pith. sign in

arxiv: 2605.18793 · v1 · pith:7CS6SV7Znew · submitted 2026-05-11 · 💻 cs.LG · cs.AI

Dimensional Balance Improves Large Scale Spatiotemporal Prediction Performance

Jing Chen , Shixiang Pan , Yujie Fan , Haocheng Ye , Haitao Xu , Wenqiang Xu This is my paper

Pith reviewed 2026-05-20 22:02 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords spatiotemporal predictionlow-rank embeddingtemporal horizondimensional balanceentropy diagnosticsurban trafficmeteorologyepidemic forecasting
0
0 comments X

The pith

Balancing spatial and temporal dimensions through low-rank compression and extended horizons improves large-scale spatiotemporal prediction accuracy.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines why existing spatiotemporal prediction methods often produce only small gains and limited transfer across fields such as traffic and weather. It treats spatial and temporal entropy as diagnostic tools to detect complexity mismatches that correlate with higher forecast uncertainty under fixed model size. The proposed framework compresses spatial features with low-rank embeddings to retain core structure while lengthening the temporal window to reduce error accumulation from heterogeneity. Experiments on traffic, meteorological, and epidemic datasets show measurable accuracy increases and domain transfer. If this holds, it offers a direct route to stronger performance by aligning representation dimensions rather than scaling model capacity alone.

Core claim

The authors establish that a framework harmonizing spatial and temporal feature representations by applying low-rank matrix embedding to compress spatial dimensionality while extending the temporal horizon produces substantial accuracy gains and demonstrates applicability across urban traffic, meteorological, and epidemic datasets.

What carries the argument

Low-rank matrix embedding for spatial compression paired with an extended temporal horizon, guided by entropy measures that diagnose spatiotemporal complexity mismatch.

If this is right

  • Prediction error decreases on urban traffic datasets under the same model capacity.
  • Meteorological forecasts exhibit higher accuracy with the harmonized representations.
  • Epidemic modeling tasks gain reliability from the reduced cumulative temporal errors.
  • The approach transfers to multiple domains without requiring domain-specific redesigns.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same compression-plus-extension pattern could be tested on video frame prediction or multi-sensor time series where spatial and temporal scales also compete.
  • Entropy mismatch scores might serve as an automatic signal for choosing rank and horizon values in new architectures.

Load-bearing premise

Entropy measures of spatial and temporal complexity can reliably flag mismatches that, once corrected through dimension adjustments, produce better forecasts.

What would settle it

Applying the framework to a new spatiotemporal dataset that exhibits high entropy mismatch yet shows no accuracy gain or a loss relative to baselines would falsify the central claim.

Figures

Figures reproduced from arXiv: 2605.18793 by Haitao Xu, Haocheng Ye, Jing Chen, Shixiang Pan, Wenqiang Xu, Yujie Fan.

