Pith. sign in

REVIEW 73 references

Proxy Reconstruction Pre-training for Ramp Flow Prediction at Highway Interchanges

T0 review · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that ramp flows at complex highway interchanges can be predicted in real time using only mainline ETC gantry data, by pre-training a spatial-temporal decoupled autoencoder to reconstruct historical ramp flows, and that the

arxiv 2510.03381 v3 pith:HYBCYF46 submitted 2025-10-03 cs.LG cs.AI

Yongchao Li , Jun Chen , Zhuoxuan Li , Chao Gao , Yang Li , Chu Zhang , Changyin Dong This is my paper · ORCID
classification cs.LGcs.AI
keywords trafficflowpredictionramphighwayinterchangepre-trainingautoencoderETCdataspatio-temporalmodelingmissing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that real-time ramp-flow prediction at highway interchanges can be done without any ramp sensors or live license-plate matching, using only mainline toll-gantry data. It proposes a two-stage framework: first, a decoupled autoencoder reconstructs historical ramp flows from long mainline series; then the learned spatial and temporal representations are injected into a downstream forecaster. On three real cloverleaf interchanges and at 3-, 5-, and 10-minute sampling intervals, the combined model ranks first or near-first among thirteen baselines and matches or beats models that directly consume historical ramp data. If correct, this would remove a practical blind spot for traffic management and make fine-grained ramp control feasible wherever mainline ETC gantries exist.

What carries the argument

The central object is STDAE, a Spatio-Temporal Decoupled Autoencoder with two parallel Transformer-based branches: a Spatial AutoEncoder (SAE) and a Temporal AutoEncoder (TAE). Mainline features are fused per ramp by concatenating the upstream and downstream gantry features, then patched, normalized, and given 2D sinusoidal positional encodings; optional binary masks hide parts of the input. SAE computes attention across ramps at a fixed time, TAE computes attention across time for a fixed ramp, and each branch reconstructs the target ramp flow. The two representations are complementary and are fused into the downstream predictor's hidden state through MLP projections, which is what carries

What would settle it

A direct test is cross-interchange transfer: pre-train STDAE on mainline data from two interchanges, withhold all ramp labels from a third interchange, and predict that interchange's ramps; if accuracy collapses to the non-pretrained baseline, the model has not learned a general mainline-to-ramp mapping. A second concrete test is to corrupt the license-plate-matched ramp labels (for example, by randomly dropping a fraction of matched vehicle records) and check whether the reported MAE gains shrink proportionally, which would reveal how much of the apparent improvement depends on label quality.

Watch

Extended reading notes

Core claim

The central claim is that a proxy reconstruction pre-training task can fill the real-time ramp-data blind spot. During pre-training, the STDAE autoencoder consumes long mainline series (volume, speed, time-of-day, day-of-week) from eight upstream and downstream ETC gantries and reconstructs the paired historical ramp-flow sequences; optional spatial and temporal masks simulate missing gantries. A spatial autoencoder branch attends across ramps at a fixed time step, while a temporal autoencoder branch attends across time for each ramp; both branches reconstruct the same ramp target, forcing the model to learn mainline-to-ramp mappings. In the prediction stage, the learned representations are

Load-bearing premise

The result stands or falls on whether ramp flows at the three interchanges are actually recoverable from the eight selected mainline ETC gantries and whether the license-plate-matched ramp labels used for training are accurate; if either fails, the proxy reconstruction task has no learnable signal and the reported gains could be artifacts of longer input sequences and larger model capacity.

