CausalPOI: Spatio-Temporal Graph-Based Causal Modeling for Cold-Start POI Check-in Forecasting

Gao Cong; Linyou Cai; Miao Xie; Siqiang Luo; Yi Li; Zhaoqi Zhang

arxiv: 2606.05413 · v1 · pith:3IWVQKD5new · submitted 2026-06-03 · 💻 cs.LG · cs.AI

CausalPOI: Spatio-Temporal Graph-Based Causal Modeling for Cold-Start POI Check-in Forecasting

Zhaoqi Zhang , Miao Xie , Yi Li , Linyou Cai , Siqiang Luo , Gao Cong This is my paper

Pith reviewed 2026-06-28 06:56 UTC · model grok-4.3

classification 💻 cs.LG cs.AI

keywords cold-start forecastingPOI check-in predictioncausal modelingspatio-temporal graphsfunctional interactionscounterfactual estimationurban computing

0 comments

The pith

CausalPOI forecasts check-in patterns for new POIs by building functional interaction graphs and simulating causal effects with aligned treatment and control graphs.

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

The paper introduces cold-start POI check-in forecasting as a distinct task and presents CausalPOI to solve it. Existing spatio-temporal graph methods rely on proximity and correlations, but CausalPOI instead constructs a Spatio-Temporal Functional Interaction Graph to encode semantic and spatial relationships among POIs. It then creates structurally aligned treatment and control graphs to distinguish factual from counterfactual scenarios. Experiments on SafeGraph data show gains over baselines in forecasting accuracy, interaction modeling, and causal estimation.

Core claim

By constructing a Spatio-Temporal Functional Interaction Graph to capture semantic and spatial relationships and building structurally aligned treatment and control graphs to simulate factual and counterfactual outcomes, CausalPOI enables accurate prediction of temporal check-in evolution for newly introduced POIs while estimating causal effects of urban interventions.

What carries the argument

Spatio-Temporal Functional Interaction Graph that encodes semantic and spatial dependencies between POIs, together with structurally aligned treatment and control graphs that separate factual from counterfactual scenarios for causal effect estimation.

If this is right

Forecasts become usable for evaluating the expected impact of opening a new POI before it exists.
Urban planners gain an interpretable way to compare alternative intervention locations based on estimated causal effects.
Models can separate functional dependencies from proximity-driven correlations when predicting activity at new sites.
Commercial decisions about site selection can incorporate counterfactual check-in trajectories rather than historical averages alone.

Where Pith is reading between the lines

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

The same graph-construction approach could be tested on other cold-start location tasks such as new transit stops or pop-up retail.
If the treatment-control alignment proves robust, the method could support online policy simulation where hypothetical POIs are inserted into live city graphs.
Extending the framework to multi-city transfer might reveal whether functional interaction patterns generalize beyond a single urban dataset.

Load-bearing premise

The functional interaction graph and the structurally aligned treatment and control graphs capture genuine causal dependencies between POIs rather than spurious correlations.

What would settle it

A controlled experiment on held-out cold-start POIs where removing the causal graph components causes performance to fall to the level of standard correlation-based spatio-temporal baselines.

Figures

Figures reproduced from arXiv: 2606.05413 by Gao Cong, Linyou Cai, Miao Xie, Siqiang Luo, Yi Li, Zhaoqi Zhang.

**Figure 2.** Figure 2: Parameter sensitivity analysis of CausalPOIs. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

**Figure 3.** Figure 3: Sanity check of estimated uplift of CausalPOI. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

