pith. sign in

arxiv: 2602.02528 · v2 · submitted 2026-01-27 · 💻 cs.LG · cs.AI

Incident-Guided Spatiotemporal Traffic Forecasting

Lixiang Fan , Bohao Li , Tao Zou , Junchen Ye , Bowen Du This is my paper

Pith reviewed 2026-05-16 10:42 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords traffic forecastingspatiotemporal graph neural networksincident impactdeep learning for transportationgraph neural networksexternal disturbances
0
0 comments X

The pith

Incorporating time-aligned incident records into graph neural networks improves traffic forecasts by modeling their spatial spread and temporal decay.

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

Standard traffic prediction models rely only on historical flow data and therefore miss the abrupt shifts triggered by incidents such as accidents or severe weather. The paper introduces IGSTGNN, which adds two modules: ICSF to fuse the initial, location-specific spatial effects of each incident across the road network, and TIID to track how those effects weaken over successive time steps. A new large dataset is released that pairs incident logs with traffic time series. On this benchmark the full framework reaches state-of-the-art accuracy, and the two modules can be dropped into other existing models to raise their performance as well.

Core claim

The IGSTGNN framework explicitly models the impact of transportation incidents through an Incident-Context Spatial Fusion module that captures heterogeneous spatial influences and a Temporal Incident Impact Decay module that models dynamic dissipation, achieving state-of-the-art performance on a new benchmark dataset with time-aligned incident and traffic data.

What carries the argument

Incident-Context Spatial Fusion (ICSF) module to capture initial heterogeneous spatial influence of incidents and Temporal Incident Impact Decay (TIID) module to model subsequent dynamic dissipation of those effects on traffic graphs.

If this is right

  • Traffic forecasts become more accurate during periods containing incidents because the model no longer treats those periods as ordinary historical sequences.
  • The ICSF and TIID modules can be inserted into a wide range of existing spatiotemporal models and still produce measurable gains.
  • The released dataset supplies the first large-scale resource for studying how incidents alter traffic dynamics in both space and time.
  • Real-time transportation systems can now respond to external disturbances without retraining the entire forecasting pipeline from scratch.

Where Pith is reading between the lines

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

  • The same two-module pattern could be tested on other sudden-event domains such as power-grid load or social-media cascade forecasting.
  • If the decay schedule in TIID proves stable across cities, the module could be frozen and reused without city-specific retraining.
  • Pairing the framework with an incident-detection front end would allow the model to act on events before they fully propagate.

Load-bearing premise

The spatial and temporal effects of incidents can be captured by the ICSF and TIID modules without requiring the model to predict future incidents or handle unobserved confounding factors.

What would settle it

An experiment in which IGSTGNN without the ICSF and TIID modules shows no accuracy gain over standard spatiotemporal baselines on the incident-aligned dataset would falsify the claim that these modules are necessary to capture the relevant effects.

Figures

Figures reproduced from arXiv: 2602.02528 by Bohao Li, Bowen Du, Junchen Ye, Lixiang Fan, Tao Zou.

Figure 1
Figure 1. Figure 1: Illustration of a conventional model’s failure under [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The overall architecture of our proposed IGSTGNN framework. (a) illustrates the main pipeline of the model, while [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Performance comparison (in terms of average MAE [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Effectiveness of the ICSF and TIID modules. The fig￾ure compares the performance (MAE and RMSE) of baseline models in their original form (Raw) versus after integrat￾ing the modules separately (ICSF, TIID) and in combination (BOTH) on the Alameda and Contra Costa datasets. performance improvement soared to 13.23%, proving their func￾tional complementarity. This trend is highly consistent across di￾verse mo… view at source ↗
Figure 5
Figure 5. Figure 5: MAPE (%) results of the ICSF module superiority study. The MAPE results, shown in [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
read the original abstract

