An interpretable and efficient multi-scale spatio-temporal neural network for traffic flow forecasting

Wenzhu Zhao, Guan Yuan, Yanmei Zhang, Xiao Liu, Shang Liu, Lei Zhang

Abstract: Highlights•Dividing traffic sequence into patches to preserve multi-scale temporal features.•Two advanced KAN are designed to improve the interpretability of traffic forecasting.•Integrating multi-scale temporal features to capture latent patterns in traffic data.