pith. sign in

arxiv: 2412.00167 · v2 · submitted 2024-11-29 · 💻 cs.LG · cs.AI

Origin-Destination Demand Prediction: An Urban Radiation and Attraction Perspective

Xuan Ma , Zepeng Bao , Ming Zhong , Yuanyuan Zhu , Chenliang Li , Jiawei Jiang , Qing Li , Tieyun Qian This is my paper

Pith reviewed 2026-05-23 08:29 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords origin-destination demand predictionradiation capacityattraction capacitydeep learninghypergraphadversarial learningurban region functionsnominal attributes
0
0 comments X

The pith

A deep learning model embeds urban regions' radiation and attraction capacities to capture their functional differences and transformations for origin-destination demand prediction.

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

The paper seeks to improve origin-destination demand forecasts by bringing the physical notions of radiation and attraction capacities into neural networks so that models can respect how regions differ in function. Current data-driven methods ignore these functional distinctions while older physical approaches rely only on counts like population and overlook both nominal labels such as residential or industrial and the way one capacity turns into the other over a day. The authors therefore build a bilateral network that represents each capacity from region attributes, add a hypergraph step to generate parameters that encode capacity transformations inside the same region, and apply cluster-based adversarial training to model competition among regions that share the same attraction role.

Core claim

We generalize the physical radiation and attraction capacities into the deep learning framework with the extended capability to fulfil regions' functions, and present a new model that captures the relationships between two types of capacities through bilateral branches with attribute representations, hypergraph-based parameter generation for transformations, and cluster-based adversarial learning for competitions, yielding consistent gains on two datasets together with explainability from nominal attributes.

What carries the argument

Bilateral branch network that assigns separate paths to radiation and attraction capacities, each conditioned on region attribute representations, with hypergraph parameter generation handling intra-region capacity transformations and cluster adversarial training handling inter-region competition.

If this is right

  • Prediction accuracy improves consistently over state-of-the-art baselines on the two evaluated datasets.
  • The learned capacities provide direct explainability tied to each region's nominal attributes.
  • The same framework simultaneously encodes within-region capacity transformations and between-region competition for the same capacity type.
  • The approach extends physical capacity concepts so they become trainable components inside modern neural architectures.

Where Pith is reading between the lines

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

  • Planners could test how rezoning a district from residential to commercial would alter overall daily demand flows without new data collection.
  • The capacity representation might transfer to other spatial forecasting tasks where objects switch roles over time, such as logistics or energy load.
  • If the hypergraph step proves stable across time periods, retraining frequency could drop when city functions evolve gradually.

Load-bearing premise

Nominal region attributes plus the hypergraph and adversarial mechanisms suffice to represent functional differences and dynamic transformations between radiation and attraction capacities without introducing artifacts.

What would settle it

Run the model on a third city dataset and check whether prediction error stays lower than strong baselines or whether the learned capacities fail to match known nominal labels such as residential versus industrial districts.

Figures

Figures reproduced from arXiv: 2412.00167 by Chenliang Li, Jiawei Jiang, Ming Zhong, Qing Li, Tieyun Qian, Xuan Ma, Yuanyuan Zhu, Zepeng Bao.

Figure 1
Figure 1. Figure 1: (a) An illustration of the region partition in Manhattan, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The overall architecture of our proposed RACTC framework. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The attribute hypergraph and its incidence matrix. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: An example of two types of competition relationship. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of attention scores on New York Taxi. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of attention scores on Chicago Taxi. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Impacts of hyperparameters k1, k2, γ1, and γ2 in terms of RMSE on New York (-N) and Chicago (-C) dataset. TABLE VI: Results for computational cost analysis Model New York Taxi Chicago Taxi Para. Num Runtime Para. Num Runtime GEML 0.2M 3.39s 0.2M 4.83s MPGCN 0.2M 17.25s 0.2M 75.73s CMOD 0.3M 76.84s 0.3M 100.21s HMOD 1.5M 117.01s 1.6M 125.00s DyHSL 0.6M 11.81s 0.6M 12.24s ST-SSL 0.3M 6.36s 0.3M 6.17s RACTC-C… view at source ↗
read the original abstract