read the original abstract

As urban environments continue to evolve rapidly, accurately modeling the dynamic behaviour of Points of Interest is essential for supporting data-driven urban planning and commercial decision-making. While recent advancements in spatio-temporal graph learning have improved POI forecasting, most methods rely on proximity-based graphs and correlation-driven modeling, which overlook the functional dependencies between POIs and fail to capture the causal effects of urban interventions. In this paper, we introduce a novel research problem -- cold-start POI check-in forecasting, which aims to predict the future check-in pattern of a newly introduced POI, by modeling its temporal evolution and functional interactions with nearby POIs in a structured urban spatial context. To address these challenges, we propose CausalPOI, a spatio-temporal graph-based causal representation learning framework. CausalPOI leverages Spatio-Temporal Functional Interaction Graph to model semantic and spatial relationships between POIs, and constructs structurally aligned treatment and control graphs to simulate factual and counterfactual scenarios. Extensive experiments on real-world SafeGraph datasets demonstrate that CausalPOI significantly outperforms state-of-the-art baselines across the board, validating its effectiveness in spatio-temporal forecasting, semantic interaction modeling, and causal effect estimation, providing a more interpretable and actionable foundation for urban intervention analysis. Source code is available at Github.

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.

Desk Editor's Note private letter to a colleague

The paper defines a fresh cold-start POI forecasting task and builds treatment/control graphs on top of a functional interaction graph, but the causal interpretation rests on an untested assumption that structural alignment isolates effects rather than correlations.

read the letter

The main thing here is a new problem statement—predicting check-ins for brand-new POIs by modeling their functional ties to existing ones—plus a graph construction that tries to simulate factual and counterfactual worlds via aligned treatment and control graphs. That framing moves beyond the usual proximity or correlation baselines in POI work, and the SafeGraph experiments plus released code give it some empirical grounding.

What the paper does cleanly is lay out the cold-start setting and show that adding the functional interaction graph plus the paired graphs improves forecasting numbers over standard spatio-temporal baselines. The idea of using structural alignment to stand in for interventions is a reasonable direction for urban data.

The soft spot is the causal part. The abstract and setup claim these graphs let them estimate causal effects for urban interventions, yet there is no identification argument, no explicit handling of confounders, and no sensitivity analysis shown. Without that, the performance gains could just come from richer correlation modeling, which undercuts the interpretability and intervention claims. The weakest assumption is that the Spatio-Temporal Functional Interaction Graph plus alignment actually separates causation from association.

This is worth sending to referees. The problem is new enough and the graph idea has enough structure that a serious review could tighten the causal claims or clarify what is really being measured. Readers working on spatio-temporal forecasting or urban analytics would get value from the problem definition and the empirical comparison, even if they end up treating the causal language as aspirational.

Referee Report

2 major / 1 minor

Summary. The paper introduces cold-start POI check-in forecasting as a new problem and proposes CausalPOI, a spatio-temporal graph-based causal representation learning framework. It constructs a Spatio-Temporal Functional Interaction Graph to capture semantic and spatial POI relationships and builds structurally aligned treatment/control graphs to simulate factual and counterfactual scenarios. Experiments on SafeGraph datasets claim significant outperformance over baselines in forecasting accuracy, semantic interaction modeling, and causal effect estimation, with source code released.

Significance. If the causal claims hold, the framework could advance interpretable urban planning by distinguishing causal effects of interventions from correlations in POI data. The release of source code supports reproducibility, a strength for the work.

major comments (2)

[Abstract] Abstract: the claim that structurally aligned treatment and control graphs 'simulate factual and counterfactual scenarios' for causal effect estimation lacks any identification strategy, confounder discussion, do-calculus justification, or sensitivity analysis. This is load-bearing for the central distinction between causal modeling and improved correlation capture.
[Abstract] Abstract and experiments description: no equations, data splits, ablation details, or error analysis are supplied, preventing verification that reported outperformance supports causal validity rather than better predictive modeling alone.

minor comments (1)

[Abstract] The GitHub link for source code is mentioned but not provided, which hinders immediate reproducibility assessment.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the presentation of our causal claims. We respond to each major comment below.

read point-by-point responses

Referee: [Abstract] Abstract: the claim that structurally aligned treatment and control graphs 'simulate factual and counterfactual scenarios' for causal effect estimation lacks any identification strategy, confounder discussion, do-calculus justification, or sensitivity analysis. This is load-bearing for the central distinction between causal modeling and improved correlation capture.