Figure 1
Figure 1. Figure 1: Scatter plots of spatial entropy vs. temporal entropy for four distinct time horizons (12-horizon, 1-day, 7-day, and 14-day). Marker size is proportional to the number of nodes in each network, and a diagonal reference line indicates where spatial and temporal entropies are equal. Points further away from the diagonal indicate greater mismatch between spatial and temporal entropy, which is empirically asso… view at source ↗
Figure 2
Figure 2. Figure 2: Framework overview integrating spatial dimensionality reduction, temporal window expansion, and fusion. The left side indicates the initial dimensional imbalance between the short-horizon temporal input and the prior graphs, while the right side indicates the balanced representations used for fusion. A significant challenge emerges from the disproportionate scale between the spatial dimension, represented … view at source ↗
Figure 3
Figure 3. Figure 3: Temporal enhancement module. The short-term window shown here is one instantiation of the Temporal Enhancement strategy. 4.2.1. Temporal Window Extension By extending the temporal window, we enhance the model’s ability to capture long-term dependencies, thereby increasing temporal capacity. For a given node i, let the original time-series input be Xi ∈ R T ×F , and we extend it to T ′ > T to capture longer… view at source ↗
Figure 4
Figure 4. Figure 4: Hierarchical spatiotemporal fusion model. 4.3.1. Hierarchical Fusion Strategy The framework stacks L layers of STFM, each comprising two stages: single-ST fusion (SF) and multi-ST fusion (MF). We denote the reduced spatial embedding by H ∈ R N×M. For each prior graph j ∈ {1, . . . , J}, we initialize graph-specific spatial features by G j 0 = ϕ j (H), where ϕ j (·) is a lightweight (e.g., linear) projectio… view at source ↗
Figure 5
Figure 5. Figure 5: Statistical comparison of multi-indicator performance between ST￾Balance and baseline models across diverse traffic datasets. Radar plots illustrate statistical results for Pearson correlation coefficient (PCC), coefficient of determination (R2 ), Kling-Gupta efficiency (KGE), modified Nash–Sutcliffe efficiency (mNSE), and per￾centage of Nash–Sutcliffe efficiency (PNSE). ST-Balance demonstrates superior co… view at source ↗
Figure 6
Figure 6. Figure 6: Our findings indicate that moderate dimensionality reduction effec [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 6
Figure 6. Figure 6: Prediction error under varying spatial dimensionality reduction meth￾ods. Comparison of prediction performance on the LargeST SD dataset when reducing spatial embedding dimension from 716 to 16 using PCA[27], UMAP[28], Node2Vec[29], HOPE[30], and our low-rank embedding method. All methods initially benefit when the dimension is reduced from 716 to 512. However, classical linear methods degrade markedly bel… view at source ↗
Figure 7
Figure 7. Figure 7: t-SNE views of SD traffic-node embeddings. Colors denote flow patterns. (a) Original embeddings are scattered. (b) Low-rank dimensionality reduction yields clear, flow-consistent clusters. (c) Zoomed hub: similar nodes (B,C) co-locate; dissimilar nodes (A,D) separate [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Scale-dependent contributions of graph structures. Performance (MAE) changes when selectively removing prior or adaptive graphs from GWNet, highlighting the critical role of prior structures in larger networks. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Scale-dependent benefits of dimensionality reduction. Performance im￾provements (MAE) from incorporating low-rank dimensionality reduction into STAEformer, highlighting pronounced gains in larger networks. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Impact of temporal module ablation on ST-Balance performance. Prediction accuracy decreases notably without the long-window module, highlighting the module’s critical contribution to capturing periodic traffic patterns [PITH_FULL_IMAGE:figures/full_fig_p024_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Sensitivity of ST-Balance performance to temporal window length. MAE varies with increasing time window length, identifying optimal scales for different datasets. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Model comparisons for meteorological forecasting. (a) Aggregate performance (MSE) of ST-Balance versus benchmark models on Wind and Temp datasets. (b) Temporal evolution of prediction errors for wind speed forecasts, highlighting ST￾Balance’s reduced errors relative to Corrformer over extended periods. (c) Detailed temperature predictions at two spatial locations, emphasizing ST-Balance’s superior ability… view at source ↗
Figure 13
Figure 13. Figure 13: Comparative performance of different models in epidemic forecasting tasks. (a) Average MAE of different models of Infection at the county-level. (b) Average infection MAE and PCC of different models on state and county level. (c) The forecasting local results of infections from Oct 16, 2021 to Nov 30, 2021. 28 [PITH_FULL_IMAGE:figures/full_fig_p028_13.png] view at source ↗
read the original abstract

Accurate spatiotemporal pattern analysis is critical in fields such as urban traffic, meteorology, and public health monitoring. However, existing methods face performance bottlenecks, typically yielding only incremental gains and often exhibiting limited cross-domain transferability. We analyze this bottleneck through spatial and temporal entropy measures, which are used as diagnostic indicators of spatiotemporal complexity mismatch rather than as guarantees that entropy alignment alone yields better forecasting. Empirically, larger mismatch is often accompanied by higher prediction uncertainty, especially under a fixed model-capacity budget. Guided by this diagnostic, we propose a scalable, adaptive framework that harmonizes spatial and temporal feature representations. Spatial dimensionality is compressed via low-rank matrix embedding to preserve essential structure, while an extended temporal horizon captures long-range dependencies and mitigates cumulative errors arising from temporal heterogeneity. Extensive experiments on urban traffic, meteorological, and epidemic datasets demonstrate substantial accuracy gains and broad applicability across the evaluated domains, suggesting that the framework is promising for a wide range of spatiotemporal tasks beyond the current study. The code is available on GitHub at https://github.com/ST-Balance/ST-Balance.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The paper claims that spatial and temporal entropy measures serve as diagnostics for spatiotemporal complexity mismatch, which guides a framework using low-rank spatial matrix embedding for dimensionality compression and extended temporal horizons to capture long-range dependencies; this harmonization is said to yield substantial accuracy gains with broad applicability on urban traffic, meteorological, and epidemic datasets.

