Оценка качества графовых моделей машинного обучения на разнообразных промышленных данных

Хотя данные, которые можно естественным образом представить в виде графов, широко распространены в реальных приложениях самых разных отраслей, популярные наборы задач машинного обучения на графах (МО на графах) для предсказания свойств вершин охватывают удивительно узкий набор предметных областей, а графовые нейронные сети (GNN) нередко оцениваются лишь на нескольких академических сетях цитирований. Эта проблема особенно насущна в свете растущего интереса к разработке графовых фундаментальных моделей (GFM). Предполагается, что такие модели способны переносить знания на разнообразные графовые наборы данных из различных областей, однако предлагаемые GFM нередко оцениваются лишь на весьма ограниченном наборе данных из узкого круга приложений. Для восполнения этого пробела мы представляем GraphLand – 14 разнообразных графовых наборов данных для предсказания свойств вершин из широкого спектра различных промышленных приложений. GraphLand позволяет оценивать модели МО на графах на широком спектре графов с разнообразными размерами, структурными характеристиками и наборами признаков в едином формате. Помимо этого, GraphLand позволяет исследовать ранее слабо изученные научные вопросы, в частности как реалистичные временны́е сдвиги распределения в трансдуктивном и индуктивном режимах влияют на качество моделей МО на графах. Симулируя реалистичные промышленные условия, мы используем GraphLand для сравнения GNN с моделями градиентного бустинга над решающими деревьями (GBDT), популярными в промышленных приложениях, и показываем, что GBDT, снабжённые дополнительными признаками на основе графа, могут в ряде случаев быть весьма сильными базовыми моделями. Также мы оцениваем современные универсальные GFM и обнаруживаем, что они не достигают конкурентных результатов на предложенных наборах данных. Исходный код и наборы данных доступны в нашем репозитории на GitHub. Кроме того, наборы данных интегрированы в PyTorch Geometric.

1. Bechler-Speicher M. et al., “Position: Graph learning will lose relevance due to poor benchmarks,” in International conference on machine learning, 2025.

2. Palowitch J., Tsitsulin A., Mayer B., Perozzi B. GraphWorld: Fake graphs bring real insights for GNNs. In Proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining, 2022, pp. 3691-3701.

3. Maekawa S., Noda K., Sasaki Y., Onizuka M. Beyond real-world benchmark datasets: An empirical study of node classification with GNNs. Advances in Neural Information Processing Systems, 2022, vol. 35, pp. 5562-5574.

4. Errica F., Podda M., Bacciu D., Micheli A. A fair comparison of graph neural networks for graph classification. In International conference on learning representations, 2020.

5. Li Z., Cao Y., Shuai K., Miao Y., Hwang K. Rethinking the effectiveness of graph classification datasets in benchmarks for assessing GNNs. Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, 2024.

6. Coupette C., Wayland J., Simons E., Rieck B. No metric to rule them all: Toward principled evaluations of graph-learning datasets. In International conference on machine learning, 2025.

7. Li Y., Xiong M., Hooi B. GraphCleaner: Detecting mislabelled samples in popular graph learning benchmarks. In International conference on machine learning, 2023, pp. 20195-20209.

8. Platonov O., Kuznedelev D., Diskin M., Babenko A., Prokhorenkova L. A critical look at the evaluation of GNNs under heterophily: Are we really making progress? In International conference on learning representations, 2023.

9. Wang Y., Fan W., Wang S., Ma Y. Towards graph foundation models: A transferability perspective. 2025. DOI: 10.48550/arXiv.2503.09363.

10. Mao H. et al. Position: Graph foundation models are already here. In International conference on machine learning, 2024.

11. Yoon M., Wu Y., Palowitch J., Perozzi B., Salakhutdinov R. Graph generative model for benchmarking graph neural networks. In International conference on machine learning, 2023.

12. Friedman J. H. Greedy function approximation: A gradient boosting machine. Annals of Statistics, 2001, pp. 1189-1232.

13. Giles C.L., Bollacker K.D., Lawrence S. CiteSeer: An automatic citation indexing system. In Proceedings of the third ACM conference on digital libraries, 1998, pp. 89-98.

14. McCallum A. K., Nigam K., Rennie J., Seymore K. Automating the construction of internet portals with machine learning. Information Retrieval, 2000, vol. 3, no. 2, pp. 127-163.

15. Sen P., Namata G., Bilgic M., Getoor L., Galligher B., Eliassi-Rad T. Collective classification in network data. AI Magazine, 2008, vol. 29, no. 3, pp. 93-93.

16. Namata G., London B., Getoor L., Huang B. Query-driven active surveying for collective classification. In Proceedings of the workshop on mining and learning with graphs (MLG-2012), 2012.

17. Yang Z., Cohen W., Salakhudinov R. Revisiting semi-supervised learning with graph embeddings. In International conference on machine learning, 2016, pp. 40-48.

18. Kipf T.N., Welling M. Semi-supervised classification with Graph Convolutional Networks. In International conference on learning representations, 2017.

19. Shchur O., Mumme M., Bojchevski A., Günnemann S. Pitfalls of graph neural network evaluation, 2018. DOI: 10.48550/arXiv.1811.05868.