In recent years, origin-destination (OD) demand prediction has gained significant attention for its profound implications in urban development. Existing data-driven deep learning methods primarily focus on the spatial or temporal dependency between regions yet neglecting regions' fundamental functional difference. Though knowledge-driven physical methods have characterised regions' functions by their radiation and attraction capacities, these functions are defined on numerical factors like population without considering regions' intrinsic nominal attributes, e.g., a region is a residential or industrial district. Moreover, the complicated relationships between two types of capacities, e.g., the radiation capacity of a residential district in the morning will be transformed into the attraction capacity in the evening, are totally missing from physical methods. In this paper, we not only generalize the physical radiation and attraction capacities into the deep learning framework with the extended capability to fulfil regions' functions, but also present a new model that captures the relationships between two types of capacities. Specifically, we first model regions' radiation and attraction capacities using a bilateral branch network, each equipped with regions' attribute representations. We then describe the transformation relationship of different capacities of the same region using a hypergraph-based parameter generation method. We finally unveil the competition relationship of different regions with the same attraction capacity through cluster-based adversarial learning. Extensive experiments on two datasets demonstrate the consistent improvements of our method over the state-of-the-art baselines, as well as the good explainability of regions' functions using their nominal attributes.

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 / 3 minor

Summary. The paper proposes a deep learning model for origin-destination demand prediction that generalizes physical radiation and attraction capacities. It employs bilateral branch networks conditioned on nominal region attributes (e.g., residential/industrial), a hypergraph-based parameter generation method to model intra-region capacity transformations, and cluster-based adversarial learning to capture inter-region competition for the same capacity type. The central claims are that this architecture captures dynamic relationships missing from prior physical and data-driven methods, yields consistent improvements over SOTA baselines on two datasets, and provides explainability of region functions via nominal attributes.

Significance. If the experimental claims hold under rigorous controls, the work offers a principled way to inject physics-inspired functional capacities into neural OD models while adding mechanisms for their temporal transformation and competition. This could improve both predictive accuracy and interpretability in urban mobility applications. The bilateral conditioning on nominal attributes and the hypergraph/adversarial components are presented as load-bearing innovations; credit is due for attempting to address a clear gap between knowledge-driven and purely data-driven approaches.

major comments (2)
  1. [§4] §4 (Experiments): The abstract and summary assert 'consistent improvements' and 'good explainability' on two datasets, yet no information is provided on train/validation/test splits, number of random seeds, error bars or statistical significance tests, or whether hyperparameter selection was performed on held-out validation data. These controls are load-bearing for the superiority claim over baselines.
  2. [§3.2–3.3] §3.2–3.3 (Hypergraph parameter generation and cluster adversarial learning): The transformation and competition mechanisms introduce additional free parameters (hypergraph weights and adversarial cluster parameters). The manuscript should include an ablation that isolates their contribution versus simply increasing model capacity; without it, it remains unclear whether the claimed capture of dynamic relationships is due to the specific inductive biases or to extra expressivity.
minor comments (3)
  1. Notation: Define the radiation and attraction capacity vectors (e.g., r_i and a_i) explicitly at first use and ensure consistent symbol usage across equations and figures.
  2. Figure clarity: Ensure that any visualization of hypergraph edges or adversarial clusters is accompanied by a legend that maps colors/symbols to region attribute types.
  3. References: Add citations to recent hypergraph neural network and adversarial domain-adaptation works in urban computing to better situate the technical choices.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive evaluation and the detailed comments on experimental reporting and ablation studies. We address each point below and will revise the manuscript accordingly to strengthen the presentation.

read point-by-point responses

  1. Referee: [§4] §4 (Experiments): The abstract and summary assert 'consistent improvements' and 'good explainability' on two datasets, yet no information is provided on train/validation/test splits, number of random seeds, error bars or statistical significance tests, or whether hyperparameter selection was performed on held-out validation data. These controls are load-bearing for the superiority claim over baselines.