Significance. If the claimed gains prove robust and the entropy diagnostic is shown to have a causal rather than post-hoc role, the work could offer a practical heuristic for balancing spatial-temporal representations under fixed model capacity, with potential transfer to other large-scale prediction domains.

major comments (3)
  1. [Abstract] Abstract: the claim of 'substantial accuracy gains' on three domains is unsupported by any quantitative metrics, error bars, baseline comparisons, or statistical significance tests, making it impossible to evaluate the magnitude or reliability of the reported improvements.
  2. [Framework and Experiments] Framework and Experiments: no before/after entropy values, mismatch reduction measurements, or ablation isolating the diagnostic-to-design step are provided; without these, the central claim that low-rank compression plus horizon extension improves performance specifically because of diagnosed entropy mismatch cannot be distinguished from generic regularization or capacity effects.
  3. [Framework] Parameter selection: the free parameters (spatial embedding rank and temporal horizon length) are described as guided by entropy diagnostics, yet the manuscript does not demonstrate an explicit separation between diagnostic use and post-hoc fitting to observed performance, weakening the non-circularity of the design process.
minor comments (1)
  1. [Abstract] The GitHub link is given but the manuscript provides no details on code structure, exact experimental protocols, or data preprocessing steps needed for reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thoughtful comments, which have helped us identify areas where the manuscript can be strengthened. We address each major comment below, indicating the revisions we intend to make in the updated version.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'substantial accuracy gains' on three domains is unsupported by any quantitative metrics, error bars, baseline comparisons, or statistical significance tests, making it impossible to evaluate the magnitude or reliability of the reported improvements.

    Authors: We agree that the abstract would be more informative with quantitative details. In the revised manuscript, we will incorporate specific metrics from our experiments, including accuracy improvements with error bars, comparisons to baselines, and references to statistical tests where performed. This will allow readers to better assess the gains. revision: yes

  2. Referee: [Framework and Experiments] Framework and Experiments: no before/after entropy values, mismatch reduction measurements, or ablation isolating the diagnostic-to-design step are provided; without these, the central claim that low-rank compression plus horizon extension improves performance specifically because of diagnosed entropy mismatch cannot be distinguished from generic regularization or capacity effects.

    Authors: We acknowledge the need for more direct evidence linking the entropy diagnostic to the design choices. We will include before-and-after entropy values and mismatch reduction measurements in the experiments section. Furthermore, we will add ablation studies that compare the full framework against versions without the entropy-guided components to isolate the effect and distinguish it from generic regularization or capacity increases. revision: yes

  3. Referee: [Framework] Parameter selection: the free parameters (spatial embedding rank and temporal horizon length) are described as guided by entropy diagnostics, yet the manuscript does not demonstrate an explicit separation between diagnostic use and post-hoc fitting to observed performance, weakening the non-circularity of the design process.

