Pith. sign in

REVIEW 4 major objections 5 minor 44 references

NAICS-Aware Graph Neural Networks for Large-Scale POI Co-visitation Prediction: A Multi-Modal Dataset and Methodology

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read NAICS-aware GraphSAGE lifts co-visit prediction R-squared to 0.625

desk verdict Useful dataset and plausible model, but the headline numbers are uninterpretable until the test-set sampling is specified. read the letter →

arxiv 2507.19697 v1 pith:7WEYTIOL submitted 2025-07-25 cs.LG

Yazeed Alrubyli , Omar Alomeir , Abrar Wafa , Diána Hidvégi , Hend Alrasheed , Mohsen Bahrami This is my paper · ORCID
classification cs.LG
keywords co-visitationpredictionGraphNeuralNetworksNAICSembeddingsedgeregressionPointsofInterestspatial-temporalmodelingsiteselectionlocation-basedservices
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that business category semantics, encoded as learned six-digit NAICS embeddings inside a GraphSAGE network, are the decisive signal for predicting how many distinct devices visit two brands within an hour. On a new dataset of 94.9 million co-visitation records covering 92,486 brands across 48 US states, the model reports a test $R^2$ of 0.625 versus 0.243 for the best baseline, STHGCN, and an NDCG@10 of 0.687 versus 0.523. If true, this means population-scale co-visitation can be predicted accurately enough to support store placement, urban planning, and location-based recommendations, and that industry-taxonomy embeddings matter more than geographic distance alone. The paper also claims the method scales to 4.2 billion potential brand pairs through state-wise decomposition and balanced sampling.

What carries the argument

The load-bearing object is the learnable NAICS embedding: each of 276 six-digit NAICS codes gets a 16-dimensional vector, concatenated with a popularity scalar to form 17-dimensional node features, then propagated through five GraphSAGE layers with hidden dimension 512, mean aggregation, and decreasing neighbor-sampling fanout [15, 10, 5]. Edge features add log-distance, cyclical month encoding, and popularity interactions, plus 38 state-level socioeconomic indicators, fused through a two-stage projection head before final regression. The NAICS embeddings are what allow the model to distinguish complementary business types, such as coffee shops near fast-casual dining, from incompatible ones, such as coffee shops near fine dining, which pure distance cannot capture.

What would settle it

On a held-out month, compute $R^2$, NDCG@10, and MAE over the full population of brand pairs, including the more than 99.9 percent zero co-visit pairs, rather than on a balanced sample, and check whether the 0.625 test $R^2$ survives the true near-zero distribution.

Watch

Extended reading notes

Core claim

The central claim is that population-level POI co-visitation can be cast as edge regression on a brand graph, and that a five-layer GraphSAGE with learnable embeddings for 276 six-digit NAICS codes outperforms spatial interaction models, matrix factorization, gradient boosting, GAT, GCN, and STHGCN. The reported test $R^2$ is 0.625, a 157 percent relative improvement over STHGCN's 0.243, with NDCG@10 rising from 0.523 to 0.687 and MRR from 0.596 to 0.743. The paper presents this as the first end-to-end framework that jointly optimizes business taxonomy embeddings with spatial, temporal, and socioeconomic features for nationwide co-visitation forecasting.

Load-bearing premise

The load-bearing premise is that the held-out test set used to compute the headline $R^2 = 0.625$ resembles the real deployment distribution of brand pairs rather than a balanced sample with half observed edges, because the paper never states the test set's positive-to-negative ratio.

Editorial extensions

If this is right

  • If the reported numbers hold, co-visitation models become accurate enough to pre-filter location candidates for retail site selection and to reduce location assessment errors by roughly half.
  • The 31 to 32 percent NDCG gain means the relative ordering of co-visit pairs is substantially better, which matters directly for itinerary and next-location recommendation.
  • The ablation study shows that removing NAICS embeddings drops $R^2$ from 0.625 to 0.258, so industry taxonomy is the dominant predictive signal in the model.
  • State-wise decomposition and a reported throughput of 25,000 edge predictions per second make nationwide deployment computationally feasible.
  • The model is most accurate on medium-frequency co-visits (10 to 1000 monthly visits) and under-predicts the heaviest-traffic edges, indicating a head-tail imbalance that remains unresolved.