    Authors: We agree these details are necessary to substantiate the performance claims. In the revised manuscript, Section 4 will be expanded to specify the train/validation/test split ratios and selection procedure for both datasets, report all results as means over at least five random seeds with standard deviations and error bars, include statistical significance tests (e.g., paired t-tests) against baselines, and explicitly state that hyperparameters were selected via grid search on a held-out validation set. These additions will provide the required rigor without altering the core experimental outcomes. revision: yes

  2. Referee: [§3.2–3.3] §3.2–3.3 (Hypergraph parameter generation and cluster adversarial learning): The transformation and competition mechanisms introduce additional free parameters (hypergraph weights and adversarial cluster parameters). The manuscript should include an ablation that isolates their contribution versus simply increasing model capacity; without it, it remains unclear whether the claimed capture of dynamic relationships is due to the specific inductive biases or to extra expressivity.

    Authors: This is a fair and important concern. To isolate the effect of the proposed mechanisms, the revised manuscript will include new ablation experiments. We will compare the full model against capacity-matched variants in which the hypergraph parameter generation and cluster adversarial components are removed and replaced by equivalent increases in the width or depth of the bilateral branch networks (keeping total parameter count comparable). These results will clarify whether the observed gains arise from the specific inductive biases rather than added expressivity alone. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's core contribution is an architectural extension that maps physical radiation/attraction capacities into a bilateral-branch DL model augmented by hypergraph parameter generation and cluster adversarial learning. No equation or modeling step in the provided abstract or description reduces a claimed prediction to a fitted parameter or self-citation by construction. The derivation chain remains independent of its inputs; experimental validation on external datasets is asserted to support the claims.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on the domain assumption that nominal attributes encode functional differences and that the chosen neural mechanisms faithfully capture capacity transformations and competition; several learned parameters in the hypergraph and adversarial modules are fitted to data.

free parameters (2)
  • hypergraph parameters
    Parameters generated to model capacity transformations are learned from training data.
  • adversarial cluster parameters
    Cluster assignments and adversarial objectives introduce additional learned quantities.
axioms (2)
  • domain assumption Regions possess intrinsic nominal attributes that determine their radiation and attraction functions independently of numerical factors such as population.
    Invoked to justify the bilateral branch network design.
  • ad hoc to paper The transformation between radiation and attraction capacities for the same region and the competition between regions can be adequately represented by hypergraph parameter generation and cluster adversarial learning.
    Core modeling choice introduced in the paper.

pith-pipeline@v0.9.0 · 5806 in / 1378 out tokens · 25357 ms · 2026-05-23T08:29:32.265409+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

49 extracted references · 49 canonical work pages · 1 internal anchor

  1. [1]

    Multi-semantic path representation learning for travel time estimation,

    L. Han, B. Du, J. Lin, L. Sun, X. Li, and Y . Peng, “Multi-semantic path representation learning for travel time estimation,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 8, pp. 13 108–13 117, 2021

    work page 2021

  2. [2]

    Trafficpredict: Trajectory prediction for heterogeneous traffic-agents,

    Y . Ma, X. Zhu, S. Zhang, R. Yang, W. Wang, and D. Manocha, “Trafficpredict: Trajectory prediction for heterogeneous traffic-agents,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 33, no. 01, 2019, pp. 6120–6127

    work page 2019

  3. [3]

    Deep multi-view spatial-temporal network for taxi demand prediction,

    H. Yao, F. Wu, J. Ke, X. Tang, Y . Jia, S. Lu, P. Gong, J. Ye, and Z. Li, “Deep multi-view spatial-temporal network for taxi demand prediction,” in Proceedings of the AAAI conference on artificial intelligence, vol. 32, no. 1, 2018

    work page 2018

  4. [4]

    Origin- destination matrix prediction via graph convolution: a new perspective of passenger demand modeling,

