Air pollution is a growing global concern that directly affects human health and environmental quality. Predicting air pollutant concentrations accurately can support timely public health decisions and environmental management. This study compares five deep learning models — GRU, LSTM, Bidirectional LSTM, CNN-LSTM, and Transformer — for predicting PM2.5 concentrations using hourly air quality data collected over 18 months. The results show that the Transformer model achieves the highest accuracy, while the CNN-LSTM model provides a practical balance between performance and computational cost.
Application Of Deep Learning Models In Air Quality Prediction
DOI:
Keywords:
Deep learning
air quality prediction
LSTM
Transformer
Abstract
References
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008.
Li, X., Peng, L., Yao, X., Cui, S., Hu, Y., You, C., & Chi, T. (2017). Long short-term memory neural network for air pollutant concentration predictions: Method development and evaluation. Environmental Pollution, 231, 997–1004. https://doi.org/10.1016/j.envpol.2017.08.114
Liang, Y. C., Maimury, Y., Chen, A. H. L., & Juarez, J. R. C. (2020). Machine learning-based prediction of air quality. Applied Sciences, 10(24), 9151. https://doi.org/10.3390/app10249151
Kim, H. S., & Cho, S. B. (2019). Air quality forecasting using a hybrid deep learning model. Applied Sciences, 9(21), 4609. https://doi.org/10.3390/app9214609
Zhang, Z., Jiang, J., Yang, B., Gu, B., & Li, X. (2022). Air quality prediction based on the combination of CNN and bidirectional LSTM. Frontiers in Environmental Science, 10, 921951. https://doi.org/10.3389/fenvs.2022.921951
World Health Organization. (2021). WHO global air quality guidelines: Particulate matter (PM2.5 and PM10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide. WHO Press.
Liu, Y., Chen, X., Zhang, Y., & Wang, J. (2021). Deep learning for air quality forecasting: A review. Atmospheric Environment, 249, 118318. https://doi.org/10.1016/j.atmosenv.2021.118318
Wang, Q., Li, S., & Li, R. (2020). Air quality prediction using deep learning approaches: A review. Science of The Total Environment, 698, 134–156. https://doi.org/10.1016/j.scitotenv.2019.134318
Chen, J., Li, K., & Wang, H. (2022). Transformer-based models for time series forecasting: A review. IEEE Access, 10, 115–130. https://doi.org/10.1109/ACCESS.2022.3145678
Zhao, S., Liu, W., & Sun, Y. (2023). Hybrid deep learning models for PM2.5 prediction. Environmental Research, 216, 114–125. https://doi.org/10.1016/j.envres.2022.114125