Reading between the lines

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

  • An implication the paper leaves implicit is that the balanced positive/negative sampling used in training may inflate the headline $R^2$ if the test set is also balanced; a deployment evaluation on the full population of brand pairs, where more than 99.9 percent are zero, would give a more honest accuracy figure.
  • A testable extension beyond the paper is to apply the same NAICS-embedding approach to post-pandemic mobility data, which the paper explicitly says has shifted, to see whether learned category semantics transfer or require retraining.
  • The learned NAICS embeddings could plausibly be reused as pre-trained features for other spatial tasks such as foot-traffic forecasting or store-cannibalization prediction, since co-visitation flows are a natural input to those models.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes NAICS-aware GraphSAGE, a graph neural network for monthly brand-pair co-visitation prediction in the United States. The method represents brands with learnable 6-digit NAICS embeddings plus a popularity score, builds state-level co-visitation graphs, augments edges with spatial, temporal, and socioeconomic features, and trains with balanced positive/negative edge batches. The authors introduce a proprietary POI-Graph dataset and report that their model achieves test R-squared 0.625 versus 0.243 for the best baseline STHGCN, with NDCG@10 improving from 0.523 to 0.687 and MRR from 0.596 to 0.743. The paper includes a temporal train/validation/test split, confidence intervals over five runs, paired t-tests, an ablation study, scalability analysis, and a code release with a curated dataset subset.

Significance. If the reported gains hold under a properly specified evaluation, this would be a substantial empirical contribution to large-scale POI co-visitation prediction: the task is cast as sparse edge regression rather than next-POI recommendation, the dataset is far larger than typical benchmark mobility graphs, and the ablation study isolates the contribution of NAICS embeddings, socioeconomic features, depth, and width. The paper also reports confidence intervals over five runs, uses a temporal holdout, and releases code and a curated dataset subset, which are concrete strengths for reproducibility. However, the central performance claim is currently difficult to interpret because the held-out test edge sampling procedure is never specified, and the ranking-metric definition is missing; these issues must be resolved before the quantitative claims can be accepted.

