Ensemble stacking of machine learning models for air quality prediction for Hyderabad city in India

Hyderabad, one of the rapidly developing cities in India, is facing with severe air pollution due to rapid urbanization, industrial operations, and climatic factors. To alleviate the negative impact on human health and the environment, accurate monitoring and forecasting of air quality are essential. This research utilized various machine learning models, such as XGBoost, LarsCV, Bayesian Ridge, AdaBoost, and ensemble stacking methods, to forecast the air quality index (AQI) using data from August 2016 to October 2023, which included 18 different air pollutants, including meteorological parameters. The ensemble stacking method showed excellent performance, attaining high training (R2 = 0.994) and validation (R2 = 0.999) accuracy with low error metrics (mean absolute error [MAE] = 0.496, mean square error [MSE] = 0.429, root-mean-square error [RMSE] = 0.655). These results highlight the efficacy of ensemble stacking for AQI prediction, providing crucial information for policymakers to formulate strategies to reduce air pollution's effects on public health and environmental sustainability.