20. Hu W. et al. Open graph benchmark: Datasets for machine learning on graphs. Advances in Neural Information Processing Systems, 2020, vol. 33, pp. 22118-22133.

21. Huang Q., He H., Singh A., Lim S.-N., Benson A.R. Combining label propagation and simple models out-performs Graph Neural Networks. In International conference on learning representations, 2021.

22. Pei H., Wei B., Chang K.C.-C., Lei Y., Yang B. Geom-GCN: Geometric graph convolutional networks. In International conference on learning representations, 2020.

23. Liu H. et al. One for all: Towards training one graph model for all classification tasks. In International conference on learning representations, 2024.

24. Li Y., Wang P., Li Z., Yu J.X., and J. Li. ZeroG: Investigating cross-dataset zero-shot transferability in graphs. In Proceedings of the 30th ACM SIGKDD conference on knowledge discovery and data mining, 2024, pp. 1725-1735.

25. He Y., Sui Y., He X., Hooi B. UniGraph: Learning a unified cross-domain foundation model for text-attributed graphs. In Proceedings of the 31st ACM SIGKDD conference on knowledge discovery and data mining, 2025, pp. 448-459.

26. Hollmann N., Müller S., Eggensperger K., Hutter F. TabPFN: A transformer that solves small tabular classification problems in a second. In International conference on learning representations, 2023.

27. Hollmann N. et al. Accurate predictions on small data with a tabular foundation model. Nature, 2025, vol. 637, no. 8045, pp. 319-326.

28. Ye H.-J., Zhou Q., Yin H.-H., Zhan D.-C., Chao W.-L. Rethinking pre-training in tabular data: A neighborhood embedding perspective, 2023. DOI: 10.48550/arXiv.2311.00055.

29. Ye H.-J., Liu S.-Y., Chao W.-L. A closer look at TabPFN v2: Understanding its strengths and extending its capabilities, 2025. DOI: 10.48550/arXiv.2502.17361

30. Mueller A.C., Curino C.A., Ramakrishnan R. MotherNet: Fast training and inference via hyper-network transformers. In International conference on learning representations, 2025.

31. Ma J. et al. TabDPT: Scaling tabular foundation models on real data, 2024. DOI: 10.48550/arXiv.2410.18164.

32. Qu J., Holzmüller D., Varoquaux G., Le Morvan M. TabICL: A tabular foundation model for in-context learning on large data. In International conference on machine learning, 2025.

33. Wang S., Cukierski W. Click-through rate prediction. Kaggle, 2014. Available at: https://kaggle.com/competitions/avazu-ctr-prediction, accessed 29.08.2026.

34. García Ling C. et al. H&M personalized fashion recommendations. Kaggle, 2022. Available at: https://kaggle.com/competitions/h-and-m-personalized-fashion-recommendations, accessed 29.08.2026.

35. Takac L., Zabovsky M. Data analysis in public social networks. In International scientific conference and international workshop present day trends of innovations, 2012.

36. Rozemberczki B., Sarkar R. Twitch gamers: A dataset for evaluating proximity preserving and structural role-based node embeddings, 2021. DOI: 10.48550/arXiv.2101.03091

37. Likhobaba D., Pavlichenko N., Ustalov D. Toloker graph: Interaction of crowd annotators. Zenodo, 2023. DOI: 10.5281/zenodo.7620795.

38. Gui S., Li X., Wang L., Ji S. GOOD: A graph out-of-distribution benchmark. Advances in Neural Information Processing Systems, 2022, vol. 35, pp. 2059-2073.

39. Bazhenov G., Kuznedelev D., Malinin A., Babenko A., Prokhorenkova L. Evaluating robustness and uncertainty of graph models under structural distributional shifts. Advances in Neural Information Processing Systems, 2023, vol. 36, pp. 75567-75594.

40. Hamilton W., Ying Z., Leskovec J. Inductive representation learning on large graphs. Advances in Neural Information Processing Systems, 2017, vol. 30.

41. He K., Zhang X., Ren S., Sun J. Deep residual learning for image recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2016, pp. 770-778.

42. Ba J.L., Kiros J.R., Hinton G.E. Layer normalization, 2016. DOI: 10.48550/arXiv.1607.06450.

43. Gorishniy Y., Rubachev I., Khrulkov V., Babenko A. Revisiting deep learning models for tabular data. Advances in Neural Information Processing Systems, 2021, vol. 34, pp. 18932-18943.

44. Gorishniy Y., Rubachev I., Babenko A. On embeddings for numerical features in tabular deep learning. Advances in Neural Information Processing Systems, 2022, vol. 35, pp. 24991-25004.

45. S. Zhang, L. Yang, M. B. Mi, X. Zheng, and A. Yao, “Improving deep regression with ordinal entropy,” in International conference on learning representations, 2023.

46. Chen T., Guestrin C. 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.

47. Ke G. et al. LightGBM: A highly efficient gradient boosting decision tree. Advances in Neural Information Processing Systems, 2017, vol. 30.

