Data from: Spatio-Temporal Graph Neural Network for Urban Spaces: Interpolating Citywide Traffic Volume

Urban Traffic Volume Dataset – Berlin (Strava) & New York City (Taxi) Associated paper: Spatio-Temporal Graph Neural Network for Urban Spaces: Interpolating Citywide Traffic VolumeAuthors: Silke K. Kaiser, Filipe Rodrigues, Carlos Lima Azevedo, Lynn H. Kaack Citation Request If you use this dataset, please cite our paper: Kaiser, S. K., Rodrigues, F., Azevedo Lima, C., & Kaack, L.H. (2025). Spatio-Temporal Graph Neural Network for Urban Spaces: Interpolating Citywide Traffic Volume. [published on arXiv]. Dataset Overview This dataset includes street-level traffic volume data for two major urban areas: Berlin (Strava Cycling Data): Daily bicycle traffic volumes from 2019–2023, aggregated from publicly shared Strava user data. New York City (Taxi Data): Hourly motorized traffic volumes from Manhattan for January–February 2016, derived from GPS trajectories of yellow taxis. Both datasets are provided at the street-segment level and come with rich auxiliary features capturing spatial, temporal, infrastructure, and contextual information. Each city includes: <City>_data:Full feature table for each street segment, including traffic volume and auxiliary features. <City>_graph_geometry:Geometry for each street segment. <City>_adjacency_binary:Binary adjacency matrix. <City>_adjacency_similarity:Adjacency matrix weighted by node feature similarity. <City>_adjacency_distancebird:Adjacency matrix based on Euclidean (bird’s-eye) distance. <City>_adjacency_distanceroad:Adjacency matrix based on real-world road network distance. <City>_adjacency_distancetime:Adjacency matrix weighted by estimated travel time over the road network. Key Features and Methodology Volume Estimation: Strava volumes are rounded aggregates of bike trips; NYC volumes are computed from reconstructed taxi trajectories. Filtering: Extreme outliers (e.g., from special events) are filtered per segment to focus on typical traffic conditions. Auxiliary Features: Built environment (e.g., speed limits, road types, lane counts) Points of Interest (e.g., shops, schools, transit stops) Network connectivity metrics (degree, betweenness, etc.) Temporal indicators (weekday, holidays, hour, month) Weather data (sunshine, precipitation, temperature) Socioeconomic indicators (Berlin only) Proxy motorized traffic metrics (Berlin only) See the paper for a complete list of features and detailed methodology. ------------------- We are grateful the European Union’s Horizon Europe research and innovation program funded this project under Grant Agreement No 101057131, Climate Action To Advance HeaLthY Societies in Europe (CATALYSE).