Air Quality Assessment and AQI Prediction Using Machine Learning: A Comparative Study of Major Indian Cities

Air pollution remains one of the most critical environmental and public health challenges in India, particularly in rapidly urbanizing regions. This study presents a comprehensive assessment of air quality by integrating a machine learning–based prediction framework with a detailed one-year comparative analysis of four major metropolitan cities—Pune, Mumbai, Bengaluru, and Delhi—for the year 2024. Publicly available air quality data from the Government Open Data (OGD) platform were processed and transformed into a structured dataset. The Air Quality Index (AQI) was computed using the official Central Pollution Control Board (CPCB) National Air Quality Index (NAQI) methodology. Two supervised machine learning models, Linear Regression and Random Forest, were applied to predict AQI values and evaluate model performance. The results indicate that Random Forest significantly outperforms Linear Regression, demonstrating a stronger ability to capture nonlinear relationships among pollutants. Feature importance analysis highlights PM2.5, CO, and PM10 as the most influential contributors to AQI. In addition, city-level statistical analysis reveals substantial regional disparities, with Delhi consistently experiencing higher pollution levels, while Bengaluru maintains relatively stable air quality. Seasonal trend analysis further shows that AQI peaks during winter due to unfavourable meteorological conditions and decreases during the monsoon season due to natural atmospheric cleansing. Comparative analysis across cities emphasizes the role of geography, emission sources, and climate in shaping air quality patterns. Overall, this study demonstrates the effectiveness of combining open environmental data, standardized AQI computation, and machine learning techniques to support transparent air quality assessment and data-driven environmental decision-making in India.