Authors: The abstract is space-constrained and therefore omits these details. The full manuscript (Section 3) constructs the Spatio-Temporal Functional Interaction Graph from observed semantic and spatial features that serve as observed confounders, then enforces structural alignment so that treatment and control graphs differ only by the presence of the new POI. This design approximates the counterfactual by holding the rest of the graph fixed. We agree that an explicit identification discussion, confounder enumeration, and sensitivity analysis would strengthen the causal framing. We will revise the abstract to reference the identification assumptions and add a short subsection on these points in the methodology. revision: yes
Referee: [Abstract] Abstract and experiments description: no equations, data splits, ablation details, or error analysis are supplied, preventing verification that reported outperformance supports causal validity rather than better predictive modeling alone.

Authors: Abstracts conventionally omit equations and experimental minutiae. The manuscript body supplies the model equations and graph-construction formalisms in Section 3, the train/validation/test splits and SafeGraph preprocessing in Section 4.1, ablation studies that isolate the contribution of the treatment-control alignment in Section 4.3, and error bars with statistical tests in Section 4.4. We will revise the abstract to point to these sections and will ensure the experiments narrative explicitly contrasts predictive gains against the causal-effect estimates. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The provided abstract and context introduce a new problem and framework (CausalPOI with Spatio-Temporal Functional Interaction Graph and aligned treatment/control graphs) but contain no equations, no fitted parameters renamed as predictions, and no self-citations that bear the load of the central claims. The derivation chain is not shown in sufficient detail to identify any reduction of outputs to inputs by construction. This is the expected honest non-finding when the manuscript text supplies no load-bearing steps that match the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Abstract-only review prevents identification of specific fitted parameters or background axioms; the framework appears to rest on the unstated premise that the constructed graphs encode true causal structure.

invented entities (1)

Spatio-Temporal Functional Interaction Graph no independent evidence
purpose: Model semantic and spatial relationships between POIs for causal simulation
Introduced as the core modeling structure in the proposed framework.

pith-pipeline@v0.9.1-grok · 5765 in / 1131 out tokens · 25995 ms · 2026-06-28T06:56:35.588331+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

36 extracted references · 10 canonical work pages · 7 internal anchors

[1]

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

2020
[2]

Shaked Brody, Uri Alon, and Eran Yahav. 2021. How attentive are graph attention networks?arXiv preprint arXiv:2105.14491(2021)

work page internal anchor Pith review Pith/arXiv arXiv 2021
[3]

Xu Chen, Junshan Wang, and Kunqing Xie. 2021. TrafficStream: A streaming traffic flow forecasting framework based on graph neural networks and continual learning.arXiv preprint arXiv:2106.06273(2021)

work page arXiv 2021
[4]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding.arXiv preprint arXiv:1810.04805(2018)

work page internal anchor Pith review Pith/arXiv arXiv 2018
[5]

Yanjie Fu, Pengyang Wang, Jiadi Du, Le Wu, and Xiaolin Li. 2019. Efficient region embedding with multi-view spatial networks: A perspective of locality- constrained spatial autocorrelations. InProceedings of the AAAI conference on artificial intelligence, Vol. 33. 906–913

2019
[6]

Shengnan Guo, Youfang Lin, Ning Feng, Chao Song, and Huaiyu Wan. 2019. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting. InProceedings of the AAAI conference on artificial intelligence, Vol. 33. 922–929

2019
[7]

Arash Hajisafi, Haowen Lin, Sina Shaham, Haoji Hu, Maria Despoina Siampou, Yao-Yi Chiang, and Cyrus Shahabi. 2023. Learning dynamic graphs from all con- textual information for accurate point-of-interest visit forecasting. InProceedings of the 31st ACM International Conference on Advances in Geographic Information Systems. 1–12

2023
[8]

Ming Jin, Shiyu Wang, Lintao Ma, Zhixuan Chu, James Y Zhang, Xiaoming Shi, Pin-Yu Chen, Yuxuan Liang, Yuan-Fang Li, Shirui Pan, et al . 2023. Time-llm: Time series forecasting by reprogramming large language models.arXiv preprint arXiv:2310.01728(2023)

work page internal anchor Pith review Pith/arXiv arXiv 2023
[9]