Editorial extensions

If this is right

  • Real-time ramp-flow prediction becomes possible at interchanges that have only mainline ETC gantries, with no requirement for ramp-mounted detectors or live license-plate matching.
  • The pretraining module is architecture-agnostic: adding STDAE representations improves several downstream predictors, including recurrent, graph-convolutional, and Transformer-based forecasters.
  • Masking during pretraining yields robustness to missing mainline data; the paper reports an average MAE reduction of about 2.23% over the non-pretrained baseline under simulated directional and temporal missingness.
  • The method adapts to multiple sampling intervals (3, 5, and 10 minutes), suggesting it can be tuned to different operational forecasting horizons.
  • Because only mainline data is needed at deployment, agencies could avoid the privacy and integration delays associated with deriving ramp flows from vehicle-path matching.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: If the learned mainline-to-ramp mapping generalizes beyond the three studied interchanges, the same proxy-reconstruction idea could be applied to other unobserved flow quantities estimated from nearby observed sensors, such as turning movements at intersections or origin-destination flows.
  • Inference: The per-ramp results suggest the pretraining gains concentrate on high-volume, complex movements; for low-volume or structurally simple ramps, the extra context may add little, so a selective or adaptive application could reduce computational cost without sacrificing accuracy.
  • Inference: A natural testable extension is cross-interchange transfer: pre-train on one or two interchanges and predict at a held-out interchange with no ramp labels at all, which would indicate whether the learned relations are tied to specific geometry or capture a general traffic-coupling mechanism.
  • Inference: The paper's own stated limitations, such as ignoring weather, incidents, and urban-network generalization, define the key open questions; the practical value of the method depends on whether the reported gains persist under those realistic perturbations.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the core claim rests on a held-out time split, with only non-load-bearing self-citations in related work.

full rationale

The central claim is empirical rather than derived: STDAE reconstructs historical ramp flows from mainline ETC features in pre-training (Eqs. 3-19), and the downstream GWNet predictor is evaluated on a temporally held-out test set (last 3 days of a 17:3:3 split, Section 4.3). No fitted parameter or label from the pre-training stage is reused as the prediction output; the reconstruction target (historical ramp flow) and the evaluation target (future ramp flow on unseen days) are distinct, and the prediction stage only injects representations extracted from mainline inputs (Eq. 22). The decoupled TAE/SAE architecture is credited to the external STDMAE paper [28], not to the authors' own prior work, and the cited self-references ([41], [42], [52]) support only generic LSTM/Transformer background, not the paper's novel proxy-reconstruction claim; they are therefore not load-bearing. The limitations stated in Section 6 (weather/incidents not modeled, limited interchange scope, deployment cost) are acknowledged scope restrictions, not evidence of circularity. Any concern about the accuracy of license-plate-matching labels (Section 4.3) is a data-quality assumption, not a constructional equivalence between inputs and predictions. Consequently, no step reduces by definition to its own input.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced; STDAE is a neural architecture. The free parameters are standard hyperparameters, but the long input horizon T_long=1 day is a particularly important choice because it gives the augmented model access to far more context than the baselines. The domain assumptions about learnability of ramp flows from mainline data and the accuracy of plate-matching labels are load-bearing and unvalidated.

free parameters (6)
  • T_long = 1 day (long input horizon) = 288/480/144 steps depending on interval
    Chosen in §5.1 for pre-training; this long context is a major source of the STDAE advantage and is not matched in baselines.
  • T = 12 (short prediction input) = 12 steps
    Set in §5.1 following [28,73]; controls baseline input length.
  • Patch length L = 12 = 12
    Set in §5.1; determines how many patches P = T_long/L and the T'=1 fusion.
  • Embedding dimension G = 96 = 96
    Set in §5.1; capacity hyperparameter.
  • Encoder/decoder layers and heads (4/1, 4 heads) = 4 encoder layers, 1 decoder layer, 4 heads
    Set in §5.1; architecture capacity hyperparameters.
  • Initial learning rate 0.002 = 0.002
    Set in §5.1; Adam hyperparameter.
assumptions (4)
  • domain assumption Ramp flow at each of the 12 movements is a learnable function of the eight selected upstream/downstream mainline ETC gantry features (flow, speed, time-of-day, day-of-week).
    Invoked in §3.2.1 Eq. (3) and the pre-training objective; if mainline counts do not determine ramp volumes, the proxy reconstruction task is unlearnable.
  • domain assumption Ramp-flow ground truth obtained by license-plate matching upstream/downstream gantries is accurate enough to train and evaluate on.
    Section 4.3 step G computes ramp flows by plate matching; no validation against observed ramp counts is reported, and data silos/missing records are acknowledged in §1.
  • domain assumption The 23-day ETC sample (17 train / 3 validation / 3 test days) is representative of traffic regimes for generalization.
    Section 4.3 reports the split; the test period is only three days, so the reported ranks may not generalize.
  • standard math Transformer attention, LayerNorm, Conv2D patch embedding, and sinusoidal positional encoding work as specified.
    Used throughout §3.2 without proof; standard components from cited literature.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Proxy Reconstruction Pre-training for Ramp Flow Prediction at Highway Interchanges." pith.science (2026). https://pith.science/paper/HYBCYF46