Recent years have witnessed the rapid development of deep-learning-based, graph-neural-network-based forecasting methods for modern intelligent transportation systems. However, most existing work focuses exclusively on capturing spatio-temporal dependencies from historical traffic data, while overlooking the fact that suddenly occurring transportation incidents, such as traffic accidents and adverse weather, serve as external disturbances that can substantially alter temporal patterns. We argue that this issue has become a major obstacle to modeling the dynamics of traffic systems and improving prediction accuracy, but the unpredictability of incidents makes it difficult to observe patterns from historical sequences. To address these challenges, this paper proposes a novel framework named the Incident-Guided Spatiotemporal Graph Neural Network (IGSTGNN). IGSTGNN explicitly models the incident's impact through two core components: an Incident-Context Spatial Fusion (ICSF) module to capture the initial heterogeneous spatial influence, and a Temporal Incident Impact Decay (TIID) module to model the subsequent dynamic dissipation. To facilitate research on the spatio-temporal impact of incidents on traffic flow, a large-scale dataset is constructed and released, featuring incident records that are time-aligned with traffic time series. On this new benchmark, the proposed IGSTGNN framework is demonstrated to achieve state-of-the-art performance. Furthermore, the generalizability of the ICSF and TIID modules is validated by integrating them into various existing models.

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

2 major / 2 minor

Summary. The paper proposes the Incident-Guided Spatiotemporal Graph Neural Network (IGSTGNN) for traffic forecasting. It introduces an Incident-Context Spatial Fusion (ICSF) module to capture initial heterogeneous spatial effects of incidents and a Temporal Incident Impact Decay (TIID) module to model subsequent dynamic dissipation of those effects. A new large-scale dataset with time-aligned incident records and traffic time series is released. The authors claim that IGSTGNN achieves state-of-the-art performance on this benchmark and that the ICSF and TIID modules are generalizable when integrated into existing models.

Significance. If the evaluation protocol uses only historical incident information and the reported gains are not due to future leakage, the work would address a clear gap in spatiotemporal traffic modeling by explicitly incorporating external disturbances. The dataset release is a concrete positive contribution that enables reproducible study of incident effects.

major comments (2)
  1. [Evaluation and Experiments] The evaluation protocol must be clarified regarding whether incident records whose timestamps fall inside the prediction horizon are supplied as input to ICSF and TIID at test time. If future incidents are available, the SOTA improvements and plug-in gains cannot be attributed to learned capture of dynamic dissipation from observable history alone, undermining the central claim that the modules model incident impact without oracle access.
  2. [Experiments] The comparison to baselines is load-bearing for the generalizability claim: baselines receive only historical traffic data while augmented models receive additional incident features. Without an explicit statement that all models are given identical input information (restricted to past incidents), the reported improvements in §4 and the ablation tables cannot be interpreted as evidence for the architectural value of ICSF/TIID.
minor comments (2)
  1. [Methods] Notation for the decay function in the TIID module should be defined with an explicit equation rather than prose description to allow direct reproduction.
  2. [Abstract and Results] The abstract states SOTA results but the main text should include error bars, statistical significance tests, and full ablation tables with all baselines to support the quantitative claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the insightful and constructive comments. We address the concerns regarding the evaluation protocol and baseline comparisons below, and will revise the manuscript accordingly to ensure full clarity.

read point-by-point responses

  1. Referee: [Evaluation and Experiments] The evaluation protocol must be clarified regarding whether incident records whose timestamps fall inside the prediction horizon are supplied as input to ICSF and TIID at test time. If future incidents are available, the SOTA improvements and plug-in gains cannot be attributed to learned capture of dynamic dissipation from observable history alone, undermining the central claim that the modules model incident impact without oracle access.

    Authors: We confirm that only incident records with timestamps up to the start of the prediction horizon are supplied as input to ICSF and TIID during testing; no future incidents within the horizon are provided. The TIID module learns to model dynamic dissipation solely from observable historical incident data. We will add an explicit description of this protocol (including a diagram of the input timeline) in the revised experimental setup section to eliminate any ambiguity. revision: yes

  2. Referee: [Experiments] The comparison to baselines is load-bearing for the generalizability claim: baselines receive only historical traffic data while augmented models receive additional incident features. Without an explicit statement that all models are given identical input information (restricted to past incidents), the reported improvements in §4 and the ablation tables cannot be interpreted as evidence for the architectural value of ICSF/TIID.

