Exploiting dynamic spatio-temporal correlations for origin-destination demand prediction

A deep learning framework for metro Origin-Destination (OD) matrix prediction with temporal-spatial attention mechanisms and delayed flow completion strategies.

Architecture

The framework implements a multi-granularity temporal encoding architecture combining:

• Recurrent graph convolutional units
• Dual-stream attention mechanisms
• Bidirectional temporal encoding
• Adaptive flow completion modules

Citation

If you use this work, please cite:

@article{gong2025exploiting,
  title={Exploiting dynamic spatio-temporal correlations for origin-destination demand prediction},
  author={Gong, Yongshun and Yu, Piao and Zhang, Xu and Zhang, Xinxin and Nie, Xiushan and Sun, Haoliang},
  journal={Expert Systems With Applications},
  pages={130095},
  year={2025},
  publisher={Elsevier}
}