Forecasting PM2.5 concentrations in Hanoi during the period 2022-2025 using a transformer model
DOI:
https://doi.org/10.5281/zenodo.19767482Keywords:
PM2.5, time-series forecasting, deep learning, Transformer, air qualityAbstract
Air pollution caused by fine particulate matter (PM2.5) has become a serious environmental problem in many large urban areas, especially in rapidly urbanizing regions. Accurate prediction of PM2.5 concentrations plays an important role in air quality management and the development of early warning systems for air pollution. This study evaluates the applicability of machine learning and deep learning approaches for forecasting PM2.5 concentrations using time-series data combined with meteorological variables. The dataset includes PM2.5 concentrations together with meteorological variables such as temperature, relative humidity, and wind speed collected in Hanoi. Data preprocessing steps include outlier detection using the Interquartile Range (IQR) method, data normalization using the Z-score approach, and the construction of time-series features. Several forecasting models were implemented and compared, including ARIMA, Random Forest, LSTM, GRU, and Transformer models. The experimental results show that deep learning models outperform traditional statistical approaches in PM2.5 prediction. Among the evaluated models, the Transformer model achieved the best performance with lower prediction errors and a better ability to capture temporal variations in PM2.5 concentrations. The results demonstrate the potential of deep learning techniques for air quality forecasting and provide a scientific basis for developing early warning systems for air pollution in large urban areas.
Downloads
References
[1] Naz F., Mccann C., Fahim M., Cao T.V., Hunter R., Viet N.T., Nguyen L.D., Duong T.Q. (2023), Comparative analysis of deep learning and statistical models for air pollutants prediction in urban areas, IEEE Access, 11, 64016–64025.
[2] Mahajan S., Chen L.J., Tsai T.C. (2018), Short-term PM2.5 forecasting using exponential smoothing method: A comparative analysis, Sensors, 18(10), 3223.
[3] Russell S., Norvig P. (2009), Artificial Intelligence: A Modern Approach, 3rd Edition, Prentice Hall, Upper Saddle River, NJ.
[4] Oprea M., Mihalache S.F., Popescu M. (2017), Computational intelligence-based PM2.5 air pollution forecasting, International Journal of Computers Communications & Control, 12(3), 365–380.
[5] Vaswani A., Shazeer N., Parmar N., Uszkoreit J., Jones L., Gomez A.N., Kaiser L., Polosukhin I. (2017), Attention Is All You Need, Proceedings of the 31st Conference on Neural Information Processing Systems (NeurIPS), Long Beach, USA.
[6] Nguyen M.H., Le Nguyen P., Nguyen K., Le V.A., Nguyen T.H., Ji Y. (2021), PM2.5 prediction using genetic algorithm-based feature selection and encoder–decoder model, IEEE Access, 9, 57338–57350.
[7] Hien P.D., Bac V.T., Tham H.C., Nhan D.D., Vinh L.D. (2002), Influence of meteorological conditions on PM2.5 and PM2.5–10 concentrations during the monsoon season in Hanoi, Vietnam, Atmospheric Environment, 36(21), 3473–3484.
[8] Zhou X., Cao Z., Ma Y., Wang L., Wu R., Wang W. (2016), Concentrations, correlations and chemical species of PM2.5 and PM10 based on published data in China: Potential implications for the revised particulate standard, Chemosphere, 144, 518–526.
[9] Zhao D., Chen H., Yu E., Luo T. (2019), PM2.5/PM10 ratios in eight economic regions and their relationship with meteorology in China, Advances in Meteorology, 2019, 1–15.
[10] Yan L., Wu Y., Yan L., Zhou M. (2018), Encoder–decoder model for forecast of PM2.5 concentration per hour, Proceedings of the 1st International Cognitive Cities Conference (IC3), 45–50.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
CC 4.0