48. Prokhorenkova L., Gusev G., Vorobev A., Dorogush A.V., Gulin A. CatBoost: Unbiased boosting with categorical features. Advances in Neural Information Processing Systems, 2018, vol. 31.

49. Veličković P., Cucurull G., Casanova A., Romero A., Lio P., Bengio Y. Graph Attention Networks. In International conference on learning representations, 2018.

50. Shi Y., Huang Z., Feng S., Zhong H., Wang W., Sun Y. Masked label prediction: Unified message passing model for semi-supervised classification. Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, 2021.

51. Yu X., Gong Z., Zhou C., Fang Y., Zhang H. SAMGPT: Text-free graph foundation model for multi-domain pre-training and cross-domain adaptation. In Proceedings of the ACM on web conference 2025, 2025, pp. 1142-1153.

52. Zhao H., Chen A., Sun X., Cheng H., Li J. All in one and one for all: A simple yet effective method towards cross-domain graph pretraining. In Proceedings of the 30th ACM SIGKDD conference on knowledge discovery and data mining, 2024, pp. 4443-4454.

53. L. Xia, B. Kao, and C. Huang, “OpenGraph: Towards open graph foundation models,” Proceedings of the Conference on Empirical Methods in Natural Language Processing, 2024.

54. Lachi D., Azabou M., Arora V., Dyer E. GraphFM: A scalable framework for multi-graph pretraining. 2024. DOI: 10.48550/arXiv.2407.11907.

55. Xia L., Huang C. AnyGraph: Graph foundation model in the wild, 2024. DOI: 10.48550/arXiv.2408.10700.

56. Finkelshtein B., Ceylan İ. İ., Bronstein M., Levie R. Equivariance everywhere all at once: A recipe for graph foundation models, 2025. DOI: 10.48550/arXiv.2506.14291.

57. Eremeev D., Bazhenov G., Platonov O., Babenko A., Prokhorenkova L. Turning tabular foundation models into graph foundation models, 2025. DOI: 10.48550/arXiv.2508.20906.

58. Eremeev D., Platonov O., Bazhenov G., Babenko A., Prokhorenkova L. GraphPFN: A prior-data fitted graph foundation model, 2025. DOI: 10.48550/arXiv.2509.21489.

59. Fey M., Lenssen J. E. Fast graph representation learning with PyTorch Geometric. Representation Learning on Graphs and Manifolds Workshop at The Seventh International Conference on Learning Representations, 2019.

60. Ivanov S., Prokhorenkova L. Boost then convolve: Gradient boosting meets graph neural networks. In International conference on learning representations, 2021.

61. Lim D. et al. Large scale learning on non-homophilous graphs: New benchmarks and strong simple methods. Advances in Neural Information Processing Systems, 2021, vol. 34, pp. 20887-20902.

62. Boccaletti S. et al. The structure and dynamics of multilayer networks. Physics Reports, 2014, vol. 544, no. 1, pp. 1-122.

63. Platonov O., Kuznedelev D., Babenko A., Prokhorenkova L. Characterizing graph datasets for node classification: Homophily–Heterophily dichotomy and beyond. In Advances in Neural Information Processing Systems, 2023, pp. 523-548.

64. Mironov M., Prokhorenkova L. Revisiting graph homophily measures. Learning on Graphs Conference, 2024.

65. Gilmer J., Schoenholz S.S., Riley P.F., Vinyals O., Dahl G.E. Neural message passing for quantum chemistry. In International conference on machine learning, 2017, pp. 1263-1272.

66. Hendrycks D., Gimpel K. Gaussian error linear units (GELUs). 2016. DOI: 10.48550/arXiv.1606.08415.

67. You Y., Chen T., Sui Y., Chen T., Wang Z., Shen Y. Graph contrastive learning with augmentations. Advances in Neural Information Processing Systems, 2020, vol. 33, pp. 5812-5823.

68. Xia J., Wu L., Chen J., Hu B., Li S.Z. SimGRACE: A simple framework for graph contrastive learning without data augmentation. In Proceedings of the ACM web conference, 2022, pp. 1070-1079.

69. Akiba T., Sano S., Yanase T., Ohta T., Koyama M., Optuna: A next-generation hyperparameter optimization framework. In Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery and data mining, 2019, pp. 2623-2631.

70. Srivastava N., Hinton G., Krizhevsky A., Sutskever I., Salakhutdinov R. Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 2014, vol. 15, no. 56, pp. 1929-1958.

71. Kingma D.P., Ba J. Adam: A method for stochastic optimization. In International conference on learning representations, 2015.

72. Paszke A. et al. PyTorch: An imperative style, high-performance deep learning library. Advances in Neural Information Processing Systems, 2019, vol. 32.

73. Wang M. et al. Deep Graph Library: A graph-centric, highly-performant package for graph neural networks. Representation Learning on Graphs and Manifolds Workshop at The Seventh International Conference on Learning Representations, 2019.

74. https://github.com/yandex-research/graphland

75. https://pytorch-geometric.readthedocs.io/en/latest/generated/torch_geometric.datasets.GraphLandDataset.html

76. https://zenodo.org/records/16895532

77. https://www.kaggle.com/datasets/bazhenovgleb/graphland