, Singapore, Houston, Sabine Pass, Ras Laffan), scored against the crew-broadcast ETA with an open, published RMSE/MAE benchmark computed over exactly the same closed voyages for both series — the operator states the broadcast baseline excludes sentinel placeholder ETAs beyond 7 days so the comparison is honest. Derived layers: live congestion indices and waiting times per port; transit counts, dwell time and deviation at 12-13 chokepoints (Hormuz, Suez, Bosphorus, Malacca); a per-voyage lifecycle state machine (open, arrival-moored, departure) derived from navigation status and motion; cargo-class assignment fusing AIS ship type with a name-and-destination text heuristic tuned per port; dwell-at-anchor anomaly flags with thresholds tuned by cargo class; delay attribution; multi-regime sanctions screening across four regimes in one pass with audit trail; and behavioural AIS-gap and dark-event detection in contested waters. Architecture matters for a buyer: source AIS is a community real-time stream plus vessel-broadcast static messages, with no proprietary feed integrated in v1 (commercial satellite AIS is roadmap, not integrated), so every asset here is derived — but the platform also stores rate-limited per-vessel position snapshots explicitly to enable backtesting and model replay, plus relational storage with timestamped lineage so any metric is reproducible at a given instant.