    Authors: We agree this requires explicit clarification. Standard baselines receive only historical traffic data, as they have no mechanism for incident inputs. When demonstrating generalizability, we integrate ICSF and TIID into those same models and provide them with identical historical traffic plus past incident records (restricted to pre-horizon timestamps). We will add a clear statement in the revised Section 4 and experimental setup confirming that all models share the same historical inputs, with the value of ICSF/TIID shown through their ability to leverage the additional past incident features. revision: yes

Circularity Check

0 steps flagged

No significant circularity; new modules and dataset are independent of target results.

full rationale

The paper introduces ICSF and TIID as novel components to model incident effects, releases a new time-aligned dataset, and validates via SOTA benchmarks plus plug-in experiments on existing models. No equations, fitted parameters, or self-citations are shown that reduce any prediction or uniqueness claim to the inputs by construction. The derivation chain adds external structure (incident records) rather than redefining the forecasting target in terms of itself.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are detailed. The central claim rests on the assumption that incident impacts are separable into initial spatial fusion and subsequent temporal decay.

pith-pipeline@v0.9.0 · 5546 in / 1082 out tokens · 21073 ms · 2026-05-16T10:42:21.417781+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages

  1. [1]

    Abadi, Tooraj Rajabioun, and Petros A

    A. Abadi, Tooraj Rajabioun, and Petros A. Ioannou. 2015. Traffic Flow Prediction for Road Transportation Networks With Limited Traffic Data.IEEE Transactions on Intelligent Transportation Systems16 (2015), 653–662. https://doi.org/10.1109/ TITS.2014.2337238

    work page arXiv 2015

  2. [2]

    Shams Forruque Ahmed, Sweety Angela Kuldeep, Sabiha Jannat Rafa, Javeria Fazal, Mahfara Hoque, Gang Liu, and Amir H. Gandomi. 2024. Enhancement of traffic forecasting through graph neural network-based information fusion techniques.Inf. Fusion110 (2024), 102466. https://doi.org/10.1016/j.inffus.2024. 102466

    work page doi:10.1016/j.inffus.2024 2024

  3. [3]

    Al-Thani, Ziyu Sheng, Yuting Cao, and Yin Yang

    Mansoor G. Al-Thani, Ziyu Sheng, Yuting Cao, and Yin Yang. 2024. Traffic Transformer: Transformer-based framework for temporal traffic accident predic- tion.AIMS Mathematics9, 5 (2024), 12610–12629. https://doi.org/10.3934/math. 2024617

    work page doi:10.3934/math 2024

  4. [4]

    Ahmad Ali, Yanmin Zhu, and M. Zakarya. 2021. Exploiting dynamic spatio- temporal graph convolutional neural networks for citywide traffic flows predic- tion.Neural networks : the official journal of the International Neural Network Society145 (2021), 233–247. https://doi.org/10.1016/j.neunet.2021.10.021

    work page doi:10.1016/j.neunet.2021.10.021 2021

  5. [5]

    Jiyao An, Liang Guo, Wei Liu, Zhiqiang Fu, Ping Ren, Xinzhi Liu, and Tao Li

    work page

  6. [6]

    IGAGCN: Information geometry and attention-based spatiotemporal graph convolutional networks for traffic flow prediction.Neural Networks143 (2021), 355–367

    work page 2021

  7. [7]

    Lei Bai, Lina Yao, Can Li, Xianzhi Wang, and Can Wang. 2020. Adaptive graph convolutional recurrent network for traffic forecasting.Advances in neural information processing systems33 (2020), 17804–17815

    work page 2020

  8. [8]

    Razvan-Gabriel Cirstea, Bin Yang, Chenjuan Guo, Tung Kieu, and Shirui Pan

    work page

  9. [9]

    In38th IEEE International Conference on Data Engineering, ICDE 2022, Kuala Lumpur, Malaysia, May 9-12, 2022

    Towards Spatio- Temporal Aware Traffic Time Series Forecasting. In38th IEEE International Conference on Data Engineering, ICDE 2022, Kuala Lumpur, Malaysia, May 9-12, 2022. IEEE, Kuala Lumpur, Malaysia, 2900–2913. https: //doi.org/10.1109/ICDE53745.2022.00262

    work page doi:10.1109/icde53745.2022.00262 2022

  10. [10]

    Khosravi

    Fei Dai, Penggui Huang, Xiaolong Xu, Lianyong Qi, and M. Khosravi. 2020. Spatio-Temporal Deep Learning Framework for Traffic Speed Forecasting in IoT. IEEE Internet of Things Magazine3 (2020), 66–69. https://doi.org/10.1109/IOTM. 0001.2000031

    work page doi:10.1109/iotm 2020

  11. [11]