Shuangli Li, Jingbo Zhou, Tong Xu, Hao Liu, Xinjiang Lu, and Hui Xiong. 2020. Competitive analysis for points of interest. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining. 1265–1274

2020
[10]

Yaguang Li, Rose Yu, Cyrus Shahabi, and Yan Liu. 2017. Diffusion convolu- tional recurrent neural network: Data-driven traffic forecasting.arXiv preprint arXiv:1707.01926(2017)

work page internal anchor Pith review Pith/arXiv arXiv 2017
[11]

Ziyao Li, Shang-Ling Hsu, and Cyrus Shahabi. 2024. Forecasting Unseen Points of Interest Visits Using Context and Proximity Priors. In2024 IEEE International Conference on Big Data (BigData). IEEE, 5812–5818

2024
[12]

Chenxi Liu, Qianxiong Xu, Hao Miao, Sun Yang, Lingzheng Zhang, Cheng Long, Ziyue Li, and Rui Zhao. 2025. Timecma: Towards llm-empowered multivariate time series forecasting via cross-modality alignment. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 18780–18788

2025
[13]

Jia Liu, Tianrui Li, Shenggong Ji, Peng Xie, Shengdong Du, Fei Teng, and Junbo Zhang. 2021. Urban flow pattern mining based on multi-source heterogeneous data fusion and knowledge graph embedding.IEEE Transactions on Knowledge and Data Engineering35, 2 (2021), 2133–2146

2021
[14]

Ruoqi Liu, Changchang Yin, and Ping Zhang. 2020. Estimating individual treat- ment effects with time-varying confounders. In2020 IEEE international conference on data mining (ICDM). IEEE, 382–391

2020
[15]

Jing Ma, Mengting Wan, Longqi Yang, Jundong Li, Brent Hecht, and Jaime Teevan
[16]

InProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

Learning causal effects on hypergraphs. InProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 1202–1212
[17]

Christof Naumzik, Patrick Zoechbauer, and Stefan Feuerriegel. 2020. Mining points-of-interest for explaining urban phenomena: A scalable variational infer- ence approach. InProceedings of The Web Conference 2020. 2342–2353

2020
[18]

Aaron van den Oord, Yazhe Li, and Oriol Vinyals. 2018. Representation learning with contrastive predictive coding.arXiv preprint arXiv:1807.03748(2018)

work page internal anchor Pith review Pith/arXiv arXiv 2018
[19]

Uri Shalit, Fredrik D Johansson, and David Sontag. 2017. Estimating individual treatment effect: generalization bounds and algorithms. InInternational conference on machine learning. PMLR, 3076–3085

2017
[20]

Claudia Shi, David Blei, and Victor Veitch. 2019. Adapting neural networks for the estimation of treatment effects.Advances in neural information processing systems32 (2019)

2019
[21]

Daniel Tschernutter and Stefan Feuerriegel. 2021. A latent customer flow model for interpretable predictions of check-in counts. In2021 IEEE International Con- ference on Big Data (Big Data). IEEE, 529–539

2021
[22]

Lijing Wang, Aniruddha Adiga, Jiangzhuo Chen, Adam Sadilek, Srinivasan Venka- tramanan, and Madhav Marathe. 2022. Causalgnn: Causal-based graph neural networks for spatio-temporal epidemic forecasting. InProceedings of the AAAI conference on artificial intelligence, Vol. 36. 12191–12199

2022
[23]

Zi’ang Wang, Lei Chen, Yuanchang Jin, Pan Deng, Shuangshuang Pang, Junting Liu, and Yu Zhao. 2026. Knowledge Graph Guided Heterogeneity-Informed Diffusion Model for Spatio-Temporal Generation. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 40. 15915–15923

2026
[24]

Shangbin Wu, Xu Yan, Xiaoliang Fan, Shirui Pan, Shichao Zhu, Chuanpan Zheng, Ming Cheng, and Cheng Wang. 2022. Multi-graph fusion networks for urban region embedding.arXiv preprint arXiv:2201.09760(2022)

work page arXiv 2022
[25]

Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, Xiaojun Chang, and Chengqi Zhang. 2020. Connecting the dots: Multivariate time series forecasting with graph neural networks. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining. 753–763

2020
[26]

Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, and Chengqi Zhang. 2019. Graph wavenet for deep spatial-temporal graph modeling.arXiv preprint arXiv:1906.00121(2019)

work page internal anchor Pith review Pith/arXiv arXiv 2019
[27]

Guolei Yang, Ying Cai, and Chandan K Reddy. 2018. Recurrent spatio-temporal point process for check-in time prediction. InProceedings of the 27th ACM Inter- national Conference on Information and Knowledge Management. 2203–2211

2018
[28]

Bing Yu, Haoteng Yin, and Zhanxing Zhu. 2017. Spatio-temporal graph con- volutional networks: A deep learning framework for traffic forecasting.arXiv preprint arXiv:1709.04875(2017)

work page internal anchor Pith review Pith/arXiv arXiv 2017
[29]

Shilu Yuan, Dongfeng Li, Wei Liu, Xinxin Zhang, Meng Chen, Junjie Zhang, and Yongshun Gong. 2024. Fine-Grained Urban Flow Inference with Multi-scale Representation Learning.arXiv preprint arXiv:2406.09710(2024)

work page arXiv 2024
[30]

Qianru Zhang, Xinyi Gao, Haixin Wang, Siu Ming Yiu, and Hongzhi Yin. 2025. Efficient traffic prediction through spatio-temporal distillation. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 1093–1101

2025
[31]

Yunchao Zhang, Yanjie Fu, Pengyang Wang, Xiaolin Li, and Yu Zheng. 2019. Unifying inter-region autocorrelation and intra-region structures for spatial embedding via collective adversarial learning. InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 1700– 1708

2019
[32]

Yu Zhang, Yonghui Xu, Lizhen Cui, and Zhongmin Yan. 2023. Multi-view graph contrastive learning for urban region representation. In2023 International Joint Conference on Neural Networks (IJCNN). IEEE, 1–8

2023
[33]

Zhaoqi Zhang, Pasquale Balsebre, Siqiang Luo, Zhen Hai, and Jiangping Huang
[34]

InProceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)

StructAM: Enhancing Address Matching through Semantic Understanding of Structure-aware Information. InProceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). 15350–15361

2024
[35]

Zhaoqi Zhang, Miao Xie, Pasquale Balsebre, Weiming Huang, Siqiang Luo, and Gao Cong. 2026. UrbanMFM: Spatial Graph-Based Multiscale Foundation Models for Learning Generalized Urban Representation.IEEE Transactions on Knowledge and Data Engineering(2026)

2026
[36]

Yu Zhao, Pan Deng, Junting Liu, Xiaofeng Jia, and Jianwei Zhang. 2023. Gener- ative causal interpretation model for spatio-temporal representation learning. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 3537–3548. A Data Example We hereby provide data examples of POI and check-in sequence in Table 4. B Baselines •...

2023

[1] [1]

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

2020

[2] [2]

Shaked Brody, Uri Alon, and Eran Yahav. 2021. How attentive are graph attention networks?arXiv preprint arXiv:2105.14491(2021)

work page internal anchor Pith review Pith/arXiv arXiv 2021

[3] [3]

Xu Chen, Junshan Wang, and Kunqing Xie. 2021. TrafficStream: A streaming traffic flow forecasting framework based on graph neural networks and continual learning.arXiv preprint arXiv:2106.06273(2021)

work page arXiv 2021

[4] [4]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding.arXiv preprint arXiv:1810.04805(2018)

work page internal anchor Pith review Pith/arXiv arXiv 2018

[5] [5]

Yanjie Fu, Pengyang Wang, Jiadi Du, Le Wu, and Xiaolin Li. 2019. Efficient region embedding with multi-view spatial networks: A perspective of locality- constrained spatial autocorrelations. InProceedings of the AAAI conference on artificial intelligence, Vol. 33. 906–913

2019

[6] [6]

Shengnan Guo, Youfang Lin, Ning Feng, Chao Song, and Huaiyu Wan. 2019. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting. InProceedings of the AAAI conference on artificial intelligence, Vol. 33. 922–929