@misc{pith2026251003381,
  author       = {Pith},
  title        = {Pith review of: Proxy Reconstruction Pre-training for Ramp Flow Prediction at Highway Interchanges},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HYBCYF46}},
  note         = {Machine review of arXiv:2510.03381}
}
read the original abstract

Interchanges are crucial nodes for vehicle transfers between highways, yet the lack of real-time ramp detectors creates blind spots in traffic prediction. To address this, we propose a Spatio-Temporal Decoupled Autoencoder (STDAE), a two-stage framework that leverages cross-modal reconstruction pretraining. In the first stage, STDAE reconstructs historical ramp flows from mainline data, forcing the model to capture intrinsic spatio-temporal relations. Its decoupled architecture with parallel spatial and temporal autoencoders efficiently extracts heterogeneous features. In the prediction stage, the learned representations are integrated with models such as GWNet to enhance accuracy. Experiments on three real-world interchange datasets show that STDAE-GWNET consistently outperforms thirteen state-of-the-art baselines and achieves performance comparable to models using historical ramp data. This demonstrates its effectiveness in overcoming detector scarcity and its plug-and-play potential for diverse forecasting pipelines.

Figures

Figures reproduced from arXiv: 2510.03381 by the authors.

Figure 1
Figure 1. Illustration of the three primary challenges faced by ramp flow prediction based on mainline data. (a) Data silos: vehicle records are [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. STDAE Pre-training and Prediction Framework [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Data processing workflow, including the determination of the data collection scope, data collection, data cleaning, and feature extraction. [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison of MAPE across different sampling intervals. (a) Average MAPE of 14 models. (b) MAPE of the STDAEGWNET model. [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Comparison of STDAE and its ablated versions on three datasets with 5 min sampling interval. [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: MAE Comparison of STDAEGWNET with and without STDAE under Different Sampling Intervals and Significance Analysis. (a) 3 [PITH_FULL_IMAGE:figures/full_fig_p019_6.png]
Figure 7
Figure 7. Figure 7: Ramp-level Prediction of the QiLin Interchange with a 5 min Sampling Interval (step = 3). Subfigure labels correspond to the source [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 2 linked inside Pith

  1. [1]

    W. Guan, S. He, Statistical features of traffic flow on urban freeways, Physica A: Statistical Mechanics and its Applications 387 (4) (2008) 944–954

    2008

  2. [2]

    H. C. Manual, Highway capacity manual, Washington, DC 2 (1) (2000) 1

    2000

  3. [3]

    N. J. Garber, M. D. Fontaine, et al., Guidelines for preliminary selection of the optimum interchange type for a specific location, Tech. rep., Virginia Transportation Research Council (VTRC) (1999)

    1999

  4. [4]

    Q. Ma, X. Wang, S. Niu, H. Zeng, S. Ullah, Analysis on congestion mechanism of cavs around traffic accident zones, Accident Analysis & Prevention 205 (2024) 107663

    2024

  5. [5]

    C. Mu, L. Du, Y. An, X. Zhao, Compliance-constrained resilient system optimal trajectory planning for cavs at on-ramp intersection with multiple lanes, Transportation Research Part B: Methodological 194 (2025) 103173

    2025

  6. [6]

    C.D.Higgins,M.N.Sweet,P.S.Kanaroglou,Allminutesarenotequal: traveltimeandtheeffectsofcongestion on commute satisfaction in canadian cities, Transportation 45 (5) (2018) 1249–1268

    2018

  7. [7]

    L. Hu, J. X. Zhu, Y. Wang, L. H. Lee, Joint design of fleet size, hub locations, and hub capacities for third-party logisticsnetworkswithroadcongestionconstraints,TransportationResearchPartE:LogisticsandTransportation Review 118 (2018) 568–588

    2018

  8. [8]

    X. Ma, A. Karimpour, Y.-J. Wu, Statistical evaluation of data requirement for ramp metering performance assessment, Transportation Research Part A: Policy and Practice 141 (2020) 248–261

    2020

Show all 73 references
  1. [9]

    S.Liu,H.Hellendoorn,B.DeSchutter,Modelpredictivecontrolforfreewaynetworksbasedonmulti-classtraffic flow and emission models, IEEE Transactions on Intelligent Transportation Systems 18 (2) (2016) 306–320