    Bowen Du, Hao Peng, Senzhang Wang, Md Zakirul Alam Bhuiyan, Lihong Wang, Qiran Gong, Lin Liu, and Jing Li. 2019. Deep irregular convolutional residual LSTM for urban traffic passenger flows prediction.IEEE Transactions on Intelligent Transportation Systems21, 3 (2019), 972–985

    work page 2019

  12. [12]

    Shengdong Du, Tao Yang, Fei Teng, Junbo Zhang, Tianrui Li, and Yu Zheng

    work page

  13. [13]

    Masset, R

    Multi-scale feature enhanced spatio-temporal learning for traffic flow forecasting.Knowl. Based Syst.294 (2024), 111787. https://doi.org/10.1016/j. knosys.2024.111787

    work page doi:10.1016/j 2024

  14. [14]

    Yuchen Fang, Fang Zhao, Yanjun Qin, Haiyong Luo, and Chenxing Wang. 2022. Learning All Dynamics: Traffic Forecasting via Locality-Aware Spatio-Temporal Joint Transformer.IEEE Transactions on Intelligent Transportation Systems23 (2022), 23433–23446. https://doi.org/10.1109/TITS.2022.3197640

    work page doi:10.1109/tits.2022.3197640 2022

  15. [15]

    Siyuan Feng, Shuqing Wei, Junbo Zhang, Yexin Li, Jintao Ke, Gaode Chen, Yu Zheng, and Hai Yang. 2023. A macro–micro spatio-temporal neural network for traffic prediction.Transportation Research Part C: Emerging Technologies156 (2023), 104331. https://doi.org/10.1016/j.trc.2023.104331

    work page doi:10.1016/j.trc.2023.104331 2023

  16. [16]

    Banishree Ghosh and J. Dauwels. 2021. Comparison of different Bayesian meth- ods for estimating error bars with incident duration prediction.Journal of Intelligent Transportation Systems26 (2021), 420 – 431. https://doi.org/10.1080/ 15472450.2021.1894936

    work page arXiv 2021

  17. [17]

    Xiaochuan Gou, Ziyue Li, Tian Lan, Junpeng Lin, Zhishuai Li, Bingyu Zhao, Chen Zhang, Di Wang, and Xiangliang Zhang. 2024. XTraffic: A Dataset Where Traffic Meets Incidents with Explainability and More. arXiv:2407.11477 [cs.LG] https://arxiv.org/abs/2407.11477

    work page arXiv 2024

  18. [18]

    Shengnan Guo, Youfang Lin, Ning Feng, Chao Song, and Huaiyu Wan. 2019. Attention Based Spatial-Temporal Graph Convolutional Networks for Traf- fic Flow Forecasting. InThe Thirty-Third AAAI Conference on Artificial In- telligence, AAAI 2019, The Thirty-First Innovative Applications of Artificial In- telligence Conference, IAAI 2019, The Ninth AAAI Symposiu...

    work page doi:10.1609/aaai.v33i01.3301922 2019

  19. [19]

    Liangzhe Han, Bowen Du, Leilei Sun, Yanjie Fu, Yisheng Lv, and Hui Xiong

    work page

  20. [20]

    InKDD ’21: The 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Virtual Event, Singapore, August 14-18, 2021, Feida Zhu, Beng Chin Ooi, and Chunyan Miao (Eds.)

    Dynamic and Multi-faceted Spatio-temporal Deep Learning for Traffic Speed Forecasting. InKDD ’21: The 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Virtual Event, Singapore, August 14-18, 2021, Feida Zhu, Beng Chin Ooi, and Chunyan Miao (Eds.). ACM, Singapore, 547–555. https: //doi.org/10.1145/3447548.3467275

    work page doi:10.1145/3447548.3467275 2021

  21. [21]

    Xiaolin Han, Tobias Grubenmann, Reynold Cheng, Sze Chun Wong, Xiaodong Li, and Wenya Sun. 2020. Traffic Incident Detection: A Trajectory-based Approach. In36th IEEE International Conference on Data Engineering, ICDE 2020, Dallas, TX, USA, April 20-24, 2020. IEEE, Dallas, TX, USA, 1866–1869. https://doi.org/10. 1109/ICDE48307.2020.00190

    work page arXiv 2020

  22. [22]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural computation9, 8 (1997), 1735–1780

    work page 1997

  23. [23]

