Prognosis of air quality index and air pollution using machine learning techniques

Air pollution constitutes a significant challenge for both public health and environmental sustainability. Pollutants like PM, O₃, NO₂, SO₂, and CO cause serious health problems and ecological damage. This study utilizes five machine learning (ML) models, which are Gaussian Process Regression (GPR), Ensemble Regression (ER), Support Vector Machine (SVM), Regression Tree (RT), and Kernel Approximation Regression (KAR), which are developed and compared to predict the Air Quality Index (AQI). The publicly available historical air pollution dataset, collected from 1st January to 31st December 2022, was obtained from the online source titled 'A Real-time Dataset of Air Pollution Monitoring Generated Using IoT-Mendeley Data', developed by the Department of Software Engineering, Daffodil International University. While the dataset includes six pollutants (PM₁₀, PM_2.5, NO₂, SO₂, CO, and O₃), only three-PM_2.5, PM₁₀, and CO-were selected for AQI prediction based on their higher feature importance as determined using the Random Forest technique. To streamline the time and cost consumed in measuring and analyzing these pollutants, the five ML models were employed to predict the AQI using only these three essential features. The findings reveal that GPR, ER, SVM, and RT ML models exhibited higher accuracy levels, achieving over 96% AQI prediction, whereas the KAR model was less accurate, with an accuracy of 82.36%. The comparative analysis revealed that the GPR model outperformed the other ML models with a minimum Root Mean Square Error (RMSE) of 0.87 and 1.219 during the training and testing, respectively. The findings highlight the value of ML in enhancing air quality prediction and monitoring, offering accurate tools for hourly data analysis and potential real-time application. Such tools can assist in devising more efficient air pollution control strategies, contributing to improved public health and environmental sustainability.