    2016

  2. [10]

    H. Wang, J. Lai, X. Zhang, Y. Zhou, S. Li, J. Hu, Make space to change lane: A cooperative adaptive cruise control lane change controller, Transportation research part C: emerging technologies 143 (2022) 103847. 21

    2022

  3. [11]

    Zhang, F

    W. Zhang, F. Zhu, Y. Lv, C. Tan, W. Liu, X. Zhang, F.-Y. Wang, Adapgl: An adaptive graph learning algo- rithm for traffic prediction based on spatiotemporal neural networks, Transportation Research Part C: Emerging Technologies 139 (2022) 103659

    2022

  4. [12]

    Y. Han, M. Ramezani, A. Hegyi, Y. Yuan, S. Hoogendoorn, Hierarchical ramp metering in freeways: An aggregated modeling and control approach, Transportation research part C: emerging technologies 110 (2020) 1–19

    2020

  5. [13]

    Y. Kan, Y. Wang, D. Wang, J. Sun, C. Shao, M. Papageorgiou, A novel approach to estimating missing pairs of on/off ramp flows, IEEE Transactions on Intelligent Transportation Systems 22 (2) (2020) 1287–1305

    2020

  6. [14]

    Nohekhan, S

    A. Nohekhan, S. Zahedian, A. Haghani, A deep learning model for off-ramp hourly traffic volume estimation, Transportation Research Record 2675 (7) (2021) 350–362

    2021

  7. [15]

    Y. Duan, Y. Lv, Y.-L. Liu, F.-Y. Wang, An efficient realization of deep learning for traffic data imputation, Transportation research part C: emerging technologies 72 (2016) 168–181

    2016

  8. [16]

    Marzano, A

    V. Marzano, A. Papola, F. Simonelli, Limits and perspectives of effective o–d matrix correction using traffic counts, Transportation Research Part C: Emerging Technologies 17 (2) (2009) 120–132

    2009

  9. [17]

    R. Shi, Z. Mo, K. Huang, X. Di, Q. Du, A physics-informed deep learning paradigm for traffic state and fundamentaldiagramestimation,IEEETransactionsonIntelligentTransportationSystems23(8)(2021)11688– 11698

    2021

  10. [18]

    M.Pereira,A.Lang,B.Kulcsár,Short-termtrafficpredictionusingphysics-awareneuralnetworks,Transportation research part C: emerging technologies 142 (2022) 103772

    2022

  11. [19]

    J.Zhang, C.Song, Z.Mo, S.Cao, Atransferlearning-basedapproachtoestimatingmissingpairsofon/offramp flows, IEEE Transactions on Intelligent Transportation Systems 25 (2) (2023) 1247–1262

    2023

  12. [20]

    X. Ma, A. Karimpour, Y.-J. Wu, Data-driven transfer learning framework for estimating on-ramp and off-ramp traffic flows, Journal of Intelligent Transportation Systems 29 (1) (2025) 67–80

    2025

  13. [21]

    Tsung, C.-T

    C.-K. Tsung, C.-T. Yang, S.-W. Yang, Visualizing potential transportation demand from etc log analysis using elk stack, IEEE Internet of Things Journal 7 (7) (2020) 6623–6633

    2020

  14. [22]

    J. Zhou, Y. Zheng, Z. Tan, B. Wei, X. Zhou, X. Ruan, Feasibility study of bridge load testing using ongoing traffic through rapid estimation of traffic load effects for network-level highway bridges assisted by site-specific etc data, Journal of Civil Structural Health Monitor...

    2025

  15. [23]

    Pospisil, J

    K. Pospisil, J. Fric, Implementation of strategic management tools to operation of highway agencies, in: 3rd International multidisciplinary scientific conference on social sciences and arts, SGEM, 2016

    2016

  16. [24]

    Jiang, J

    W. Jiang, J. Luo, Graph neural network for traffic forecasting: A survey, Expert systems with applications 207 (2022) 117921

    2022

  17. [25]

    D. A. Tedjopurnomo, Z. Bao, B. Zheng, F. M. Choudhury, A. K. Qin, A survey on modern deep neural network for traffic prediction: Trends, methods and challenges, IEEE Transactions on Knowledge and Data Engineering 34 (4) (2020) 1544–1561

    2020

  18. [26]

    6778–6786, survey Track

    Q.Wen,T.Zhou,C.Zhang,W.Chen,Z.Ma,J.Yan,L.Sun,Transformersintimeseries: Asurvey,in: E.Elkind (Ed.), Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, IJCAI-23, International Joint Conferences on Artificial Intelligence Organization, 202...

