The Air Quality Index (AQI)Air quality index is an important indicator of short-term air pollution health impacts. With increasing air pollution in Indian cities, precise AQI forecasting is necessary for environmental and public healthMonitoring monitoringHealth monitoring. This paper suggests a machine learningMachine learning-based approach to forecast AQI from real-world data from four Indian cities: Delhi, Amritsar, Chandigarh, and Visakhapatnam. These cities were selected to reflect various geographic, climatic, and pollution profiles. Support Vector Regression (SVRSupport Vector Regression (SVR)), Random ForestRandom forest Regression (RFR), and XGBoostXGBoost were the three regression models used. For correcting class imbalance of AQI categories, the Synthetic Minority Over-sampling Technique (SMOTE)Synthetic Minority Over-sampling Technique (SMOTE) technique was employed. On the raw datasetDataset, RFR performed best with the smallest RMSE and highest accuracy for most of the cities: Delhi (RMSE: 21.90, Accuracy: 95.32%), Amritsar (19.00, 94.30%), Chandigarh (18.74, 95.24%), and Visakhapatnam (16.64, 96.55%). XGBoostXGBoost provided the highest accuracy in Amritsar (96.28%) and Visakhapatnam (96.70%). On the SMOTESynthetic Minority Over-sampling Technique (SMOTE)-balanced datasetDataset, all models had improved performance, and RFR led consistently. This research emphasizes that the incorporation of SMOTESynthetic Minority Over-sampling Technique (SMOTE) with machine learningMachine learning enhances the forecast of AQIAir quality index and facilitates efficient environmental planning and public health planning.