    Renhe Jiang, Zhaonan Wang, Jiawei Yong, Puneet Jeph, Quanjun Chen, Yasumasa Kobayashi, Xuan Song, Shintaro Fukushima, and Toyotaro Suzumura. 2023. Spatio-Temporal Meta-Graph Learning for Traffic Forecasting. InThirty-Seventh AAAI Conference on Artificial Intelligence, AAAI 2023, Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence...

    work page doi:10.1609/aaai 2023

  24. [24]

    Wei Ju, Yusheng Zhao, Yifang Qin, Siyu Yi, Jingyang Yuan, Zhiping Xiao, Xiao Luo, Xiting Yan, and Ming Zhang. 2024. COOL: A Conjoint Perspective on Spatio-Temporal Graph Neural Network for Traffic Forecasting.Inf. Fusion107 (2024), 102341. https://doi.org/10.48550/arXiv.2403.01091

    work page doi:10.48550/arxiv.2403.01091 2024

  25. [25]

    Shiyong Lan, Yitong Ma, Weikang Huang, Wenwu Wang, Hongyu Yang, and Pyang Li. 2022. DSTAGNN: Dynamic Spatial-Temporal Aware Graph Neural Network for Traffic Flow Forecasting. InInternational Conference on Machine Learning, ICML 2022, 17-23 July 2022, Baltimore, Maryland, USA (Proceedings of Machine Learning Research, Vol. 162), Kamalika Chaudhuri, Stefani...

    work page 2022

  26. [26]

    Fuxian Li, Jie Feng, Huan Yan, Guangyin Jin, Fan Yang, Funing Sun, Depeng Jin, and Yong Li. 2023. Dynamic graph convolutional recurrent network for traffic prediction: Benchmark and solution.ACM Transactions on Knowledge Discovery from Data17, 1 (2023), 1–21

    work page 2023

  27. [27]

    Yaguang Li, Rose Yu, Cyrus Shahabi, and Yan Liu. 2018. Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting. In6th International Conference on Learning Representations, ICLR 2018, Vancouver, BC, Canada, April 30 - May 3, 2018, Conference Track Proceedings. OpenReview.net, Vancouver, BC, Canada. https://openreview.net/forum?id=...

    work page 2018

  28. [28]

    Rosenblum

    Yuxuan Liang, Kun Ouyang, Yiwei Wang, Ye Liu, Junbo Zhang, Yu Zheng, and David S. Rosenblum. 2020. Revisiting Convolutional Neural Networks for City- wide Crowd Flow Analytics. InMachine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2020, Ghent, Belgium, September 14-18, 2020, Proceedings, Part I (Lecture Notes in Computer...

    work page doi:10.1007/978-3-030-67658-2_33 2020

  29. [29]

    Xu Liu, Yutong Xia, Yuxuan Liang, Junfeng Hu, Yiwei Wang, Lei Bai, Chao Huang, Zhenguang Liu, Bryan Hooi, and Roger Zimmermann. 2023. Largest: A benchmark dataset for large-scale traffic forecasting.Advances in Neural Information Processing Systems36 (2023), 75354–75371

    work page 2023

  30. [30]

    Jiaming Ma, Binwu Wang, Pengkun Wang, Zhengyang Zhou, Xu Wang, and Yang Wang. 2025. BiST: A Lightweight and Efficient Bi-directional Model for Spatiotemporal Prediction.Proc. VLDB Endow.18, 6 (2025), 1663–1676. https: //www.vldb.org/pvldb/vol18/p1663-wang.pdf

    work page 2025

  31. [31]

    Qiwei Ma, Wei Sun, Junbo Gao, Pengwei Ma, and Mengjie Shi. 2022. Spatio- temporal adaptive graph convolutional networks for traffic flow forecasting.IET Intelligent Transport Systems17, 4 (Oct. 2022), 691–703. https://doi.org/10.1049/ itr2.12296

    work page 2022

  32. [32]

    Boris Medina-Salgado, Eddy Sánchez-Delacruz, Maria del Pilar Pozos Parra, and J. Sierra. 2022. Urban traffic flow prediction techniques: A review.Sustain. Comput. Informatics Syst.35 (2022), 100739. https://doi.org/10.1016/j.suscom.2022.100739

    work page doi:10.1016/j.suscom.2022.100739 2022

  33. [33]

    Zheyi Pan, Yuxuan Liang, Weifeng Wang, Yong Yu, Yu Zheng, and Junbo Zhang

    work page

  34. [34]

