Air Quality Index Prediction of Bangalore City Using Various Machine Learning Methods

Abstract

Air pollution is a severe problem in the present time whose implications are getting more dire with each passing day. In India alone, over 1.6 million people have died due to air pollution and its related factors in 2019. To clearly understand air pollution, we need to identify the primary factors contributing to it and to try and predict the conditions in future. Therefore, the assessment and forecasting of air quality is crucial. The Air Quality Index (AQI) is the standard measure of air quality and is calculated based on the average concentration of particular pollutants over a set time interval. Prediction models obtained using eight different Machine Learning methods are created to predict the AQI levels using the pollutants data obtained from India’s Central Control Room for Air Quality Management of Bangalore City over the past 10 years. The prediction models were for their accuracy using evaluation metrics such as Root Mean Squared Error, Mean Absolute Error and R2, and the best performing models were determined. From the metrics, it is found that Stacking Ensemble, XGBoost, Random Forest and Decision Trees perform the best, with Stacking Ensemble performing the best in terms of R2 and RMSE with R2 value of 0.991 and RMSE value of 6.353, whereas Random Forest has the lowest MAE value with 1.415. The models showing features importance also found that carbon monoxide (CO) and PM2.5 are the most important factors to the AQI.