mlsanityData to 14 Oct 2026

MP3: Multi-Period Pattern Pre-training for Spatio-Temporal Forecasting

arXiv 2606.13119 ↗First posted 11 Jun 2026cs.LG cs.AI cs.NEcited by 0

Not on the map. Why?

Abstract (arXiv)

Spatio-Temporal forecasting is crucial in diverse fields, such as transportation, climate, and energy. Urban spatio-temporal data exhibits temporal mirage: similar short-window inputs have divergent future trends, and vice versa. Existing spatio-temporal graph neural networks (STGNNs) cannot effectively identify such mirages. We argue that the core reason lies in the short-window inputs that have incomplete period observation, heterogeneous global spatial correlation, and cross-period superposition causality. To bridge this gap, we develop a novel Multi- Period Pattern Pre-training (MP3), a plug-and-play pre-training plugin for distinguishing temporal mirages. MP3 presents two core innovations: (1) The multi-period pattern learning is designed to learn multi-period patterns from long time series. Specifically, multi-period temporal modeling leverages edge convolution to identify different multi-period patterns. Multi-period spatial modeling uses a bottleneck project and a global memory bank to capture heterogeneous global spatial relations efficiently. Cross-period pattern interaction employs a causality-enhanced Transformer to capture dependencies across different period patterns. (2) This plugin can seamlessly integrate into existing STGNN backbones to strengthen their forecasting performance. The experiment on five STGNN baselines across five real-world datasets (including a large-scale dataset CA) verify the effectiveness, superior scalability and strong adaptability of MP3, which brings consistent and robust performance improvements across all evaluated baselines. On average, MP3 reduces the MAE 4.7% and the RMSE 5.0%. The code can be available at https://github.com/YAN-outlook/MP3.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 31 of its 56 references that are in the map. First posted on arXiv 11 Jun 2026.

Builds on

31 of 56 references.

References · 31

  1. 2016Semi-Supervised Classification with Graph Convolutional Networkscited by 11,361
  2. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
  3. 2014Going Deeper with Convolutionscited by 4,241
  4. 2017Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecastingcited by 1,136
  5. 2022TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysiscited by 1,058
  6. 2019Graph WaveNet for Deep Spatial-Temporal Graph Modelingcited by 829
  7. 2020Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networkscited by 566
  8. 2020Adaptive Graph Convolutional Recurrent Network for Traffic Forecastingcited by 530
  9. 2019GMAN: A Graph Multi-Attention Network for Traffic Predictioncited by 434
  10. 2018T-GCN: A Temporal Graph ConvolutionalNetwork for Traffic Predictioncited by 401
  11. 2016Deep Spatio-Temporal Residual Networks for Citywide Crowd Flows Predictioncited by 333
  12. 2021Spatial-Temporal Graph ODE Networks for Traffic Flow Forecastingcited by 210
19 more
  1. 2023Spatio-Temporal Graph Neural Networks for Predictive Learning in Urban Computing: A Surveycited by 140
  2. 2023PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Predictioncited by 137
  3. 2022Pre-training Enhanced Spatial-temporal Graph Neural Network for Multivariate Time Series Forecastingcited by 117
  4. 2022Spatio-Temporal Meta-Graph Learning for Traffic Forecastingcited by 88
  5. 2022AirFormer: Predicting Nationwide Air Quality in China with Transformerscited by 73
  6. 2022Towards Spatio-Temporal Aware Traffic Time Series Forecasting--Full Versioncited by 59
  7. 2020Coupled Layer-wise Graph Convolution for Transportation Demand Predictioncited by 54
  8. 2023MSGNet: Learning Multi-Scale Inter-Series Correlations for Multivariate Time Series Forecastingcited by 50
  9. 2022Spatial-Temporal Hypergraph Self-Supervised Learning for Crime Predictioncited by 39
  10. 2022Learning the Evolutionary and Multi-scale Graph Structure for Multivariate Time Series Forecastingcited by 36
  11. 2023Spatial-Temporal-Decoupled Masked Pre-training for Spatiotemporal Forecastingcited by 27
  12. 2024AirPhyNet: Harnessing Physics-Guided Neural Networks for Air Quality Predictioncited by 26
  13. 2025Spatiotemporal-aware Trend-Seasonality Decomposition Network for Traffic Flow Forecastingcited by 19
  14. 2024Heterogeneity-Informed Meta-Parameter Learning for Spatiotemporal Time Series Forecastingcited by 18
  15. 2024FlashST: A Simple and Universal Prompt-Tuning Framework for Traffic Predictioncited by 18
  16. 2023GPT-ST: Generative Pre-Training of Spatio-Temporal Graph Neural Networkscited by 14
  17. 2020Auto-STGCN: Autonomous Spatial-Temporal Graph Convolutional Network Search Based on Reinforcement Learning and Existing Research Resultscited by 9
  18. 2024Multi-Modality Spatio-Temporal Forecasting via Self-Supervised Learningcited by 9
  19. 2023Learning Time-aware Graph Structures for Spatially Correlated Time Series Forecastingcited by 7

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Causal Spatio-Temporal Prediction: An Effective and Efficient Multi-Modal Approach
  2. 2024Causal Deciphering and Inpainting in Spatio-Temporal Dynamics via Diffusion Model
  3. 2026LLMODE: Aligning ODEs with LLMs via Gated Token Injection for Irregular Spatio-Temporal Forecasting
  4. 2024Diffusion Transformers as Open-World Spatiotemporal Foundation Models
  5. 2025Collaborative Deterministic-Probabilistic Forecasting for Diverse Spatiotemporal Systems
  6. 2025Adaptive Spatio-Temporal Graphs with Self-Supervised Pretraining for Multi-Horizon Weather Forecasting
  7. 2024ProGen: Revisiting Probabilistic Spatial-Temporal Time Series Forecasting from a Continuous Generative Perspective Using Stochastic Differential Equations
  8. 2026UniMamba: A Unified Spatial-Temporal Modeling Framework with State-Space and Attention Integration