2019

[7] [7]

Arash Hajisafi, Haowen Lin, Sina Shaham, Haoji Hu, Maria Despoina Siampou, Yao-Yi Chiang, and Cyrus Shahabi. 2023. Learning dynamic graphs from all con- textual information for accurate point-of-interest visit forecasting. InProceedings of the 31st ACM International Conference on Advances in Geographic Information Systems. 1–12

2023

[8] [8]

Ming Jin, Shiyu Wang, Lintao Ma, Zhixuan Chu, James Y Zhang, Xiaoming Shi, Pin-Yu Chen, Yuxuan Liang, Yuan-Fang Li, Shirui Pan, et al . 2023. Time-llm: Time series forecasting by reprogramming large language models.arXiv preprint arXiv:2310.01728(2023)

work page internal anchor Pith review Pith/arXiv arXiv 2023

[9] [9]

Shuangli Li, Jingbo Zhou, Tong Xu, Hao Liu, Xinjiang Lu, and Hui Xiong. 2020. Competitive analysis for points of interest. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining. 1265–1274

2020

[10] [10]

Yaguang Li, Rose Yu, Cyrus Shahabi, and Yan Liu. 2017. Diffusion convolu- tional recurrent neural network: Data-driven traffic forecasting.arXiv preprint arXiv:1707.01926(2017)

work page internal anchor Pith review Pith/arXiv arXiv 2017

[11] [11]

Ziyao Li, Shang-Ling Hsu, and Cyrus Shahabi. 2024. Forecasting Unseen Points of Interest Visits Using Context and Proximity Priors. In2024 IEEE International Conference on Big Data (BigData). IEEE, 5812–5818

2024

[12] [12]

Chenxi Liu, Qianxiong Xu, Hao Miao, Sun Yang, Lingzheng Zhang, Cheng Long, Ziyue Li, and Rui Zhao. 2025. Timecma: Towards llm-empowered multivariate time series forecasting via cross-modality alignment. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 18780–18788

2025

[13] [13]

Jia Liu, Tianrui Li, Shenggong Ji, Peng Xie, Shengdong Du, Fei Teng, and Junbo Zhang. 2021. Urban flow pattern mining based on multi-source heterogeneous data fusion and knowledge graph embedding.IEEE Transactions on Knowledge and Data Engineering35, 2 (2021), 2133–2146

2021

[14] [14]

Ruoqi Liu, Changchang Yin, and Ping Zhang. 2020. Estimating individual treat- ment effects with time-varying confounders. In2020 IEEE international conference on data mining (ICDM). IEEE, 382–391

2020

[15] [15]

Jing Ma, Mengting Wan, Longqi Yang, Jundong Li, Brent Hecht, and Jaime Teevan

[16] [16]

InProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

Learning causal effects on hypergraphs. InProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 1202–1212

[17] [17]

Christof Naumzik, Patrick Zoechbauer, and Stefan Feuerriegel. 2020. Mining points-of-interest for explaining urban phenomena: A scalable variational infer- ence approach. InProceedings of The Web Conference 2020. 2342–2353

2020

[18] [18]

Aaron van den Oord, Yazhe Li, and Oriol Vinyals. 2018. Representation learning with contrastive predictive coding.arXiv preprint arXiv:1807.03748(2018)

work page internal anchor Pith review Pith/arXiv arXiv 2018

[19] [19]

Uri Shalit, Fredrik D Johansson, and David Sontag. 2017. Estimating individual treatment effect: generalization bounds and algorithms. InInternational conference on machine learning. PMLR, 3076–3085

2017

[20] [20]

Claudia Shi, David Blei, and Victor Veitch. 2019. Adapting neural networks for the estimation of treatment effects.Advances in neural information processing systems32 (2019)

2019

[21] [21]

Daniel Tschernutter and Stefan Feuerriegel. 2021. A latent customer flow model for interpretable predictions of check-in counts. In2021 IEEE International Con- ference on Big Data (Big Data). IEEE, 529–539

2021

[22] [22]