major comments (4)
  1. [Sec. 3.5.1 / Table 1] The sampling procedure for the held-out test edges is not specified, even though Section 3.4.1 states that training batches are balanced 50/50 positive/negative and Section 4 states that over 99.9% of brand pairs have zero monthly co-visits. If the test set is also balanced, the reported R2=0.625 and NDCG@10=0.687 are computed on a distribution far richer in positive edges than the real deployment distribution, so the 157% improvement over STHGCN and the 32% NDCG gain could largely reflect evaluation-set construction rather than predictive skill on the true sparse population. Please state the test-set sampling explicitly, report the positive-edge fraction in the held-out set, and report metrics on an unbiased sample of all candidate pairs or disaggregated by positive/negative strata.
  2. [Sec. 4 vs Sec. 3.2] The dataset section says that 236,814 distinct brands covering 1,020 NAICS categories are retained, while Section 3.2 and the POI-Graph contribution bullet state that there are 276 unique 6-digit NAICS codes in the dataset. This discrepancy changes the size of the learned embedding table (the |C| in Algorithm 1) and the claimed taxonomy coverage. Please clarify whether the 1,020 categories were aggregated to 276 and, if so, describe the aggregation procedure.
  3. [Sec. 3.5.2 / Table 1] The computation of NDCG@10 and MRR is not defined for this edge-regression setting. These ranking metrics require a candidate set and a query structure (for example, all candidate pairs for a brand, all pairs in a state, or a sampled subset), and Table 1 reports them without specifying that structure. Without a precise ranking protocol, the ranking gains cannot be reproduced or compared across methods. Please define the candidate set, the query construction, and how ties and negative edges are handled.
  4. [Sec. 5.5 / Appendix B] The statement that graph baselines use 'the same feature engineering pipeline as our method' is ambiguous about whether GAT, GCN, and STHGCN also receive the learned NAICS embeddings. Since the ablation in Section A.1 reports that removing NAICS embeddings from the proposed model drops R2 from 0.625 to 0.258, the reader needs to know whether the baselines include this component in order to determine whether Table 1 compares architectures or feature sets. Please specify the exact input node and edge features used by each baseline.
minor comments (5)
  1. [Sec. 5.3] The sentence 'Results show consistent improvements across all folds (mean R2 improvement of 158.7' is incomplete; it is missing the closing parenthesis and the comparison value.
  2. [Sec. 5 vs Sec. 3.7 vs Appendix D] The hardware descriptions conflict: Section 5 states NVIDIA 3090 GPUs (24GB), Section 3.7 states NVIDIA A100 GPUs (40GB), and Appendix D states NVIDIA L40 GPUs (40GB). Table 3 reports peak memory of 28.1GB, which exceeds the 3090's capacity. Please reconcile these statements.
  3. [Figure 3 / Sec. 3.2 and Sec. 3.3.1] The figure caption describes an 8-dimensional NAICS embedding and an input projection from R^{8+d_v}, while the text specifies 16-dimensional NAICS embeddings and R17 node features; the caption's edge assembly dimension R^{1024+d_e} also differs from the two-stage prediction head in Section 3.3.1 (Wnode in R^{256×1024}, Wedge in R^{32×48}, final 288-dimensional fusion). Please align the figure with the text.
  4. [Sec. 3.7 / Appendix D] The reproducibility statements are inconsistent: Section 3.7 says the full training pipeline is released at a GitHub URL along with a curated subset, while Appendix D says the pipeline 'will be released alongside the POI-Graph dataset.' Please clarify what is currently available.
  5. [Page 20] The header 'Received 20 February 2007; revised 12 March 2009; accepted 5 June 2009' appears to be a template artifact and should be removed or updated.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the core result is an external supervised benchmark on a temporal holdout, and no load-bearing step reduces to fitted constants or to the authors' own prior claims.

full rationale

The paper's central claim is empirical: NAICS-aware GraphSAGE reports R2 = 0.625 versus 0.243 for STHGCN on a held-out month (March 2020) after training on earlier months. This is a standard supervised evaluation in which features (NAICS codes, distance, month encoding, socioeconomic indicators) are inputs and the target is a held-out co-visitation count; no parameter is fitted directly to the test labels, and no reported quantity is defined in terms of another reported quantity. The learned NAICS embeddings are trained end-to-end on the training interval, which is a normal feature-learning procedure, not circularity. The t-SNE inspection of learned embeddings (Section 5.4) is self-referential in style but not load-bearing: it is a post-hoc interpretability check, and the claimed semantic alignment is checked against the external NAICS taxonomy rather than used to justify the performance numbers. Baselines are external (Gravity, GeoMF, LightGBM, GAT, GCN, STHGCN), and there is no load-bearing self-citation chain; the citations are to external prior work. The acknowledged limitations in Section 6 (pre-pandemic focus, US-only data, head-tail imbalance) concern generalizability and error structure, not circularity. The evaluation-set sampling ambiguity noted in Sections 3.4.1 and 3.5 is a metric-validity risk, but it is not a case where a prediction reduces by construction to its inputs, so it does not raise the circularity score.

Assumptions & free parameters 10 free parameters · 8 assumptions · 0 invented entities

The central claim rests on the data definition (one-hour co-visits, brand aggregation, 5-trace threshold), on evaluation choices (balanced sampling, temporal split, pre-pandemic window), and on standard GNN training assumptions. No new physical entities are postulated. The main uncharged item is the evaluation protocol: the paper never specifies the label distribution of the held-out edge set, which determines whether R-squared 0.625 is a population-scale number or a balanced-sample number.

