Prediction of Urban PM2.5 Concentrations Using Machine Learning–Based Multiple Linear Regression: A Case Study of Jaipur, India

Fine particulate matter (PM2.5) (aerodynamic dimension (≤2.5 µm)) poses a serious risk to public health and urban environmental quality, particularly in rapidly growing cities of developing countries. Reliable prediction of PM2.5 concentrations is therefore essential for effective air quality management and policy formulation. This study develops a machine learning–based multiple linear regression (MLR) model to predict PM₂.₅ concentrations using hourly air quality data from three monitoring stations in Jaipur, India, for the year 2019. Based on Pearson correlation analysis, PM10, NOx, and benzene were selected as input variables. Initial MLR models showed moderate predictive performance (R2 = 0.27–0.71), which was significantly improved through systematic data refinement techniques, including outlier removal, logarithmic transformation, and bootstrapping. The optimized models achieved R2 values ranging from 0.77 to 0.80 across the three sites, demonstrating strong agreement between predicted and observed PM₂.₅ concentrations. The results highlight the effectiveness of combining simple machine learning techniques with robust data preprocessing to enhance air quality prediction accuracy. The proposed approach provides a practical and transferable framework for PM2.5 forecasting in data-rich urban environments and can support decision-making for air pollution mitigation in Indian cities.