    2023 doi

  19. [27]

    K. He, X. Chen, S. Xie, Y. Li, P. Dollár, R. Girshick, Masked autoencoders are scalable vision learners, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 16000–16009. 22

    2022

  20. [28]

    H. Gao, R. Jiang, Z. Dong, J. Deng, Y. Ma, X. Song, Spatial-temporal-decoupled masked pre-training for spa- tiotemporal forecasting, in: K. Larson (Ed.), PROCEEDINGS OF THE THIRTY-THIRD INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, IJCAI 2024, 2024, pp. 3998–4006,...

    2024

  21. [29]

    Z. Shao, F. Wang, Y. Xu, W. Wei, C. Yu, Z. Zhang, D. Yao, T. Sun, G. Jin, X. Cao, et al., Exploring progressinmultivariatetimeseriesforecasting: Comprehensivebenchmarkingandheterogeneityanalysis,IEEE Transactions on Knowledge and Data Engineering 37 (1) (2024) 291–305

    2024

  22. [30]

    L. Zhao, Y. Song, C. Zhang, Y. Liu, P. Wang, T. Lin, M. Deng, H. Li, T-gcn: A temporal graph convolutional network for traffic prediction, IEEE transactions on intelligent transportation systems 21 (9) (2019) 3848–3858

    2019

  23. [31]

    Z. Liu, F. Ding, Y. Dai, L. Li, T. Chen, H. Tan, Spatial-temporal graph convolution network model with traffic fundamentaldiagraminformationinformedfornetworktrafficflowprediction,ExpertSystemswithApplications 249 (2024) 123543

    2024

  24. [32]

    M. M. Hamed, H. R. Al-Masaeid, Z. M. B. Said, Short-term prediction of traffic volume in urban arterials, Journal of Transportation Engineering 121 (3) (1995) 249–254

    1995

  25. [33]

    G.P.Zhang,Timeseriesforecastingusingahybridarimaandneuralnetworkmodel,Neurocomputing50(2003) 159–175

    2003

  26. [34]

    B. M. Williams, L. A. Hoel, Modeling and forecasting vehicular traffic flow as a seasonal arima process: Theoretical basis and empirical results, Journal of transportation engineering 129 (6) (2003) 664–672

    2003

  27. [35]

    Okutani, Y

    I. Okutani, Y. J. Stephanedes, Dynamic prediction of traffic volume through kalman filtering theory, Transporta- tion Research Part B: Methodological 18 (1) (1984) 1–11

    1984

  28. [36]

    X. Xu, L. Zhang, Q. Kong, C. Gui, X. Zhang, Enhanced-historical average for long-term prediction, in: 2022 2ndInternationalConferenceonComputer,ControlandRobotics(ICCCR),2022,pp.115–119.doi:10.1109/ ICCCR54399.2022.9790148

    2022

  29. [37]

    Chen, C.-Y

    R. Chen, C.-Y. Liang, W.-C. Hong, D.-X. Gu, Forecasting holiday daily tourist flow based on seasonal support vector regression with adaptive genetic algorithm, Applied Soft Computing 26 (2015) 435–443

    2015

  30. [38]

    Zhang, G

    X.-l. Zhang, G. He, H. Lu, Short-term traffic flow forecasting based on k-nearest neighbors non-parametric regression, Journal of Systems Engineering 24 (2) (2009) 178–183

    2009

  31. [39]

    Z. Li, J. Cao, X. Shi, W. Huang, Qpso-ahes-rc: a hybrid learning model for short-term traffic flow prediction, Soft Computing 27 (14) (2023) 9347–9366

    2023

  32. [40]

    X. Yin, G. Wu, J. Wei, Y. Shen, H. Qi, B. Yin, Deep learning on traffic prediction: Methods, analysis, and future directions, IEEE Transactions on Intelligent Transportation Systems 23 (6) (2022) 4927–4943.doi: 10.1109/TITS.2021.3054840

    2022

  33. [41]