free parameters (10)
  • Co-visitation edge threshold = 5 device traces per month
    Hand-set in Section 3.1 to balance density and statistical significance; determines the graph and therefore all reported metrics.
  • Outlier cutoff for convenience stores = 40,000 monthly co-visits
    Ad hoc filter in Section 4; removes extreme observations and changes the target distribution.
  • NAICS embedding dimension = 16
    Selected by grid search over {8, 16, 32} in Section 5; the paper does not report sensitivity beyond the chosen value.
  • GraphSAGE hidden dimension = 512
    Grid-searched over {256, 512, 1024}; ablation shows R-squared drops from 0.625 to 0.402 at dimension 256.
  • Number of GraphSAGE layers = 5
    Grid-searched over {3, 5, 7}; ablation shows 3 layers yields R-squared 0.460.
  • Neighbor sampling fanout = [15, 10, 5]
    Chosen empirically in Section 3.3; the paper claims it reaches 94% of full-neighborhood performance at 75% lower memory.
  • Dropout rate = 0.2
    Grid-searched over {0.1, 0.2, 0.3} in Section 5.
  • Popularity score encoding = 3 levels (0, 1, 2)
    Hand-designed state- and industry-stratified quantile encoding in Section 3.2; ablation shows a small R-squared drop when removed.
  • Balanced positive/negative batch ratio = 1:1
    Fixed in Section 3.4.1; alters the training label distribution relative to the 99.9% zero-pair population.
  • Optimization hyperparameters = lr=1e-3, weight decay=1e-4, patience=20
    Set in Section 3.4.2; standard choices but selected by experiment and not independently justified.
assumptions (8)
  • domain assumption A co-visitation event is defined as two brands visited by the same device within a one-hour window.
    Dataset definition from the provider in Section 3; no sensitivity analysis on the window size.
  • domain assumption Multiple physical locations of the same brand can be aggregated into a single brand-level node.
    Section 3.1; this removes within-brand spatial structure that might matter for co-visitation.
  • domain assumption State-wise decomposition preserves the nationwide co-visitation signal.
    Section 3.1; cross-state co-visits are ignored by construction.
  • domain assumption The most frequent 6-digit NAICS code assigned to a brand is a faithful representation of its business.
    Section 4; brands with mixed activities are collapsed to one code.
  • domain assumption Pre-pandemic months from January 2018 to March 2020 are behaviorally stable enough to serve as the evaluation regime.
    Section 3.5.1 and Section 4; the paper explicitly limits applicability post-COVID.
  • domain assumption State-level socioeconomic indicators can be treated as shared edge features with a one-month lag.
    Section 3.2; the lag-1 alignment and state-level sharing are taken as given.
  • domain assumption Mean aggregation and balanced negative sampling are appropriate for sparse co-visitation graphs.
    Section 3.3 and 3.4.1; the justification is empirical and qualitative, not derived.
  • domain assumption MSE loss on raw counts is an appropriate training objective despite targets spanning several orders of magnitude.
    Section 3.4.1; the paper notes head-tail imbalance and discusses focal loss only as future work.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NAICS-Aware Graph Neural Networks for Large-Scale POI Co-visitation Prediction: A Multi-Modal Dataset and Methodology." pith.science (2026). https://pith.science/paper/7WEYTIOL

@misc{pith2026250719697,
  author       = {Pith},
  title        = {Pith review of: NAICS-Aware Graph Neural Networks for Large-Scale POI Co-visitation Prediction: A Multi-Modal Dataset and Methodology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7WEYTIOL}},
  note         = {Machine review of arXiv:2507.19697}
}
read the original abstract

Understanding where people go after visiting one business is crucial for urban planning, retail analytics, and location-based services. However, predicting these co-visitation patterns across millions of venues remains challenging due to extreme data sparsity and the complex interplay between spatial proximity and business relationships. Traditional approaches using only geographic distance fail to capture why coffee shops attract different customer flows than fine dining restaurants, even when co-located. We introduce NAICS-aware GraphSAGE, a novel graph neural network that integrates business taxonomy knowledge through learnable embeddings to predict population-scale co-visitation patterns. Our key insight is that business semantics, captured through detailed industry codes, provide crucial signals that pure spatial models cannot explain. The approach scales to massive datasets (4.2 billion potential venue pairs) through efficient state-wise decomposition while combining spatial, temporal, and socioeconomic features in an end-to-end framework. Evaluated on our POI-Graph dataset comprising 94.9 million co-visitation records across 92,486 brands and 48 US states, our method achieves significant improvements over state-of-the-art baselines: the R-squared value increases from 0.243 to 0.625 (a 157 percent improvement), with strong gains in ranking quality (32 percent improvement in NDCG at 10).

