Leveraging machine learning to analyze and forecast air quality trends in Kota City, India

Abstract

Air quality is a critical indicator of environmental health, directly impacting human well-being and ecological stability. Rapid urbanization and industrialization have recently exacerbated air pollution, necessitating robust monitoring and predictive frameworks. This study investigates air quality trends in Kota city of Rajasthan, India and using data from 2017 to 2023. Machine learning models, including linear regression (LR), random forest (RF), decision tree (DT), support vector regressor (SVR), and K-nearest neighbors (KNN), were employed to analyze predict air quality index (AQI) values based on key pollutants such as PM_2.5, PM₁₀, NO, NO₂, NOx, NH₃, SO₂, CO, Ozone, Benzene, Ethyl-Benzene, m & p-Xylene considering the effects of meteorological factors like relative humidity (RH), wind speed (WS), wind directions (WD), and barometric pressure (BP). Among these, the decision tree regressor shows almost perfect fit on the training set (R² score 0.9999) and excellent test performance (R² score 0.9991), suggesting a very accurate prediction model. However, it exhibits potential overfitting, limiting its generalization capabilities. On the other hand, the random forest regressor provides a balance of accuracy and robustness, achieving an R² score of 0.9831, making it the preferred model for reliable predictions. The study delves into pollutant contributions, evaluates model performances, and explores actionable insights for policymakers. By leveraging machine learning approaches, the study aims to provide a comprehensive framework for analyzing air quality trends and supporting decision-making processes.