Seasonal Analysis of Air Pollutants and Ensemble Machine Learning–Based AQI Prediction

¹ Department of Engineering and Technology, College of Engineering, University of Technology and Applied Sciences, Muscat, Oman.
² Department of Information Systems, College of Economics, Management and Information Systems, University of Nizwa, Birkat Al Mouz, Nizwa 616, Oman
³ Department of Computing, Universiti Teknologi PETRONAS, Seri Iskandar 32610, Perak Darul Ridzuan, Malaysia
⁴ Higher Institute of Forensic Sciences, AI-Nahrain University, Baghdad, Iraq
⁵ Civil Engineering Department, Vallurupalli Nageswararao Vignana Jyothi Institute of Engineering and Technology, Hyderabad, Telangana 500090
⁶ Department of Civil and Environmental Engineering, College of Engineering and Architecture, University of Nizwa, Birkat Al Mouz, Nizwa 616, Oman, This email address is being protected from spambots. You need JavaScript enabled to view it.
⁷ Department of Civil Engineering, College of Engineering, Universiti Tenaga Nasional (UNITEN), Kajang 43000, Selangor Darul Ehsan, Malaysia

^* Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

This research investigated the concentration of major air pollutants that vary seasonally and developed a machine learning approach to estimate the Air Quality Index (AQI). Monthly analysis revealed distinct temporal trends, with particulate matter (PM₁₀ and PM₂.₅) showing consistently higher concentrations compared to gaseous pollutants. Both nitrogen oxides and ammonia have moderate fluctuation based on seasons, and sulfur dioxide and carbon monoxide show less fluctuation. Ozone is also found to be significantly variable, indicating photochemical effects. For predicting AQI, four machine learning models, namely Random Forest Regressor, Gradient Boosting Regressor, Support Vector Regression (SVR), and K-Nearest Neighbors (KNN), were employed for AQI prediction. The Gradient Boosting approach was the best at predicting AQI (R² = 0.9899, RMSE = 1.2328), followed by Random Forest (R² = 0.9877, RMSE = 1.3647). Analysis of residual plots confirms that the ensemble models were robust and had very less errors, but that SVR had greater errors in predictions. A feature importance analysis using the Random Forest approach shows that particulates (PM₁₀ and PM_2.5) are the two most important features of AQI, together accounting for the majority of the predicted values of AQI. These results indicate the significance of particulate matter to the determination of air quality and illustrate the use of ensemble learning methods to accurately forecast AQI to support data-driven decision-making for air quality management.