Research map / Graph learning / Knowledge and temporal graphs
Deciphering Spatio-Temporal Graph Forecasting: A Causal Lens and Treatment
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use CaST rather than AGCRN | STG forecasting · PEMS08 (24→24) | MAE: 17.06 ± 0.14 -> 16.44 ± 0.10 | Helped |
| use CaST rather than AGCRN | STG forecasting · PEMS08 (24→24) | RMSE: 26.80 ± 0.15 -> 26.61 ± 0.15 | Helped |
| use CaST rather than AGCRN | STG forecasting · AIR-BJ (24→24) | MAE: 23.43 ± 0.29 -> 22.90 ± 0.09 | Helped |
| use CaST rather than AGCRN | STG forecasting · AIR-BJ (24→24) | RMSE: 35.66 ± 0.57 -> 34.84 ± 0.11 | Helped |
| use CaST rather than MTGNN | STG forecasting · AIR-GZ (24→24) | MAE: 12.43 ± 0.11 -> 12.36 ± 0.01 | Helped |
| use CaST rather than AGCRN | STG forecasting · AIR-GZ (24→24) | RMSE: 17.49 ± 0.01 -> 17.25 ± 0.05 | Helped |
| replace edge convolution with a self-adaptive adjacency matrix (ablation) | CaST · AIR-BJ (24→24) | MAE: 22.90 -> 24.28 | Hurt |
| replace edge convolution with graph attention for causal scores (ablation) | CaST · AIR-BJ (24→24) | MAE: 22.90 -> 23.77 | Hurt |
| remove environment features from prediction (ablation) | CaST · AIR-BJ and PEMS08 | MAE | Hurt |
| remove entity features from prediction (ablation) | CaST · AIR-BJ and PEMS08 | MAE | Hurt |
| remove causal scores that guide spatial message passing (ablation) | CaST · AIR-BJ and PEMS08 | MAE | Hurt |
| remove position embeddings (ablation) | CaST · AIR-BJ | evaluation loss | Hurt |
Closest papers by meaning
- Graph-based Forecasting with Missing Data through Spatiotemporal Downsampling (ICML 2024 · similarity 0.90)
- ST-FiT: Inductive Spatial-Temporal Forecasting with Limited Training Data (AAAI 2025 · similarity 0.90)
- Causal Spatio-Temporal Prediction: An Effective and Efficient Multi-Modal Approach (NeurIPS 2025 · similarity 0.90)
- Towards Dynamic Spatial-Temporal Graph Learning: A Decoupled Perspective (AAAI 2024 · similarity 0.89)
- Causal Deciphering and Inpainting in Spatio-Temporal Dynamics via Diffusion Model (NeurIPS 2024 · similarity 0.89)
- Expand and Compress: Exploring Tuning Principles for Continual Spatio-Temporal Graph Forecasting (ICLR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.