Machine Learning Based Prediction Models of Air Quality Index for Mid-Sized Urban Cities

Abstract

Air quality deterioration in mid-sized urban cities poses growing environmental and public-health challenges because pollutant concentrations can vary substantially with traffic activity, urban development, and atmospheric conditions. Reliable Air Quality Index (AQI) prediction is therefore important for timely air-quality management and early warning. This study develops and evaluates Machine-Learning (ML)-based prediction models for AQI using air-quality observations from a high-traffic monitoring station in Lucknow, India. A three-month dataset covering January–March 2024 and comprising 88 observations was analyzed using “PM_2.5, PM₁₀, NO₂, SO₂, CO, and O3” as predictor variables and AQI as the target. Four regression approaches—"Linear Regression, LASSO Regression, Random Forest (RF), and Extreme Gradient Boosting (XGBoost)”—were developed using an 80:20 division of the dataset into training and testing subsets. Their predictive performance was evaluated using “Mean Absolute Error (MAE), Mean Squared Error (MSE), R², Adjusted R², Akaike Information Criterion (AIC), and Schwarz’s Bayesian Criterion (SBC)”. RF achieved the best performance, with an MAE of 7.626, MSE of 108.288, R² of 0.774, and Adjusted R² of 0.650. It also produced the lowest AIC (96.326) and SBC (101.668), demonstrating its superior overall predictive performance. XGBoost showed moderate performance, while LASSO performed weakest. Findings support ensemble learning for reliable AQI prediction.