    Q.Hu, W.Liu, C.Yan, C.Zhang, J.Chen, C.Dong, Anovelapproachforunderstandingtheparkingdemandfor internal access roads in hub car parks, Engineering Applications of Artificial Intelligence 161 (2025) 112201

    2025

  34. [42]

    C. Dong, Z. Xiong, N. Li, X. Yu, M. Liang, C. Zhang, Y. Li, H. Wang, A real-time prediction framework for energy consumption of electric buses using integrated machine learning algorithms, Transportation Research Part E: Logistics and Transportation Review 194 (2025) 103884

    2025

  35. [43]

    G. Dai, C. Ma, X. Xu, Short-term traffic flow prediction method for urban road sections based on space–time analysis and gru, IEEE Access 7 (2019) 143025–143035

    2019

  36. [44]

    Jiang, L

    W. Jiang, L. Zhang, Geospatial data to images: A deep-learning framework for traffic forecasting, Tsinghua Science and Technology 24 (1) (2018) 52–64. 23

    2018

  37. [45]

    Méndez, M

    M. Méndez, M. G. Merayo, M. Núñez, Long-term traffic flow forecasting using a hybrid cnn-bilstm model, Engineering Applications of Artificial Intelligence 121 (2023) 106041

    2023

  38. [46]

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, P. S. Yu, A comprehensive survey on graph neural networks, IEEE transactions on neural networks and learning systems 32 (1) (2020) 4–24

    2020

  39. [47]

    B. Yu, H. Yin, Z. Zhu, Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting, in: J. Lang (Ed.), PROCEEDINGS OF THE TWENTY-SEVENTH INTERNATIONAL JOINT CONFERENCEONARTIFICIALINTELLIGENCE,2018,pp.3634–3640,27thInternationalJointConferenc...

    2018

  40. [48]

    Zheng, X

    C. Zheng, X. Fan, S. Pan, H. Jin, Z. Peng, Z. Wu, C. Wang, P. S. Yu, Spatio-temporal joint graph convolutional networks for traffic forecasting, IEEE Transactions on Knowledge and Data Engineering 36 (1) (2024) 372–385. doi:10.1109/TKDE.2023.3284156

    2024

  41. [49]

    G. He, W. Huang, Y. Zhu, M. Huang, Adaptive spatial-temporal dependence graph convolution neural network for traffic flow prediction, Expert Systems with Applications 281 (2025) 127564.doi:https://doi.org/10. 1016/j.eswa.2025.127564

    2025

  42. [50]

    Y. Liu, T. Hu, H. Zhang, H. Wu, S. Wang, L. Ma, M. Long, itransformer: Inverted transformers are effective for time series forecasting, in: The Twelfth International Conference on Learning Representations, 2024. URLhttps://openreview.net/forum?id=JePfAI8fah

    2024

  43. [51]

    X.-B. Jin, H. Ma, J.-Y. Xie, J. Kong, M. Deveci, S. Kadry, Ada-stgmat: An adaptive spatio-temporal graph multi-attention network for intelligent time series forecasting in smart cities, Expert Systems with Applications 269 (2025) 126428

    2025

  44. [52]

    C. Dong, Z. Xiong, C. Zhang, N. Li, Y. Li, N. Xie, J. Zhang, H. Wang, A transformer-based approach for deep feature extraction and energy consumption prediction of electric buses based on driving distances, Applied Energy 380 (2025) 123941

    2025

  45. [53]

    T.-H.Chang,Z.-Y.Li,Optimizationofmainlinetrafficviaanadaptiveco-ordinatedramp-meteringcontrolmodel with dynamic od estimation, Transportation Research Part C: Emerging Technologies 10 (2) (2002) 99–120

    2002

  46. [54]

    2114–2124

    G.Zerveas,S.Jayaraman,D.Patel,A.Bhamidipaty,C.Eickhoff,Atransformer-basedframeworkformultivariate time series representation learning, in: Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 2021, pp. 2114–2124

    2021

  47. [55]

    1567–1577

    Z.Shao,Z.Zhang,F.Wang,Y.Xu,Pre-trainingenhancedspatial-temporalgraphneuralnetworkformultivariate timeseriesforecasting,in: Proceedingsofthe28thACMSIGKDDconferenceonknowledgediscoveryanddata mining, 2022, pp. 1567–1577

    2022

  48. [56]

