DURANet: An Early-Fusion Spatiotemporal Graph Learning Framework for Quantifying Urban Vitality and Resilience — Supplementary Data and Code

Source
DataCite
Published
2026
Country
China
License
Creative Commons Attribution 4.0 International
Themes
Machine Learning & AI, Disaster Risk & Resilience, Sustainable Cities

Get the data from DataCite

About this dataset

This package provides the reproducibility materials for DURANet (Dynamic Urban Resilience Attention Network), an early-fusion spatiotemporal graph learning framework, and the Spatiotemporal Vitality–Resilience (SVR) Index, a formally specified composite metric unifying five urban dimensions (Mobility, Safety, Energy Efficiency, Environmental Vitality, and Social Vitality) with PCA-derived weights. The framework was evaluated on controlled-access municipal digital-twin data from the Nantong city region, Jiangsu Province, China, with companion data from Haimen District and Qidong City. On the held-out 2023 evaluation set, DURANet attains a Vitality F1-score of 0.94 and a Traffic MAE of 2.15 km/h, outperforming late-fusion baselines under paired Diebold–Mariano testing (Bonferroni-corrected, n = 8,760 test time steps).Because the raw municipal digital-twin, location-based-service, and...

How the data was collected

This package provides the reproducibility materials for DURANet (Dynamic Urban Resilience Attention Network), an early-fusion spatiotemporal graph learning framework, and the Spatiotemporal Vitality–Resilience (SVR) Index, a formally specified composite metric unifying five urban dimensions (Mobility, Safety, Energy Efficiency, Environmental Vitality, and Social Vitality) with PCA-derived weights. No synthetic or randomly generated values are used in any reported metric or empirical results figureThis deposit accompanies the manuscript submitted to Environment and Planning B: Urban Analytics and City Science.

Browse sustainability datasets on SusanHub