Research map / Graph learning / Knowledge and temporal graphs
Spatio-temporal graph traffic forecasting
Forecast traffic with spatio-temporal graph neural networks.
60 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Knowledge and temporal graphs, keyword cluster “dynamic · temporal graph · spatio temporal”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Few recent reports The tracked benchmarks have at most one report from 2025–26 that could beat the old record, too few to say whether they are still moving.
11 papers in 2023–24 and 19 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 4.3% to 6.7%.
No longer reported by papers here: PEMS08 (MAE, last in 2024); PEMS03 (MAE, last in 2024).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · Deep Spatio-Temporal Residual Networks for Citywide Crowd Flows Prediction (AAAI 2017 · 2,253 citations)
- 2018 · When Will You Arrive? Estimating Travel Time Based on Deep Neural Networks (AAAI 2018 · 366 citations)
- 2019 · Attention Based Spatial-Temporal Graph Convolutional Networks for Traffic Flow Forecasting (AAAI 2019 · 2,989 citations)
- 2020 · GMAN: A Graph Multi-Attention Network for Traffic Prediction (AAAI 2020 · 1,722 citations)
- 2021 · Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting (AAAI 2021 · 904 citations)
- 2022 · Graph Neural Controlled Differential Equations for Traffic Forecasting (AAAI 2022 · 351 citations)
- 2023 · PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction (AAAI 2023 · 621 citations)
- 2024 · Spatio-Temporal Pivotal Graph Neural Networks for Traffic Flow Forecasting (AAAI 2024 · 129 citations)
- 2025 · Spatiotemporal-aware Trend-Seasonality Decomposition Network for Traffic Flow Forecasting (AAAI 2025 · 49 citations)
- 2026 · HyperD: Hybrid Periodicity Decoupling Framework for Traffic Forecasting (AAAI 2026 · 3 citations)
Earliest papers here
- Deep Spatio-Temporal Residual Networks for Citywide Crowd Flows Prediction (AAAI 2017 · 2,253 citations)
- When Will You Arrive? Estimating Travel Time Based on Deep Neural Networks (AAAI 2018 · 366 citations)
- Attention Based Spatial-Temporal Graph Convolutional Networks for Traffic Flow Forecasting (AAAI 2019 · 2,989 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
PEMS04 · MAE (lower is better)
6 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 21.19 | Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting (AAAI 2020 · 1,590 citations) |
| 2021 | 19.83 | Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting (AAAI 2021 · 904 citations) |
| 2022 | 19.3 | DSTAGNN: Dynamic Spatial-Temporal Aware Graph Neural Network for Traffic Flow Forecasting (ICML 2022) |
| 2023 | 18.32 | PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction (AAAI 2023 · 621 citations) |
PEMS07 · MAE (lower is better)
6 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 24.26 | Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting (AAAI 2020 · 1,590 citations) |
| 2021 | 22.07 | Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting (AAAI 2021 · 904 citations) |
| 2022 | 21.42 | DSTAGNN: Dynamic Spatial-Temporal Aware Graph Neural Network for Traffic Flow Forecasting (ICML 2022) |
| 2023 | 19.83 | PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction (AAAI 2023 · 621 citations) |
PEMS07 · RMSE (lower is better)
6 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 39.03 | Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting (AAAI 2020 · 1,590 citations) |
| 2021 | 35.8 | Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting (AAAI 2021 · 904 citations) |
| 2022 | 34.51 | DSTAGNN: Dynamic Spatial-Temporal Aware Graph Neural Network for Traffic Flow Forecasting (ICML 2022) |
| 2023 | 32.87 | PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction (AAAI 2023 · 621 citations) |
PEMS04 · RMSE (lower is better)
5 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 31.88 | Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting (AAAI 2021 · 904 citations) |
| 2022 | 31.46 | DSTAGNN: Dynamic Spatial-Temporal Aware Graph Neural Network for Traffic Flow Forecasting (ICML 2022) |
| 2023 | 29.96 | PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction (AAAI 2023 · 621 citations) |
| 2024 | 29.64 | Spatio-Temporal Pivotal Graph Neural Networks for Traffic Flow Forecasting (AAAI 2024 · 129 citations) |
PEMS08 · MAE (lower is better)
5 papers, 2020–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 17.13 | Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting (AAAI 2020 · 1,590 citations) |
| 2021 | 16.64 | Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting (AAAI 2021 · 904 citations) |