    H. Chi, Y. Lu, Y. Zhu, W. Ke, H. Mao, Dynamic spatiotemporal graph convolutional network collaborative pre-training learning for traffic flow prediction, Knowledge-Based Systems (2025) 114339

    2025

  49. [57]

    Q. Liu, J. Ye, H. Liang, L. Sun, B. Du, Ts-mae: A masked autoencoder for time series representation learning, Information Sciences 690 (2025) 121576

    2025

  50. [58]

    S.Fang,W.Ji,S.Xiang,W.Hua,Prestnet: Pre-trainedspatio-temporalnetworkfortrafficforecasting,Information Fusion 106 (2024) 102241

    2024

  51. [59]

    D. Cai, K. Chen, Z. Lin, D. Li, T. Zhou, Y. Ling, M.-F. Leung, Jointstnet: Joint pre-training for spatial-temporal traffic forecasting, IEEE Transactions on Consumer Electronics (2024)

    2024

  52. [60]

    F.Zhuang,Z.Qi,K.Duan,D.Xi,Y.Zhu,H.Zhu,H.Xiong,Q.He,Acomprehensivesurveyontransferlearning, Proceedings of the IEEE 109 (1) (2020) 43–76. 24

    2020

  53. [61]

    P. Liu, W. Yuan, J. Fu, Z. Jiang, H. Hayashi, G. Neubig, Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing, ACM computing surveys 55 (9) (2023) 1–35

    2023

  54. [62]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural information processing systems 30 (2017)

    2017

  55. [63]

    Wang, J.-C

    Z. Wang, J.-C. Liu, Translating math formula images to latex sequences using deep neural networks with sequence-level training, International Journal on Document Analysis and Recognition (IJDAR) 24 (1) (2021) 63–75

    2021

  56. [64]

    1907–1913.doi:10.24963/ijcai.2019/264

    Z.Wu,S.Pan,G.Long,J.Jiang,C.Zhang,Graphwavenetfordeepspatial-temporalgraphmodeling,in: Proceed- ingsoftheTwenty-EighthInternationalJointConferenceonArtificialIntelligence,IJCAI-19,InternationalJoint Conferences on Artificial Intelligence Organization, 2019, pp. 1907–1913.doi:1...

    2019 doi

  57. [65]

    L. Bai, L. Yao, C. Li, X. Wang, C. Wang, Adaptive graph convolutional recurrent network for traffic forecasting, Advances in neural information processing systems 33 (2020) 17804–17815

    2020

  58. [66]

    Y.Li,R.Yu,C.Shahabi,Y.Liu,Diffusionconvolutionalrecurrentneuralnetwork: Data-driventrafficforecasting, arXiv preprint arXiv:1707.01926 (2017)

    2017 arXiv

  59. [67]

    Z. Wu, S. Pan, G. Long, J. Jiang, X. Chang, C. Zhang, Connecting the dots: Multivariate time series forecasting with graph neural networks, in: Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, 2020, pp. 753–763

    2020

  60. [68]

    38, 2024, pp

    W.Kong,Z.Guo,Y.Liu,Spatio-temporalpivotalgraphneuralnetworksfortrafficflowforecasting,in: Proceed- ings of the AAAI conference on artificial intelligence, Vol. 38, 2024, pp. 8627–8635

    2024

  61. [69]

    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, arXiv preprint arXiv:2206.09112 (2022)

    2022 arXiv

  62. [70]

    J. Deng, X. Chen, R. Jiang, X. Song, I. W. Tsang, St-norm: Spatial and temporal normalization for multi-variate time series forecasting, in: Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 2021, pp. 269–278

    2021

  63. [71]

    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, 2023, pp. 4125–4129

    2023

  64. [72]

    Z. Dong, R. Jiang, H. Gao, H. Liu, J. Deng, Q. Wen, X. Song, Heterogeneity-informed meta-parameter learning forspatiotemporaltimeseriesforecasting, in: Proceedingsofthe30thACMSIGKDDconferenceonknowledge discovery and data mining, 2024, pp. 631–641

    2024

  65. [73]

    34, 2020, pp

    C.Song,Y.Lin,S.Guo,H.Wan,Spatial-temporalsynchronousgraphconvolutionalnetworks: Anewframework for spatial-temporal network data forecasting, in: Proceedings of the AAAI conference on artificial intelligence, Vol. 34, 2020, pp. 914–921. 25

    2020

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.