    Y . Wang, H. Yin, H. Chen, T. Wo, J. Xu, and K. Zheng, “Origin- destination matrix prediction via graph convolution: a new perspective of passenger demand modeling,” in Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , 2019, pp. 1227–1235

    work page 2019

  5. [5]

    Adaptive graph convolutional recurrent network for traffic forecasting,

    L. Bai, L. Yao, C. Li, X. Wang, and C. Wang, “Adaptive graph convolutional recurrent network for traffic forecasting,” Advances in neural information processing systems, vol. 33, pp. 17 804–17 815, 2020

    work page 2020

  6. [6]

    Dynamic and multi- faceted spatio-temporal deep learning for traffic speed forecasting,

    L. Han, B. Du, L. Sun, Y . Fu, Y . Lv, and H. Xiong, “Dynamic and multi- faceted spatio-temporal deep learning for traffic speed forecasting,” in Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining , 2021, pp. 547–555

    work page 2021

  7. [7]

    St-gsp: Spatial-temporal global seman- tic representation learning for urban flow prediction,

    L. Zhao, M. Gao, and Z. Wang, “St-gsp: Spatial-temporal global seman- tic representation learning for urban flow prediction,” in Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining, 2022, pp. 1443–1451

    work page 2022

  8. [8]

    Road traffic forecasting: Recent advances and new challenges,

    I. Lana, J. Del Ser, M. Velez, and E. I. Vlahogianni, “Road traffic forecasting: Recent advances and new challenges,” IEEE Intelligent Transportation Systems Magazine, vol. 10, no. 2, pp. 93–109, 2018

    work page 2018

  9. [9]

    Continuous- time and multi-level graph representation learning for origin-destination demand prediction,

    L. Han, X. Ma, L. Sun, B. Du, Y . Fu, W. Lv, and H. Xiong, “Continuous- time and multi-level graph representation learning for origin-destination demand prediction,” in Proceedings of the 28th ACM SIGKDD Confer- ence on Knowledge Discovery and Data Mining , 2022, pp. 516–524

    work page 2022

  10. [10]

    Pattern expansion and consolidation on evolving graphs for continual traffic prediction,

    B. Wang, Y . Zhang, X. Wang, P. Wang, Z. Zhou, L. Bai, and Y . Wang, “Pattern expansion and consolidation on evolving graphs for continual traffic prediction,” inProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2023, pp. 2223–2232

    work page 2023

  11. [11]

    Bigst: Linear complexity spatio-temporal graph neural network for traffic forecasting on large-scale road networks,

    J. Han, W. Zhang, H. Liu, T. Tao, N. Tan, and H. Xiong, “Bigst: Linear complexity spatio-temporal graph neural network for traffic forecasting on large-scale road networks,” Proc. VLDB Endow. , vol. 17, no. 5, p. 1081–1090, 2024

    work page 2024

  12. [12]

    Fogs: First-order gradient supervision with learning-based graph for traffic flow forecasting,

    X. Rao, H. Wang, L. Zhang, J. Li, S. Shang, and P. Han, “Fogs: First-order gradient supervision with learning-based graph for traffic flow forecasting,” in Proceedings of International Joint Conference on Artificial Intelligence, IJCAI. ijcai. org, 2022

    work page 2022

  13. [13]

    Dstagnn: Dynamic spatial-temporal aware graph neural network for traffic flow forecasting,

    S. Lan, Y . Ma, W. Huang, W. Wang, H. Yang, and P. Li, “Dstagnn: Dynamic spatial-temporal aware graph neural network for traffic flow forecasting,” in International conference on machine learning . PMLR, 2022, pp. 11 906–11 917

    work page 2022

  14. [14]

    Graph multi-head convolution for spatio-temporal attention in origin destination tensor prediction,

    M. Bhanu, R. Kumar, S. Roy, J. Mendes-Moreira, and J. Chandra, “Graph multi-head convolution for spatio-temporal attention in origin destination tensor prediction,” in Pacific-Asia Conference on Knowledge Discovery and Data Mining . Springer, 2022, pp. 459–471

    work page 2022

  15. [15]

