PreSTGNet: Source Code and Supporting Materials for Traffic Flow Prediction

This repository contains the source code, configuration files, and supporting materials for our traffic flow prediction study based on a two-stage pre-training and fine-tuning framework. The experiments are conducted on six public benchmark datasets: PeMS03, PeMS04, PeMS07, PeMS08, BikeNYC, and TaxiNYC. The PeMS datasets record freeway traffic volume at 5-minute intervals using loop detectors, while BikeNYC and TaxiNYC represent citywide urban mobility demand flows in New York City and are aggregated into grid-based spatiotemporal sequences with 30-minute intervals. The datasets are obtained from their original public sources. This repository provides the implementation code and related materials for reproducibility and does not redistribute raw data unless otherwise stated.