    Authors: To clarify the non-circular nature of our approach, we will expand the description of the parameter selection process. We will explicitly show that entropy diagnostics are computed solely from the input data characteristics, independent of any model training or performance evaluation. The selection of spatial embedding rank and temporal horizon length will be presented as being determined based on these pre-computed diagnostics, with examples illustrating the decision process before reporting final results. revision: yes

Circularity Check

0 steps flagged

No significant circularity: entropy diagnostics are observational and framework gains rest on independent empirical validation

full rationale

The paper explicitly frames spatial and temporal entropy measures as diagnostic indicators of complexity mismatch rather than guarantees of improved forecasting. The proposed low-rank spatial compression and extended temporal horizon are presented as design choices guided by this observation, followed by direct experimental evaluation on traffic, meteorological, and epidemic datasets. No equations, fitted parameters renamed as predictions, or self-citation chains reduce the reported accuracy gains to the input diagnostics by construction. The derivation remains self-contained because the performance claims are supported by cross-domain empirical results rather than tautological re-expression of the entropy observations.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on the assumption that low-rank embedding preserves essential spatial structure and that entropy mismatch is a useful proxy for model-capacity allocation; no new entities are postulated.

free parameters (2)
  • spatial embedding rank
    Chosen to compress dimensionality while preserving structure; value not specified in abstract.
  • temporal horizon length
    Extended to capture long-range dependencies; specific length chosen empirically.
axioms (2)
  • domain assumption Low-rank matrix embedding preserves essential spatial structure for downstream prediction
    Invoked when describing spatial dimensionality compression.
  • domain assumption Entropy measures serve as valid indicators of spatiotemporal complexity mismatch
    Used as diagnostic indicators rather than guarantees.

pith-pipeline@v0.9.0 · 5725 in / 1300 out tokens · 44010 ms · 2026-05-20T22:02:59.192605+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages

  1. [1]

    M. Jin, H. Y. Koh, Q. Wen, D. Zambon, C. Alippi, G. I. Webb, I. King, S. Pan, A survey on graph neural networks for time series: Forecasting, classification, imputation, and anomaly detection, IEEE Transactions on Pattern Analysis and Machine Intelligence 46 (2024) 10466–10485

    work page 2024

  2. [2]

    J. Chen, S. Pan, W. Peng, W. Xu, Bilinear spatiotemporal fusion network: An efficient approach for traffic flow prediction, Neural Networks 187 (2025) 107382

    work page 2025

  3. [3]

    L. Chen, L. Chen, H. Wang, Spatiotemporal multi-view trend-aware network for traffic flow prediction, Knowledge-Based Systems 333 (2026) 115002

    work page 2026

  4. [4]

    K. Bi, L. Xie, H. Zhang, X. Chen, X. Gu, Q. Tian, Accurate medium- range global weather forecasting with 3D neural networks, Nature 619 (7970) (2023) 533–538

    work page 2023

  5. [5]

    H. Wu, H. Zhou, M. Long, J. Wang, Interpretable weather forecast- ing for worldwide stations with a unified deep model, Nature Machine Intelligence 5 (6) (2023) 602–611

    work page 2023

  6. [6]

    M. U. Kraemer, J. L.-H. Tsui, S. Y. Chang, S. Lytras, M. P. Khurana, S. Vanderslott, S. Bajaj, N. Scheidwasser, J. L. Curran-Sebastian, E. Se- menova, et al., Artificial intelligence for modelling infectious disease epidemics, Nature. 638 (8051) (2025) 623–635

    work page 2025

  7. [7]

    M. Liu, Y. Liu, J. Liu, Epidemiology-aware deep learning for infectious disease dynamics prediction, in: Proceedings of the 32nd ACM Interna- tional Conference on Information and Knowledge Management, ACM, 2023, p. 4084–4088.doi:10.1145/3583780.3615139

    work page doi:10.1145/3583780.3615139 2023

  8. [8]

    Y. Li, R. Yu, C. Shahabi, Y. Liu, Diffusion convolutional recurrent neural network: Data-driven traffic forecasting, in: International Conference on Learning Representations, 2018

    work page 2018

  9. [9]

    Z. Wu, S. Pan, G. Long, J. Jiang, C. Zhang, Graph wavenet for deep spatial-temporal graph modeling, in: Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, IJCAI, 2019, pp. 1907–1913.doi:10.24963/ijcai.2019/264. 30

    work page doi:10.24963/ijcai.2019/264 2019

  10. [10]