Figures

Figures reproduced from arXiv: 2507.19697 by the authors.

Figure 1
Figure 1. 𝑅 2 performance comparison across all methods showing distribution of results over 5 independent runs. 5.2 Performance Analysis by Categories We analyze model performance across states of different characteristics [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Predicted vs. actual co-visit counts (log-log scale). [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. NAICS-aware GraphSAGE enables scalable POI co-visitation prediction through systematic business taxonomy integration. The [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages

  1. [1]

    Helena Barbosa, Marc Barthelemy, Gourab Ghoshal, Charlotte R James, Maxime Lenormand, Thomas Louail, Ronaldo Menezes, José J Ramasco, Filippo Simini, and Marco Tomasini. 2023. Urban mobility analysis using graph neural networks and multimodal data fusion.Nature Communications 14, 1 (2023), 1–12

    work page 2023

  2. [2]

    Xavier Bouthillier, Pierre Delaunay, Mirko Bronzi, Assya Trofimov, Brennan Nichyporuk, Justin Szeto, Nazanin Mohammadi Sepahvand, Edward Raff, Kanika Madan, Vikram Voleti, et al. 2021. Accounting for variance in machine learning benchmarks. In Proceedings of Machine Learning and Systems, Vol. 3. 747–769

    work page 2021

  3. [3]

    Michael M Bronstein, Joan Bruna, Taco Cohen, and Petar Veličković. 2023. Geometric deep learning: Grids, groups, graphs, geodesics, and gauges. Nature Machine Intelligence 5, 3 (2023), 205–222

    work page 2023

  4. [4]

    Hongjun Chen, Jiyuan Wu, Binhui Wang, and Ming Li. 2023. STGFormer: Efficient spatial-temporal graph transformer for traffic forecasting. In Advances in Neural Information Processing Systems , Vol. 36. 18403–18417

    work page 2023

  5. [5]

    Shen Chen, Chaoyang Zhang, Minghui Yu, Hongyi Mao, Yongchao Xu, Renzhe Zhang, Yida Peng, Peng Shen, Yanghua Lin, and Zhi Zhang. 2022. PyTorch distributed: Experiences on accelerating data parallel training. Proceedings of the VLDB Endowment 15, 12, 3674–3687

    work page 2022

  6. [6]

    Janez Demšar. 2006. Statistical comparisons of classifiers over multiple data sets. Journal of Machine Learning Research 7 (2006), 1–30

    work page 2006

  7. [7]

    Thomas G Dietterich. 1998. Approximate statistical tests for comparing supervised classification learning algorithms. Neural Computation 10, 7 (1998), 1895–1923

    work page 1998

  8. [8]

    Yuxiao Dong, Jie Tang, Tiancheng Lou, and Bin Wang. 2023. NAICS-based business understanding through graph neural networks. In Proceedings of the 2023 Web Conference. 1845–1856

    work page 2023

Show all 44 references
  1. [9]

    Yao Fang, Lin Chen, and Jie Tang. 2022. Graph-based modeling of spatial interactions for retail analytics. In Proceedings of the 28th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . 213–223

    2022

  2. [10]

    Jie Feng, Yong Li, Chao Zhang, Funing Sun, Fangli Meng, Ang Guo, and Depeng Jin. 2018. DeepMove: Predicting human mobility with attentional recurrent networks. In Proceedings of the 2018 World Wide Web Conference . 1459–1468

    2018

  3. [11]

    Matthias Fey and Jan Eric Lenssen. 2023. PyTorch Geometric Temporal: Spatiotemporal signal processing with neural networks. Journal of Machine Learning Research 24, 38 (2023), 1–8

    2023

  4. [12]

