Data science & mobility
ODM Road Network
Predicting origin-destination matrices for Hungarian road networks with ML and spatial analytics.
Overview
The ODM Road Network project delivered the first graph-based origin-destination matrix prediction method on Hungarian public road data. It blends network topology, geospatial signals, and socio-economic inputs to estimate travel demand when surveys are sparse.
Outputs are designed for transport planners who need fast, interpretable forecasts without full-scale data collection.
Key features
- Engineered graph-based features from road topology and accessibility metrics.
- Integrated traffic, population, and geospatial data sources.
- Trained ML models to predict OD flows with transparent error reporting.
- Delivered reproducible notebooks for stakeholder handoff.
Technical approach
The solution combines graph analytics, regression ensembles, and spatial smoothing to infer missing flows. Model evaluation emphasizes interpretability for policy teams.
Results & impact
ODM predictions reduced model error against baselines and supported scenario analysis for infrastructure investment decisions.