    Multi- attention 3d residual neural network for origin-destination crowd flow prediction,

    J. Ma, J. Chan, S. Rajasegarar, G. Ristanoski, and C. Leckie, “Multi- attention 3d residual neural network for origin-destination crowd flow prediction,” in IEEE International conference on data mining (ICDM) . IEEE, 2020, pp. 1160–1165

    work page 2020

  16. [16]

    Origin-destination traffic prediction based on hybrid spatio-temporal network,

    T. Chen, L. Nie, J. Pan, L. Tu, B. Zheng, and X. Bai, “Origin-destination traffic prediction based on hybrid spatio-temporal network,” in 2022 IEEE International Conference on Data Mining (ICDM) . IEEE, 2022, pp. 879–884

    work page 2022

  17. [17]

    Predicting origin-destination flow via multi-perspective graph convo- lutional network,

    H. Shi, Q. Yao, Q. Guo, Y . Li, L. Zhang, J. Ye, Y . Li, and Y . Liu, “Predicting origin-destination flow via multi-perspective graph convo- lutional network,” in 2020 IEEE 36th International conference on data engineering (ICDE). IEEE, 2020, pp. 1818–1821

    work page 2020

  18. [18]

    Taxi origin- destination demand prediction with contextualized spatial-temporal net- work,

    Z. Qiu, L. Liu, G. Li, Q. Wang, N. Xiao, and L. Lin, “Taxi origin- destination demand prediction with contextualized spatial-temporal net- work,” in 2019 IEEE International Conference on Multimedia and Expo (ICME). IEEE, 2019, pp. 760–765

    work page 2019

  19. [19]

    Spatiotemporal virtual graph convolution network for key origin- destination flow prediction in metro system,

    J. Yang, X. Han, T. Ye, Y . Tang, W. Feng, A. Wang, H. Zuo, Q. Zhang et al., “Spatiotemporal virtual graph convolution network for key origin- destination flow prediction in metro system,” Mathematical Problems in Engineering, 2022

    work page 2022

  20. [20]

    Origin-destination matrix prediction via hexagon-based generated graph,

    Y . Yang, S. Zhang, C. Zhang, and J. James, “Origin-destination matrix prediction via hexagon-based generated graph,” in 2021 IEEE Interna- tional Intelligent Transportation Systems Conference (ITSC) . IEEE, 2021, pp. 1399–1404

    work page 2021

  21. [21]

    Metro od matrix prediction based on multi-view passenger flow evolution trend modeling,

    F. Zheng, J. Zhao, J. Ye, X. Gao, K. Ye, and C. Xu, “Metro od matrix prediction based on multi-view passenger flow evolution trend modeling,” IEEE Transactions on Big Data , 2022

    work page 2022

  22. [22]

    Explainable origin-destination crowd flow interpolation via variational multi-modal recurrent graph auto-encoder,

    Q. Zhou, X. Lu, J. Gu, Z. Zheng, B. Jin, and J. Zhou, “Explainable origin-destination crowd flow interpolation via variational multi-modal recurrent graph auto-encoder,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 08, 2024, pp. 9422–9430

    work page 2024

  23. [23]

    Commuting patterns: the flow and jump model and supporting data,

    L. Varga, G. T ´oth, and Z. N ´eda, “Commuting patterns: the flow and jump model and supporting data,” EPJ Data Science , vol. 7, no. 1, pp. 1–13, 2018

    work page 2018

  24. [24]

    Intervening opportunities: a theory relating mobility and distance,

    S. A. Stouffer, “Intervening opportunities: a theory relating mobility and distance,” American sociological review, vol. 5, no. 6, pp. 845–867, 1940

    work page 1940

  25. [25]

    The p 1 p 2/d hypothesis: on the intercity movement of persons,

    G. K. Zipf, “The p 1 p 2/d hypothesis: on the intercity movement of persons,” American sociological review , vol. 11, no. 6, pp. 677–686, 1946

    work page 1946

  26. [26]