Lijing Wang, Aniruddha Adiga, Jiangzhuo Chen, Adam Sadilek, Srinivasan Venka- tramanan, and Madhav Marathe. 2022. Causalgnn: Causal-based graph neural networks for spatio-temporal epidemic forecasting. InProceedings of the AAAI conference on artificial intelligence, Vol. 36. 12191–12199

2022

[23] [23]

Zi’ang Wang, Lei Chen, Yuanchang Jin, Pan Deng, Shuangshuang Pang, Junting Liu, and Yu Zhao. 2026. Knowledge Graph Guided Heterogeneity-Informed Diffusion Model for Spatio-Temporal Generation. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 40. 15915–15923

2026

[24] [24]

Shangbin Wu, Xu Yan, Xiaoliang Fan, Shirui Pan, Shichao Zhu, Chuanpan Zheng, Ming Cheng, and Cheng Wang. 2022. Multi-graph fusion networks for urban region embedding.arXiv preprint arXiv:2201.09760(2022)

work page arXiv 2022

[25] [25]

Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, Xiaojun Chang, and Chengqi Zhang. 2020. Connecting the dots: Multivariate time series forecasting with graph neural networks. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining. 753–763

2020

[26] [26]

Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, and Chengqi Zhang. 2019. Graph wavenet for deep spatial-temporal graph modeling.arXiv preprint arXiv:1906.00121(2019)

work page internal anchor Pith review Pith/arXiv arXiv 2019

[27] [27]

Guolei Yang, Ying Cai, and Chandan K Reddy. 2018. Recurrent spatio-temporal point process for check-in time prediction. InProceedings of the 27th ACM Inter- national Conference on Information and Knowledge Management. 2203–2211

2018

[28] [28]

Bing Yu, Haoteng Yin, and Zhanxing Zhu. 2017. Spatio-temporal graph con- volutional networks: A deep learning framework for traffic forecasting.arXiv preprint arXiv:1709.04875(2017)

work page internal anchor Pith review Pith/arXiv arXiv 2017

[29] [29]

Shilu Yuan, Dongfeng Li, Wei Liu, Xinxin Zhang, Meng Chen, Junjie Zhang, and Yongshun Gong. 2024. Fine-Grained Urban Flow Inference with Multi-scale Representation Learning.arXiv preprint arXiv:2406.09710(2024)

work page arXiv 2024

[30] [30]

Qianru Zhang, Xinyi Gao, Haixin Wang, Siu Ming Yiu, and Hongzhi Yin. 2025. Efficient traffic prediction through spatio-temporal distillation. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 1093–1101

2025

[31] [31]

Yunchao Zhang, Yanjie Fu, Pengyang Wang, Xiaolin Li, and Yu Zheng. 2019. Unifying inter-region autocorrelation and intra-region structures for spatial embedding via collective adversarial learning. InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 1700– 1708

2019

[32] [32]

Yu Zhang, Yonghui Xu, Lizhen Cui, and Zhongmin Yan. 2023. Multi-view graph contrastive learning for urban region representation. In2023 International Joint Conference on Neural Networks (IJCNN). IEEE, 1–8

2023

[33] [33]

Zhaoqi Zhang, Pasquale Balsebre, Siqiang Luo, Zhen Hai, and Jiangping Huang

[34] [34]

InProceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)

StructAM: Enhancing Address Matching through Semantic Understanding of Structure-aware Information. InProceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). 15350–15361

2024

[35] [35]

Zhaoqi Zhang, Miao Xie, Pasquale Balsebre, Weiming Huang, Siqiang Luo, and Gao Cong. 2026. UrbanMFM: Spatial Graph-Based Multiscale Foundation Models for Learning Generalized Urban Representation.IEEE Transactions on Knowledge and Data Engineering(2026)

2026

[36] [36]

Yu Zhao, Pan Deng, Junting Liu, Xiaofeng Jia, and Jianwei Zhang. 2023. Gener- ative causal interpretation model for spatio-temporal representation learning. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 3537–3548. A Data Example We hereby provide data examples of POI and check-in sequence in Table 4. B Baselines •...

2023