Astana Synthetic Urban Traffic Congestion Benchmark (February–May 2026)

A reproducible, city-scale synthetic benchmark for short-term urban traffic congestion forecasting, calibrated to the road network of Astana, Kazakhstan. Covers 100 virtual sensor locations (real OSM-derived positions; simulated speed/congestion values) at 15-minute resolution from 2026-02-01 to 2026-05-31 (11,520 time steps per sensor). Includes a normalized congestion index, road-class metadata, and the train/test split used in the accompanying paper. Released as a license-free public benchmark for evaluating statistical, machine-learning and deep-learning traffic forecasting methods. The dataset is synthetic: sensor locations are real, but speed and congestion values are generated by a calibrated stochastic simulation, not measured from physical sensors.