    Predicting commuter flows in spatial networks using a radiation model based on temporal ranges,

    Y . Ren, M. Ercsey-Ravasz, P. Wang, M. C. Gonz ´alez, and Z. Toroczkai, “Predicting commuter flows in spatial networks using a radiation model based on temporal ranges,” Nature communications , vol. 5, no. 1, p. 5347, 2014

    work page 2014

  27. [27]

    A universal model for mobility and migration patterns,

    F. Simini, M. C. Gonz ´alez, A. Maritan, and A.-L. Barab ´asi, “A universal model for mobility and migration patterns,” Nature, vol. 484, no. 7392, pp. 96–100, 2012

    work page 2012

  28. [28]

    Physics-infused machine learning for crowd simulation,

    G. Zhang, Z. Yu, D. Jin, and Y . Li, “Physics-infused machine learning for crowd simulation,” in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2022, pp. 2439– 2449

    work page 2022

  29. [29]

    Human mobility: Models and applications,

    H. Barbosa, M. Barthelemy, G. Ghoshal, C. R. James, M. Lenormand, T. Louail, R. Menezes, J. J. Ramasco, F. Simini, and M. Tomasini, “Human mobility: Models and applications,” Physics Reports, vol. 734, pp. 1–74, 2018

    work page 2018

  30. [30]

    Modelling the scaling properties of human mobility,

    C. Song, T. Koren, P. Wang, and A.-L. Barab ´asi, “Modelling the scaling properties of human mobility,” Nature physics, vol. 6, no. 10, pp. 818– 823, 2010

    work page 2010

  31. [31]

    A new set of spatial-interaction models: the theory of competing destinations,

    A. S. Fotheringham, “A new set of spatial-interaction models: the theory of competing destinations,” Environment and Planning A: Economy and Space, vol. 15, no. 1, pp. 15–36, 1983

    work page 1983

  32. [32]

    Web-scale k-means clustering,

    D. Sculley, “Web-scale k-means clustering,” in Proceedings of the 19th international conference on World wide web , 2010, pp. 1177–1178

    work page 2010

  33. [33]

    Unsupervised domain adaptation by back- propagation,

    Y . Ganin and V . Lempitsky, “Unsupervised domain adaptation by back- propagation,” in International conference on machine learning. PMLR, 2015, pp. 1180–1189

    work page 2015

  34. [34]

    Xgboost: A scalable tree boosting system,

    T. Chen and C. Guestrin, “Xgboost: A scalable tree boosting system,” in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining , 2016, pp. 785–794

    work page 2016

  35. [35]

    Dynamic graph learning based on hierarchical memory for origin-destination demand prediction,

    R. Zhang, L. Han, B. Liu, J. Zeng, and L. Sun, “Dynamic graph learning based on hierarchical memory for origin-destination demand prediction,” in Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, 2022, pp. 2383–2389

    work page 2022

  36. [36]

    Dynamic hypergraph structure learning for traffic flow forecasting

    Y . Zhao, X. Luo, W. Ju, C. Chen, X.-S. Hua, and M. Zhang, “Dynamic hypergraph structure learning for traffic flow forecasting.” ICDE, 2023

    work page 2023

  37. [37]

    Spatio-temporal self-supervised learning for traffic flow prediction,

    J. Ji, J. Wang, C. Huang, J. Wu, B. Xu, Z. Wu, J. Zhang, and Y . Zheng, “Spatio-temporal self-supervised learning for traffic flow prediction,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 4, 2023, pp. 4356–4364

    work page 2023

  38. [38]

    Deep multi-scale convolutional lstm network for travel demand and origin-destination predictions,

    K.-F. Chu, A. Y . Lam, and V . O. Li, “Deep multi-scale convolutional lstm network for travel demand and origin-destination predictions,”IEEE Transactions on Intelligent Transportation Systems , vol. 21, no. 8, pp. 3219–3232, 2019

    work page 2019

  39. [39]

    Contextualized spatial–temporal network for taxi origin-destination demand prediction,