    Urban Traffic Prediction from Spatio-Temporal Data Using Deep Meta Learning. InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD 2019, Anchorage, AK, USA, August 4-8, 2019, Ankur Teredesai, Vipin Kumar, Ying Li, Rómer Rosales, Evimaria Terzi, and George Karypis (Eds.). ACM, Anchorage, AK, USA, 1720–1730....

    work page doi:10.1145/3292500.3330884 2019

  35. [35]

    Zezhi Shao, Zhao Zhang, Wei Wei, Fei Wang, Yongjun Xu, Xin Cao, and Chris- tian S. Jensen. 2022. Decoupled Dynamic Spatial-Temporal Graph Neural Net- work for Traffic Forecasting.Proc. VLDB Endow.15, 11 (2022), 2733–2746. https://doi.org/10.14778/3551793.3551827

    work page doi:10.14778/3551793.3551827 2022

  36. [36]

    Xuxiang Ta, Zihan Liu, Xiao Hu, Le Yu, Leilei Sun, and Bowen Du. 2022. Adaptive Spatio-temporal Graph Neural Network for traffic forecasting.Knowl. Based Syst. 242 (2022), 108199. https://doi.org/10.1016/j.knosys.2022.108199

    work page doi:10.1016/j.knosys.2022.108199 2022

  37. [37]

    Thanh Tran, Dan He, Jiwon Kim, and Mark Hickman. 2023. MSGNN: A Multi- structured Graph Neural Network model for real-time incident prediction in large traffic networks.Transportation research part C: emerging technologies156 (2023), 104354

    work page 2023

  38. [38]

    Yi Wang, Changfeng Jing, Shisuo Xu, and Tao Guo. 2022. Attention based spatiotemporal graph attention networks for traffic flow forecasting.Inf. Sci.607 (2022), 869–883. https://doi.org/10.1016/j.ins.2022.05.127

    work page doi:10.1016/j.ins.2022.05.127 2022

  39. [39]

    Salim, Xuan Song, and Ryosuke Shibasaki

    Zhaonan Wang, Renhe Jiang, Hao Xue, Flora D. Salim, Xuan Song, and Ryosuke Shibasaki. 2022. Event-Aware Multimodal Mobility Nowcasting. InThirty-Sixth 10 Incident-Guided Spatiotemporal Traffic Forecasting AAAI Conference on Artificial Intelligence, AAAI 2022, Thirty-Fourth Confer- ence on Innovative Applications of Artificial Intelligence, IAAI 2022, The ...

    work page doi:10.1609/aaai.v36i4.20342 2022

  40. [40]

    Cong Wu, Hui Ding, Zhongwang Fu, and Ning Sun. 2024. Multi-Scale Spatio- Temporal Attention Networks for Network-Scale Traffic Learning and Forecast- ing.Sensors24, 17 (2024), 5543. https://doi.org/10.3390/S24175543

    work page doi:10.3390/s24175543 2024

  41. [41]

    Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, and Chengqi Zhang. 2019. Graph WaveNet for Deep Spatial-Temporal Graph Modeling. InProceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, IJCAI 2019, Macao, China, August 10-16, 2019, Sarit Kraus (Ed.). ijcai.org, Macao, China, 1907–1913. https://doi.org/10.24963/IJCAI.2019/264

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

  42. [42]

    Qinge Xie, Tiancheng Guo, Yang Chen, Yu Xiao, Xin Wang, and Ben Y. Zhao

    work page

  43. [43]

    Deep Graph Convolutional Networks for Incident-Driven Traffic Speed Prediction. InCIKM ’20: The 29th ACM International Conference on Information and Knowledge Management, Virtual Event, Ireland, October 19-23, 2020, Math- ieu d’Aquin, Stefan Dietze, Claudia Hauff, Edward Curry, and Philippe Cudré- Mauroux (Eds.). ACM, 1665–1674. https://doi.org/10.1145/33...

    work page doi:10.1145/3340531.3411873 2020

  44. [44]

    Mingxing Xu, Wenrui Dai, Chunmiao Liu, Xing Gao, Weiyao Lin, Guo-Jun Qi, and Hongkai Xiong. 2021. Spatial-Temporal Transformer Networks for Traffic Flow Forecasting. arXiv:2001.02908 [eess.SP] https://arxiv.org/abs/2001.02908

    work page arXiv 2021

  45. [45]