    H. Liu, Z. Dong, R. Jiang, J. Deng, J. Deng, Q. Chen, X. Song, Spatio- temporal adaptive embedding makes vanilla transformer sota for traffic forecasting, in: Proceedings of the 32nd ACM International Conference on Information and Knowledge Management, ACM, 2023, p. 4125–4129. doi:10.1145/3583780.3615160

    work page doi:10.1145/3583780.3615160 2023

  11. [11]

    J. Chen, H. Ye, Z. Ying, Y. Sun, W. Xu, Dynamic trend fusion module for traffic flow prediction, Applied Soft Computing 174 (2025) 112979

    work page 2025

  12. [12]

    C. Chen, K. Petty, A. Skabardonis, P. Varaiya, Z. Jia, Freeway perfor- mance measurement system: mining loop detector data, Transp.Res.Rec. 1748 (1) (2001) 96–102

    work page 2001

  13. [13]

    X. Liu, Y. Xia, Y. Liang, J. Hu, Y. Wang, L. Bai, C. Huang, Z. Liu, B. Hooi, R. Zimmermann, LargeST: a benchmark dataset for large-scale traffic forecasting, in: 37th Conference on Neural Information Processing Systems, NeurIPS, 2023, pp. 75354–75371

    work page 2023

  14. [14]

    Z. Shao, Z. Zhang, F. Wang, Y. Xu, Pre-training enhanced spatial- temporal graph neural network for multivariate time series forecasting, in: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD ’22, ACM, 2022, p. 1567–1577.doi: 10.1145/3534678.3539396

    work page doi:10.1145/3534678.3539396 2022

  15. [15]

    H. Gao, R. Jiang, Z. Dong, J. Deng, Y. Ma, X. Song, Spatial- temporal-decoupled masked pre-training for spatiotemporal forecast- ing, in: K. Larson (Ed.), Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, IJCAI, 2024, pp. 3998–4006. doi:10.24963/ijcai.2024/442

    work page doi:10.24963/ijcai.2024/442 2024

  16. [16]

    H. Han, M. Zhang, M. Hou, F. Zhang, Z. Wang, E. Chen, H. Wang, J. Ma, Q. Liu, STGCN: A spatial-temporal aware graph learning method for poi recommendation, in: 2020 IEEE International Conference on Data Mining, IEEE, 2020, pp. 1052–1057.doi:10.1109/ICDM50108. 2020.00124

    work page doi:10.1109/icdm50108 2020

  17. [17]

    L. Bai, L. Yao, C. Li, X. Wang, C. Wang, Adaptive graph convolutional recurrent network for traffic forecasting, in: 34th Conference on Neural Information Processing Systems, NeurIPS, 2020, pp. 17804–17815. 31

    work page 2020

  18. [18]

    Y. Fang, Y. Qin, H. Luo, F. Zhao, B. Xu, L. Zeng, C. Wang, When spatio- temporal meet wavelets: Disentangled traffic forecasting via efficient spectral graph attention networks, in: 2023 IEEE 39th International Conference on Data Engineering, IEEE, 2023, pp. 517–529

    work page 2023

  19. [19]

    J. Han, W. Zhang, H. Liu, T. Tao, N. Tan, H. Xiong, BigST: Linear complexity spatio-temporal graph neural network for traffic forecasting on large-scale road networks, Proc. VLDB Endow. 17 (5) (2024) 1081–1090. doi:10.14778/3641204.3641217

    work page doi:10.14778/3641204.3641217 2024

  20. [20]

    M. Xu, W. Dai, C. Liu, X. Gao, W. Lin, G.-J. Qi, H. Xiong, Spatial- temporal transformer networks for traffic flow forecasting, arXiv preprint arXiv:2001.02908 (2020)

    work page arXiv 2001

  21. [21]

    Z. Fang, Q. Long, G. Song, K. Xie, Spatial-temporal graph ode networks for traffic flow forecasting, in: Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, ACM, 2021, p. 364–373.doi:10.1145/3447548.3467430

    work page doi:10.1145/3447548.3467430 2021

  22. [22]

    F. Li, J. Feng, H. Yan, G. Jin, F. Yang, F. Sun, D. Jin, Y. Li, Dynamic graph convolutional recurrent network for traffic prediction: Benchmark and solution, ACM Trans.Knowl.Discov.D. 17 (1) (2023) 1–21

    work page 2023

  23. [23]