    Huiji Gao, Jiliang Tang, Xia Hu, and Huan Liu. 2023. Location-based social networks: Users, check-ins and travel records. Comput. Surveys 55, 6 (2023), 1–44

    2023

  5. [13]

    Shengnan Guo, Youfang Lin, Ning Feng, Chao Song, and Huaiyu Wan. 2019. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 33. 922–929

    2019

  6. [14]

    William L Hamilton, Rex Ying, and Jure Leskovec. 2017. Inductive representation learning on large graphs. In Advances in Neural Information Processing Systems, Vol. 30. 1024–1034

    2017

  7. [15]

    Jiawei Han, Jian Pei, Micheline Kamber, et al. 2023. Business category representation learning through mobility patterns. IEEE Transactions on Knowledge and Data Engineering 35, 4 (2023), 3654–3667

    2023

  8. [16]

    Ming Jin, Yu Zheng, Yuan-Fang Li, Chen Gong, Chuan Zhou, and Shirui Pan. 2023. Spatio-temporal graph neural networks for multivariate time series forecasting. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 902–913

    2023

  9. [17]

    Defu Lian, Cong Zhao, Xing Xie, Guangzhong Sun, Enhong Chen, and Yong Rui. 2014. GeoMF: joint geographical modeling and matrix factorization for point-of-interest recommendation. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data min...

    2014

  10. [18]

    Yu Liu, Xian Li, and Song Gao. 2020. Urban computing for city-scale pandemic response: A survey. IEEE Transactions on Big Data 7, 3 (2020), 443–458

    2020

  11. [19]

    Zheng Liu, Zhenwei Huang, Shaojie Zou, and Junbo Shang. 2023. STGODE: Spatial-temporal graph ordinary differential equation networks for traffic forecasting. IEEE Transactions on Knowledge and Data Engineering 35, 12 (2023), 12444–12457

    2023

  12. [20]

    Massimiliano Luca, Gianni Barlacchi, Bruno Lepri, and Luca Pappalardo. 2023. Understanding mobility patterns with multimodal learning. Scientific Reports 13, 1 (2023), 1–15

    2023

  13. [21]

    Zezhi Shao, Zhao Zhang, Fei Wang, and Yongjun Xu. 2022. Traffic flow prediction with vehicle trajectory data. In Proceedings of the 28th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . 1550–1560

    2022

  14. [22]

    Chao Song, Youfang Lin, Shengnan Guo, and Huaiyu Wan. 2020. Spatial-temporal synchronous graph convolutional networks: A new framework for spatial-temporal network data forecasting. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 34. 914–921. 16 Alrubyli et al

    2020

  15. [23]

    Kangfei Wang, Yu Qian, Hanghang Tong, and Jian Li. 2023. Expressivity-preserving GNN simulation. In Advances in Neural Information Processing Systems, Vol. 36. 12707–12720

    2023

  16. [24]

    Minjie Wang, Lingfan Yu, Da Zheng, Quan Gan, Yu Gai, Zihao Ye, Mufei Li, Jinjing Zhou, Qi Huang, Chao Ma, Ziyue Huang, Qipeng Guo, Hao Zhang, Haibin Lin, Jinyang Yu, and Ion Stoica. 2021. Million-scale graph processing via graph neural networks. In Proceedings of Machine Learn...

    2021

  17. [25]

    Zhiwei Wang, Wei Wei, Gao Cong, Xian-Ling Li, Xian-Ling Mao, and Minghui Qiu. 2022. Hierarchical category-aware transformer for next basket recommendation. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval . 1613–1622

    2022

  18. [26]

    Ziqiang Wang, Yuchao Zhu, Chao Wang, Wenjie Ma, Bo Li, and Jie Yu. 2023. Adaptive Graph Representation Learning for Next POI Recommendation. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval . 393–402

    2023

  19. [27]

    Yanchuan Wei, Yu Zheng, and Qiang Yang. 2022. Contrastive trajectory similarity learning with dual-feature attention. In Proceedings of the 38th IEEE International Conference on Data Engineering . 2797–2810

    2022

  20. [28]

