"Qian traffic benchmark"

"The Qian-Traffic Benchmark Dataset provides a real-world testing ground for traffic flow prediction in mountainous urban road networks. Collected from 132 key road segments in Guiyang City over a continuous three-month period (at a 2-minute sampling interval), it accumulates over 50,000 observation steps.Unlike flat-city datasets, Qian-Traffic captures the unique characteristics of southwestern China's karst topography, featuring irregular road connections, local high-density coupling, and frequent localized congestion. To preserve the natural laws of traffic evolution, the data has been preprocessed using a novel time-variant adaptive imputation mechanism (incorporating LSTM, bidirectional interpolation, and moving average techniques based on time-of-day traffic patterns). Furthermore, dual periodic labels (time-of-day and day-of-week) are injected to reflect local commuting routines. This dataset is explicitly split chronologically (6:2:2) to support the rigorous evaluation and real-world deployment of lightweight, high-precision intelligent transportation systems."