Predictive Modelling for Air Quality: A Multi-Model Regression Approach

Air pollution poses a significant threat, necessitating reliable predictive models for effective air quality management. This study employs a multi-model regression approach for AQI prediction using techniques like Extra Trees Regressor, SVR, LightGBM, and CatBoost. Real-time datasets from CPCB, combined with meteorological and pollutant data, are pre-processed to handle missing values, outliers, and imbalances. Feature selection enhances accuracy by focusing on significant variables. The performance evaluation of the models revealed that the ExtraTrees Regressor was the best performer, achieving the highest R² score of 0.8660 and the lowest RMSE of 56.98. These results highlight ExtraTrees Regressor’s superior ability to explain AQI variance and minimize prediction errors, making it the most effective model for AQI forecasting in this study. Visualization techniques compare model outcomes, while distributed computing ensures real-time AQI forecasting, supporting sustainable urban air quality strategies.