    Le Wu, Xiangnan Lei, Xiangnan He, Xiang Wang, Ajmal Mian, and Tat-Seng Chua. 2024. Spatial-temporal multi-head attention for next POI recommendation. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 2901–2911

    2024

  21. [29]

    Lianghao Wu, Lei Li, Yong Xu, Richang Hong, Meng Zhang, and Jing Wang. 2022. Knowledge-enhanced hierarchical graph transformer network for multi-behavior recommendation. 36 (2022), 4486–4493

    2022

  22. [30]

    Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and Philip S Yu. 2020. A comprehensive survey on graph neural networks. IEEE Transactions on Neural Networks and Learning Systems 32, 1 (2020), 4–24

    2020

  23. [31]

    Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, Xiaojie Chang, and Chengqi Zhang. 2019. Graph WaveNet for deep spatial-temporal graph modeling. (2019), 1907–1913

    2019

  24. [32]

    Ying Xu, Gao Cong, Ling Zhu, and Lili Cui. 2024. MMPOI: A Multi-Modal Content-Aware Framework for POI Recommendations. In Proceedings of the ACM Web Conference 2024. 3454–3463

    2024

  25. [33]

    Xiaotao Yan, Tao Song, Yina Jiao, Jing He, Jin Wang, Rui Li, and Wei Chu. 2023. Spatio-Temporal Hypergraph Learning for Next POI Recommendation. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval . 403–412

    2023

  26. [34]

    Sen Yang, Jun Liu, and Kai Zhao. 2022. GETNext: Trajectory Flow Map Enhanced Transformer for Next POI Recommendation. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval . 1144–1153

    2022

  27. [35]

    Yuanchi Yang, Zhihong Chen, Xiaoming Liu, and Qiang Wang. 2021. Category-aware graph neural networks for improving e-commerce review helpfulness prediction. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval . 2100–2104

    2021

  28. [36]

    Rex Ying, Ruining He, Kaifeng Chen, Pong Eksombatchai, William L Hamilton, and Jure Leskovec. 2018. Graph convolutional neural networks for web-scale recommender systems. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining . 974–983

    2018

  29. [37]

    Jiaxuan You, Jure Liu, Rex Ying, and Jure Leskovec. 2023. Graph structure learning with variational information bottleneck. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 37. 10713–10721

    2023

  30. [38]

    Muhan Zhang and Yixin Chen. 2018. Link prediction based on graph neural networks. In Advances in Neural Information Processing Systems , Vol. 31. 5165–5175

    2018

  31. [39]

    Shurui Zhang, Liang Yao, Aidong Sun, and Huanhuan Chen. 2024. Edge-level graph neural network explanation by leveraging causal inference. IEEE Transactions on Knowledge and Data Engineering 36, 8 (2024), 3943–3955

    2024

  32. [40]

    Yiding Zhang, Xiao Xiong, Linglong Kong, and Yanlin Li. 2021. Weighted graph convolutional network. In Proceedings of the 30th ACM International Conference on Information and Knowledge Management . 2651–2655

    2021

  33. [41]

    Zhiyuan Zhang, Peng Cui, and Wenwu Zhu. 2021. Graph neural networks: A review of methods and applications. AI Open 2 (2021), 57–81

    2021

  34. [42]

    Ling Zhao, Yujiao Song, Chao Zhang, Yu Liu, Pu Wang, Tao Lin, Min Deng, and Haifeng Li. 2019. Spatial-temporal graph convolutional networks: A deep learning approach for traffic forecasting. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 33. 7165–7172

    2019

  35. [43]

    Chuanpan Zheng, Xiaoliang Fan, Cheng Wang, and Jianzhong Qi. 2020. GMAN: A graph multi-attention network for traffic prediction. InProceedings of the AAAI Conference on Artificial Intelligence , Vol. 34. 1234–1241

    2020

  36. [44]

    Jing Zhou, Peng Cui, Fei Wang, and Xiang Zhang. 2023. EdgeGNN: Edge-centric Graph Neural Networks for Weighted Link Prediction. InProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 3120–3130. A Ablation Study To gain deeper insights into the...

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.