    L. Liu, Z. Qiu, G. Li, Q. Wang, W. Ouyang, and L. Lin, “Contextualized spatial–temporal network for taxi origin-destination demand prediction,” IEEE Transactions on Intelligent Transportation Systems, vol. 20, no. 10, pp. 3875–3887, 2019

    work page 2019

  40. [40]

    Dynamic origin- destination prediction in urban rail systems: A multi-resolution spatio- temporal deep learning approach,

    P. Noursalehi, H. N. Koutsopoulos, and J. Zhao, “Dynamic origin- destination prediction in urban rail systems: A multi-resolution spatio- temporal deep learning approach,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 6, pp. 5106–5115, 2021

    work page 2021

  41. [41]

    Dynamic origin–destination matrix prediction with line graph neural networks and kalman filter,

    X. Xiong, K. Ozbay, L. Jin, and C. Feng, “Dynamic origin–destination matrix prediction with line graph neural networks and kalman filter,” Transportation Research Record, vol. 2674, no. 8, pp. 491–503, 2020

    work page 2020

  42. [42]

    Dneat: A novel dynamic node-edge attention network for origin-destination demand prediction,

    D. Zhang, F. Xiao, M. Shen, and S. Zhong, “Dneat: A novel dynamic node-edge attention network for origin-destination demand prediction,” Transportation Research Part C: Emerging Technologies , vol. 122, p. 102851, 2021

    work page 2021

  43. [43]

    Short-term origin- destination demand prediction in urban rail transit systems: A channel- wise attentive split-convolutional neural network method,

    J. Zhang, H. Che, F. Chen, W. Ma, and Z. He, “Short-term origin- destination demand prediction in urban rail transit systems: A channel- wise attentive split-convolutional neural network method,” Transporta- tion Research Part C: Emerging Technologies, vol. 124, p. 102928, 2021

    work page 2021

  44. [44]

    Gman: A graph multi-attention network for traffic prediction,

    C. Zheng, X. Fan, C. Wang, and J. Qi, “Gman: A graph multi-attention network for traffic prediction,” in Proceedings of the AAAI conference on artificial intelligence , vol. 34, no. 01, 2020, pp. 1234–1241

    work page 2020

  45. [45]

    Multi-stgcnet: A graph convolution based spatial-temporal framework for subway passenger flow forecast- ing,

    J. Ye, J. Zhao, K. Ye, and C. Xu, “Multi-stgcnet: A graph convolution based spatial-temporal framework for subway passenger flow forecast- ing,” in2020 International joint conference on neural networks (IJCNN). IEEE, 2020, pp. 1–8

    work page 2020

  46. [46]

    Graph-Based Deep Modeling and Real Time Forecasting of Sparse Spatio-Temporal Data

    B. Wang, X. Luo, F. Zhang, B. Yuan, A. L. Bertozzi, and P. J. Brantingham, “Graph-based deep modeling and real time forecasting of sparse spatio-temporal data,” arXiv preprint arXiv:1804.00684, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  47. [47]

    Temporal multi-graph convolutional network for traffic flow prediction,

    M. Lv, Z. Hong, L. Chen, T. Chen, T. Zhu, and S. Ji, “Temporal multi-graph convolutional network for traffic flow prediction,” IEEE Transactions on Intelligent Transportation Systems , vol. 22, no. 6, pp. 3337–3348, 2020

    work page 2020

  48. [48]

    Urban region representation learning with openstreetmap building footprints,

    Y . Li, W. Huang, G. Cong, H. Wang, and Z. Wang, “Urban region representation learning with openstreetmap building footprints,” in Pro- ceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2023, pp. 1363–1373

    work page 2023

  49. [49]

    Spa- tiotemporal multi-graph convolution network for ride-hailing demand forecasting,

    X. Geng, Y . Li, L. Wang, L. Zhang, Q. Yang, J. Ye, and Y . Liu, “Spa- tiotemporal multi-graph convolution network for ride-hailing demand forecasting,” in Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence, 2019

    work page 2019