Comparative Evaluation of Machine and Deep Learning Models for Air Quality Index Prediction in Jaipur, India

Air pollution is an important environmental and public health challenge, therefore, accurate Air Quality Index (AQI) forecasting is important for timely mitigation. This study predicts AQI in Jaipur, India using seven machine learning and deep learning-based models i.e., Multiple Linear Regression (MLR), Random Forest (RF), Support Vector Regression (SVR), XGBoost, Adaboost, Artificial Neural Network (ANN), and Convolutional Neural Network (CNN). For this purpose, two years of hourly data from three monitoring sites were used, with preprocessing to address missing values and outliers. Key pollutant and meteorological variables were selected using Pearson's correlation coefficient vaules. Models were evaluated under three scenarios: pollutant parameters only (Case 1), meteorological parameters only (Case 2), and a combined dataset (Case 3). Performance was assessed using indices such as R2 and RMSE. Case 3 consistently produced the most accurate predictions, with Site 2 reflecting the best overall results. Among all models, XGBoost outperformed achieving R2 values of 0.77-0.95 and RMSE values of 16.96-20.98 across the three sites. The study demonstrates that XGBoost is a reliable approach for AQI forecasting and provides useful insights for air quality management and policymaking in rapidly urbanizing cities like Jaipur.