Models / falak-analytics

falak-analytics/port-congestion-forecaster

Port Congestion Forecaster

PublicMinimal riskCommercialForecastingLightGBMN/A1.2.0
890 7.2K/30d 42

MAE (waiting hrs)

3.6

MAPE (occupancy)

8.9%

Congestion-event recall

0.81

R² (occupancy)

0.87

About this model

Forecasts berth occupancy and vessel waiting time at Hamad Port over a 14-day horizon using scheduled calls, AIS-derived arrival estimates, and seasonal trade patterns. Outputs congestion-risk scores per terminal to support berth planning and shipper advisories.

Intended use

Berth-window planning by terminal operators and congestion advisories for freight forwarders and government logistics dashboards.

Training data lineage

Trained on port-vessel-calls covering 38,000 vessel calls at Hamad Port from 2021 through 2025.

Limitations & bias notes

Unscheduled diversions from regional ports and severe weather closures are not modelled and cause under-prediction of waiting times. Container-terminal accuracy is stronger than for general-cargo and livestock berths where call data is sparser.

Evaluation metrics

MAE (waiting hrs)3.6
MAPE (occupancy)8.9%
Congestion-event recall0.81
R² (occupancy)0.87
#logistics#ports#forecasting#hamad-port#lightgbm#supply-chain

Try it

Live sandbox

horizon=12&profile=weekday_peak

fixture

Feedback

Owning entity

faFalak Analytics
Updated
2026-06-25
Latest version
1.2.0
License
Commercial Paid
Access
Open