    Zhijian Xu, Hao Wang, and Qiang Xu. 2025. Intervention-Aware Forecasting: Breaking Historical Limits from a System Perspective. arXiv:2405.13522 [cs.LG] https://arxiv.org/abs/2405.13522

    work page arXiv 2025

  46. [46]

    Li, Yuanfei Zang, and Yixuan Zhang

    Yaqin Ye, Yue Xiao, Yuxuan Zhou, S. Li, Yuanfei Zang, and Yixuan Zhang. 2023. Dynamic multi-graph neural network for traffic flow prediction incorporating traffic accidents.Expert Syst. Appl.234 (2023), 121101. https://doi.org/10.2139/ ssrn.4382825

    work page 2023

  47. [47]

    Bing Yu, Haoteng Yin, and Zhanxing Zhu. 2018. Spatio-Temporal Graph Convolu- tional Networks: A Deep Learning Framework for Traffic Forecasting. InProceed- ings of the Twenty-Seventh International Joint Conference on Artificial Intelligence, IJCAI-18. International Joint Conferences on Artificial Intelligence Organization, Stockholm, Sweden, 3634–3640. ht...

    work page doi:10.24963/ijcai.2018/505 2018

  48. [48]

    Zhuoning Yuan, Xun Zhou, and Tianbao Yang. 2018. Hetero-ConvLSTM: A Deep Learning Approach to Traffic Accident Prediction on Heterogeneous Spatio- Temporal Data. InProceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD 2018, London, UK, August 19-23, 2018, Yike Guo and Faisal Farooq (Eds.). ACM, London, UK, 9...

    work page doi:10.1145/3219819.3219922 2018

  49. [49]

    Weibin Zhang, Yinghao Yu, Yong Qi, Feng Shu, and Yinhai Wang. 2019. Short- term traffic flow prediction based on spatio-temporal analysis and CNN deep learning.Transportmetrica A: Transport Science15 (2019), 1688 – 1711. https: //doi.org/10.1080/23249935.2019.1637966

    work page doi:10.1080/23249935.2019.1637966 2019

  50. [50]

    Xu Zhang, Shunjie Wen, Liang Yan, Jiangfan Feng, and Ying Xia. 2024. A hybrid- convolution spatial–temporal recurrent network for traffic flow prediction.Com- put. J.67, 1 (2024), 236–252

    work page 2024

  51. [51]

    Ling Zhao, Yujiao Song, Chao Zhang, Yu Liu, Pu Wang, Tao Lin, Min Deng, and Haifeng Li. 2019. T-GCN: A temporal graph convolutional network for traffic prediction.IEEE transactions on intelligent transportation systems21, 9 (2019), 3848–3858

    work page 2019

  52. [52]

    Yi Zhou, Yihan Liu, Nianwen Ning, Li Wang, Zixing Zhang, Xiaozhi Gao, and Ning Lu. 2023. Integrating knowledge representation into traffic prediction: a spatial–temporal graph neural network with adaptive fusion features.Complex & Intelligent Systems10 (2023), 2883–2900. https://doi.org/10.1007/s40747-023- 01299-7

    work page doi:10.1007/s40747-023- 2023

  53. [53]

    Jiawei Zhu, Xing Han, Hanhan Deng, Chao Tao, Ling Zhao, Pu Wang, Tao Lin, and Haifeng Li. 2020. KST-GCN: A Knowledge-Driven Spatial-Temporal Graph Convolutional Network for Traffic Forecasting.IEEE Transactions on Intelligent Transportation Systems23 (2020), 15055–15065. https://doi.org/10.1109/TITS. 2021.3136287

    work page doi:10.1109/tits 2020

  54. [54]

    Jiawei Zhu, Qiongjie Wang, Chao Tao, Hanhan Deng, Ling Zhao, and Haifeng Li. 2020. AST-GCN: Attribute-Augmented Spatiotemporal Graph Convolutional Network for Traffic Forecasting.IEEE Access9 (2020), 35973–35983. https: //doi.org/10.1109/ACCESS.2021.3062114 11 Lixiang Fan, Bohao Li, Tao Zou, Junchen Ye, and Bowen Du A Appendix A.1 Data Processing Details ...

    work page doi:10.1109/access.2021.3062114 2020