    Z. Shao, Z. Zhang, W. Wei, F. Wang, Y. Xu, X. Cao, C. S. Jensen, Decoupled dynamic spatial-temporal graph neural network for traffic forecasting, Proc. VLDB Endow. 15 (11) (2022) 2733–2746

    work page 2022

  24. [24]

    Z. Li, L. Xia, Y. Xu, C. Huang, FlashST: A simple and universal prompt- tuning framework for traffic prediction, in: R. Salakhutdinov, Z. Kolter, K. Heller, A. Weller, N. Oliver, J. Scarlett, F. Berkenkamp (Eds.), Proceedings of the 41st International Conference on Machine Learning, Vol. 235, PMLR, 2024, pp. 28978–28988

    work page 2024

  25. [25]

    Z. Shao, Z. Zhang, F. Wang, W. Wei, Y. Xu, Spatial-temporal identity: A simple yet effective baseline for multivariate time series forecasting, in: Proceedings of the 31st ACM International Conference on Information and Knowledge Management, ACM, 2022, p. 4454–4458.doi:10.1145/ 3511808.3557702

    work page arXiv 2022

  26. [26]

    J. Deng, X. Chen, R. Jiang, X. Song, I. W. Tsang, ST-Norm: Spatial and temporal normalization for multi-variate time series forecasting, 32 in: Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, ACM, 2021, p. 269–278.doi:10.1145/ 3447548.3467330

    work page arXiv 2021

  27. [27]

    H. Abdi, L. J. Williams, Principal component analysis, Wires.Comput.Stat. 2 (4) (2010) 433–459

    work page 2010

  28. [28]

    McInnes, J

    L. McInnes, J. Healy, N. Saul, L. Grossberger, Umap: Uniform manifold approximation and projection, Joss. 3 (29) (2018) 861

    work page 2018

  29. [29]

    Node2vec: Scalable feature learning for networks,

    A. Grover, J. Leskovec, node2vec: Scalable feature learning for networks, in: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, ACM, 2016, p. 855–864. doi:10.1145/2939672.2939754

    work page doi:10.1145/2939672.2939754 2016

  30. [30]

    M. Ou, P. Cui, J. Pei, Z. Zhang, W. Zhu, Asymmetric transitivity preserving graph embedding, in: Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining, 2016, pp. 1105–1114

    work page 2016

  31. [31]

    H. Wu, J. Xu, J. Wang, M. Long, Autoformer: Decomposition trans- formers with auto-correlation for long-term series forecasting, in: 35th Conference on Neural Information Processing Systems, NeurIPS, 2021, pp. 22419–22430

    work page 2021

  32. [32]

    H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, W. Zhang, Informer: Beyond efficient transformer for long sequence time-series forecasting, in: Proceedings of the AAAI Conference on Artificial In- telligence, Vol. 35, AAAI, 2021, pp. 11106–11115.doi:10.1609/aaai. v35i12.17325

    work page doi:10.1609/aaai 2021

  33. [33]

    O. D. Anderson, Time-series. 2nd edn., Journal of the Royal Statistical Society. Series D (The Statistician) 25 (4) (1976) 308–310

    work page 1976

  34. [34]

    R. J. Hyndman, G. Athanasopoulos, Forecasting: principles and practice, OTexts, 2018

    work page 2018

  35. [35]

    2024);https://www

    Global Forecast System (NOAA, accessed 1 Sept. 2024);https://www. ncei.noaa.gov/. 33

    work page 2024

  36. [36]

    Hersbach, B

    H. Hersbach, B. Bell, P. Berrisford, S. Hirahara, A. Horányi, J. Muñoz- Sabater, J. Nicolas, C. Peubey, R. Radu, D. Schepers, et al., The era5 global reanalysis, Quarterly journal of the royal meteorological society 146 (730) (2020) 1999–2049

    work page 2020

  37. [37]

    B. N. Oreshkin, G. Dudek, P. Pełka, E. Turkina, N-beats neural network for mid-term electricity load forecasting, Applied Energy 293 (2021) 116918

    work page 2021

  38. [38]