| 2022 | 15.67 | DSTAGNN: Dynamic Spatial-Temporal Aware Graph Neural Network for Traffic Flow Forecasting (ICML 2022) |
| 2023 | 13.58 | PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction (AAAI 2023 · 621 citations) |
PEMS03 · MAE (lower is better)
4 papers, 2020–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 17.48 | Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting (AAAI 2020 · 1,590 citations) |
| 2021 | 16.77 | Spatial-Temporal Fusion Graph Neural Networks for Traffic Flow Forecasting (AAAI 2021 · 904 citations) |
| 2022 | 15.57 | DSTAGNN: Dynamic Spatial-Temporal Aware Graph Neural Network for Traffic Flow Forecasting (ICML 2022) |
| 2024 | 14.37 | Spatio-Temporal Pivotal Graph Neural Networks for Traffic Flow Forecasting (AAAI 2024 · 129 citations) |
What has been tried
47 papers here report 431 measured changes (a component added, removed or swapped, compared with a baseline): helped 76%, no clear effect 5%, hurt 16%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Replace identified pivotal nodes with randomly selected pivotal nodes | PEMS08 and England · prediction accuracy | Hurt | Spatio-Temporal Pivotal Graph Neural Networks for Traffic Flow Forecasting (AAAI 2024 · 129 citations) | |
| add temporal embedding modeling | PeMS04 · MAE | 19.10 -> 18.40 | Helped | Spatiotemporal-aware Trend-Seasonality Decomposition Network for Traffic Flow Forecasting (AAAI 2025 · 49 citations) |
| Add embedding-level knowledge distillation | four traffic datasets · prediction results | Helped | Efficient Traffic Prediction Through Spatio-Temporal Distillation (AAAI 2025 · 23 citations) | |
| Add temporal decomposition into seasonal and trend patterns | Traffic, 15-min prediction · MAE | Helped | Towards Dynamic Spatial-Temporal Graph Learning: A Decoupled Perspective (AAAI 2024 · 17 citations) | |
| replace average pooling with max pooling | ISCX-VPN2016, ISCX-Tor2016, QUIC2019, CICIoT2022 · accuracy | 98.14% -> 97.50%; 98.57% -> 97.31%; 96.41% -> 96.95%; 96.91% -> 96.33% | Mixed | Dual-Channel Interactive Graph Transformer for Traffic Classification with Message-Aware Flow Representation (AAAI 2025 · 8 citations) |
| add Spatial Facet Modeling | V-GCT prediction, steps 1–12 average · MAE | 5.46 -> 5.34 | Helped | TelTrans: Applying Multi-Type Telecom Data to Transportation Evaluation and Prediction via Multifaceted Graph Modeling (AAAI 2024 · 4 citations) |
| add Redundancy Context Disentanglement | Metro flow forecasting · forecasting error | Helped | Responsive Dynamic Graph Disentanglement for Metro Flow Forecasting (AAAI 2025 · 3 citations) | |
| add in-situ gating over diverse experts | METR-LA average prediction windows · MAE | 3.00 -> 2.93 | Helped | TESTAM: A Time-Enhanced Spatio-Temporal Attention Model with Mixture of Experts (ICLR 2024 · 3 citations) |
| Reduce the prompt-pool rank k from 6 to 2 | Larger datasets · Tuning parameters | ∼ 63% | Helped | Expand and Compress: Exploring Tuning Principles for Continual Spatio-Temporal Graph Forecasting (ICLR 2025 · 1 citation) |
| Use the full Recurrent GNN instead of the Lite variant | Collision prediction test set · macro average recall | 0.73±0.0043 -> 0.76±0.0040 | Helped | I Open at the Close: A Deep Reinforcement Learning Evaluation of Open Streets Initiatives (AAAI 2024 · 1 citation) |
| Remove node-to-node graph messaging | LargeST-SD, horizon 3 · MAE | 25.72 -> 25.45 | Helped | Robust Spatio-Temporal Centralized Interaction for OOD Learning (ICML 2025) |
What reviewers will expect
Datasets and baselines named in the 9 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- STGCN · 4 of 9 papers
- DCRNN · 3 of 9 papers
- ASTGCN · 3 of 9 papers
- AGCRN · 3 of 9 papers
- HA · 2 of 9 papers
- VAR · 2 of 9 papers
- STGODE · 2 of 9 papers
- GMAN · 2 of 9 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 30 papers; from OpenAlex and the papers' affiliations).
- University of Science and Technology of China · 3 papers
- Renmin University of China · 2 papers
- Beihang University · 2 papers
Use it
Download BibTeX of 39 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Continuous-time dynamic graph learning (132 papers)
- Temporal graph neural architectures (69 papers)
Elsewhere in the topic
- Knowledge graph embeddings (110 papers)
- Graph reasoning for visual recognition (96 papers)
- Graph-based dialog systems (94 papers)
- Social and behavior graphs (87 papers)
- Temporal knowledge graph reasoning (84 papers)
- Scene graph generation (70 papers)
- Logical query answering and rule reasoning (68 papers)
- Heterogeneous graph learning (66 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).