Air Quality Prediction Based on a Deep Learning Hybrid Model

DOI: https://doi.org/10.62517/jbdc.202501422

Author(s)

Keke Han

Affiliation(s)

School of Mathematics and Statistics, North China University of Water Resources and Electric Power, Zhengzhou, Henan, China

Abstract

Accurate prediction of air quality is crucial for the management of public health and environmental. This paper uses Zhengzhou air quality data from 2018 to 2020 to predict AQI. Among them, six major pollutant concentrations, including sulfur dioxide, nitrogen dioxide, carbon monoxide, inhalable particles, ozone fine particles, are selected as key influencing points to predict the air quality index. Addressing the limitation of traditional models in fully capturing the complex spatiotemporal dependencies within air quality data, this study proposes a deep learning hybrid model integrating Convolutional Neural Networks (CNN), Long Short-Term Memory Networks (LSTM), and Attention Mechanisms. This model first utilizes CNN to effectively extract spatial correlations among pollutants. In order to ensure the long-term dependence of air quality, the air quality characteristics extracted from CNN will be input into LSTM. Finally, add the attention mechanism to the LSTM layer. In the attention mechanism, a higher weight can be automatically assigned to key information. In this process, the model's interpretation of important features can be enhanced. Experimental results show that the CNN-LSTM-Attention model achieves lower mean absolute error and root mean square error compared to other models, while exhibiting a higher R² value. This demonstrates that the prediction accuracy of the hybrid model is higher and more suitable for air quality index prediction.

Keywords

Air Quality; Convolutional Neural Network; Long Short-Term Memory Neural Network; Attention Mechanism

References

[1] Wang Siyuan, Ren Ying, Xia Bisheng, et al. PM2.5 Concentration Prediction in Yan'an City Based on Multiple Regression. Information Technology, 2024, (05): 15-21. [2] Talepour, N., Birgani, Y.T., Kelly, F.J. et al. Analyzing meteorological factors for forecasting PM10 and PM2.5 levels: a comparison between MLR and MLP models. Earth Sci Inform,2024,17: 5603–5623. [3] Zhong Lin, Yan Qisheng. Interpretable Machine Learning Modeling Method for Air Quality Prediction. Computer Simulation, 2023, 40(10): 511-518. [4] Chaturvedi, P. Air Quality Prediction System Using Machine Learning Models. Water Air Soil Pollut, 2024, 235(578). [5] Shanshan S, Qilong H. Multi-view multi-task spatiotemporal graph convolutional network for air quality prediction. The Science of the total environment,2023, 893164699-164699. [6] Wen Changjun, Chen Yangyang, He Yonghao, et al. Air Quality Index Prediction Based on Multi-Strategy SMA-BP Neural Network. Electronic Measurement Technology, 2023, 46(22): 78-86. [8] Nguyen, A.T., Pham, D.H., Oo, B. et al. Predicting air quality index using attention hybrid deep learning and quantum-inspired particle swarm optimization. Big Data, 2024, 11: 71. [9] Guo, Z., Jing, X., Ling, Y. et al. Optimized air quality management based on air quality index prediction and air pollutants identification in representative cities in China. Sci Rep, 2024, 14: 17923. [10] Pranolo A, Zhou X, Mao Y. A novel bifold-attention-LSTM for analyzing PM2.5 concentration-based multi-station data time series. International Journal of Data Science and Analytics,2024, (prepublish): 1-18. [11] Aysenur G, Selcuk A O, Atilla O. Air quality prediction using CNN+LSTM-based hybrid deep learning architecture. Environmental science and pollution research international, 2021, 29(8): 11920-11938.