    Lee-Thorp, J

    J. Lee-Thorp, J. Ainslie, I. Eckstein, S. Ontanon, FNet: Mixing tokens with Fourier transforms, in: M. Carpuat, M.-C. de Marneffe, I. V. Meza Ruiz (Eds.), Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, ACL, 2022, pp. 4296–4313.doi:10. 18653/v1/2022.naacl-main.319

    work page 2022

  39. [39]

    D. Cao, Y. Wang, J. Duan, C. Zhang, X. Zhu, C. Huang, Y. Tong, B. Xu, J. Bai, J. Tong, et al., Spectral temporal graph neural network for multivariate time-series forecasting, in: 34th Conference on Neural Information Processing Systems, NeurIPS, 2020, pp. 17766–17778

    work page 2020

  40. [40]

    T. Zhou, P. Niu, X. Wang, L. Sun, R. Jin, One fits all: power general time series analysis by pretrained lm, in: 37th Conference on Neural Information Processing Systems, NeurIPS, 2023, pp. 43322–43355

    work page 2023

  41. [41]

    S. Wang, H. Wu, X. Shi, T. Hu, H. Luo, L. Ma, J. Y. Zhang, J. ZHOU, TimeMixer: Decomposable multiscale mixing for time series forecasting, in: International Conference on Learning Representations https:// openreview.net/forum?id=7oLshfEIC2, 2024

    work page 2024

  42. [42]

    Banerjee, M

    S. Banerjee, M. Dong, W. Shi, Spatial–temporal synchronous graph transformer network (STSGT) for COVID-19 forecasting, Smart.Health. 26 (2022) 100348

    work page 2022

  43. [43]

    J. Xue, T. Yabe, K. Tsubouchi, J. Ma, S. Ukkusuri, Multiwave COVID-19 prediction from social awareness using web search and mobility data, in: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, ACM, 2022, p. 4279–4289.doi:10.1145/ 3534678.3539172. 34

    work page arXiv 2022

  44. [44]

    X. Pu, J. Zhu, Y. Wu, C. Leng, Z. Bo, H. Wang, Dynamic adap- tive spatio–temporal graph network for COVID-19 forecasting, CAAI Trans.Intell.Technol. 9 (3) (2024) 769–786

    work page 2024

  45. [45]

    S. Deng, S. Wang, H. Rangwala, L. Wang, Y. Ning, Cola-GNN: Cross- location attention based graph neural networks for long-term ili pre- diction, in: Proceedings of the 29th ACM International Conference on Information and Knowledge Management, ACM, 2020, p. 245–254. doi:10.1145/3340531.3411975

    work page doi:10.1145/3340531.3411975 2020

  46. [46]

    1085–1088.doi:10.1145/3209978.3210077

    Y.Wu, Y.Yang, H.Nishiura, M.Saitoh, Deeplearningforepidemiological predictions, in: The 41st International ACM SIGIR Conference on Research and Development in Information Retrieval, ACM, 2018, p. 1085–1088.doi:10.1145/3209978.3210077

    work page doi:10.1145/3209978.3210077 2018

  47. [47]

    J. Gao, R. Sharma, C. Qian, L. M. Glass, J. Spaeder, J. Romberg, J. Sun, C. Xiao, STAN: spatio-temporal attention network for pandemic prediction using real-world evidence, Journal of the American Medical Informatics Association 28 (4) (2021) 733–743

    work page 2021

  48. [48]

    F. Xie, Z. Zhang, L. Li, B. Zhou, Y. Tan, EpiGNN: Exploring spatial transmission with graph neural network for regional epidemic forecasting, in: Joint European Conference on Machine Learning and Knowledge Discovery in Databases, Springer, 2022, p. 469–485. doi:10.1007/ 978-3-031-26422-1_29

    work page 2022

  49. [49]

    Sutskever, O

    I. Sutskever, O. Vinyals, Q. V. Le, Sequence to sequence learning with neural networks, in: 28th International Conference on Neural Information Processing Systems, NeurIPS, 2014, pp. 3104–3112. 35

    work page 2014