COMPARATIVE ANALYSIS AND FEATURES OF GRAPH NEURAL NETWORKS FOR SPATIOTEMPORAL AIR POLLUTION FORECASTING

Authors

Z.M. Abubakerova Institute of Mechanical Engineering, Automation and Geomechanics of the National Academy of Sciences of the Kyrgyz Republic

Keywords:

CNN, GNN, air quality forecasting, spatiotemporal data, attention mechanism

Abstract

The article presents a review of modern spatiotemporal graph and hybrid neural networks used for air quality forecasting. The considered models aim to improve prediction accuracy by accounting for complex non-Euclidean spatiotemporal dependencies between monitoring stations and meteorological factors. Particular attention is paid to dynamic graph construction using mechanistic models, multi-level “region-station” architectures, adaptive attention mechanisms, and hybridization with signal decomposition. The analysis shows that these approaches outperform traditional methods, providing more accurate and stable long-term forecasting of pollutant concentrations, including rare extreme events.

References

Li P., Zhang T., Jin Y. A Spatio-Temporal Graph Convolutional Network for Air Quality Prediction // Sustainability. - 2023. - Vol. 15. No. 9. - Art. 7624. - DOI: 10.3390/su15097624.

Лыченко, Н. М. Модели прогноза качества воздуха на основе методов машинного обучения с учетом временных и пространственных зависимостей / Н. М. Лыченко // Проблемы автоматики и управления. – 2024. – № 3(51). – С. 136 –145. – EDN PJSEPL.

Devasekhar V., Natarajan P. Prediction of Air Quality and Pollution using Statistical Methods and Machine Learning Techniques // International Journal of Advanced Computer Science and Applications. - 2023. - Vol. 14. - No. 4.

Лыченко, Н. М. ПРИМЕНЕНИЕ LSTM-нейронных сетей ДЛЯ КЛАССИФИКАЦИИ индекса качества воздуха Г. БИШКЕК / Н. М. Лыченко, А. В. Сороковая // Проблемы автоматики и управления. – 2020. – № 1(38). – С. 70–80. – DOI 10.5281/zenodo.3904130. – EDN QQZNFI.

Li D., Yu H., Geng Y.-a., Li X., Li Q. DDGNet: A dual-stage dynamic spatio-temporal graph network for PM2.5 forecasting // В сб.: 2021 IEEE International Conference on Big Data (Big Data). - Piscataway, NJ: IEEE, 2021. - С. 1679-1685. - DOI: 10.1109/BigData52589.2021.9671941

Bai S., Kolter J. Z., Koltun V. An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling [Электронный ресурс]. - arXiv:1803.01271, 2018. - Режим доступа: https://arxiv.org/pdf/1803.01271 (дата обращения: 14.09.2025).

Bahdanau D., Cho K., Bengio Y. Neural Machine Translation by Jointly Learning to Align and Translate // В сб.: International Conference on Learning Representations (ICLR). – 2015. – Режим доступа: https://arxiv.org/abs/1409.0473 (дата обращения: 8.11.2025).

Veličković P., Cucurull G., Casanova A., Romero A., Liò P., Bengio Y. Graph Attention Networks // В сб.: International Conference on Learning Representations (ICLR). – 2018. – Режим доступа: https://arxiv.org/abs/1710.10903 (Дата обращения: 8.11.2025).

Iskandaryan D., Ramos F., Trilles S. Graph Neural Network for Air Quality Prediction: A Case Study in Madrid // IEEE Access. - 2023. - Vol. 11. - P. 2730-2742. - DOI: 10.1109/ACCESS.2023.3234214.

Liu H., Han Q., Sun H., Sheng J., Yang Z. Spatiotemporal adaptive attention graph convolution network for city-level air quality prediction // Scientific Reports. - 2023. - Vol. 13. - Art. 13335. - DOI: 10.1038/s41598-023-39286-0.

Kim D.-Y., Jin D.-Y., Suk H.-I. Spatiotemporal graph neural networks for predicting mid-to-long-term PM2.5 concentrations // Journal of Cleaner Production. - 2023. - Vol. 425. - Art. 138880. - DOI: 10.1016/j.jclepro.2023.138880.

Liao H., Yuan L., Wu M. A spatiotemporal graph neural network with multi granularity for air quality prediction [Электронный ресурс]. - Режим доступа: paper2.pdf (дата обращения: 17.09.2025).

Wang X., Zhang S., Chen Y., He L., Ren Y., Zhang Z., Li J., Zhang S. Air quality forecasting using a spatiotemporal hybrid deep learning model based on VMD-GAT-BiLSTM // Scientific Reports. - 2024. - Vol. 14. - Art. 17841. - DOI: 10.1038/s41598-024-68874-x.