Prediction of Air Quality Index Using Machine Learning Techniques and the Study of Its Influence on the Health Hazards at Urban Environment

Abstract

AbstractAir quality across the globe is degrading at a faster rate due to the industrialization and urbanization which leaves no access to fresh air for breathing in major industrialized regions. Specifically, the industrial regions in and around India contribute a major part in the depletion of air quality in the south Asian region. Further, air pollution plays a vital role on the harmful diseases in the metropolises in India. The pollutant levels have been monitored in real-time to initiate the control measures to keep the air quality index (AQI) in the satisfactory level. Some of the major pollutants are particulate matter (PM2.5 and PM 10), sulfur dioxide (SO2), nitrogen dioxide (NO2), ozone (O3), carbon monoxide (CO), ammonia (NH3). With the evolution of technology, it is possible to predict the future air quality using various machine learning (ML) techniques. In the present work, the air quality data of major industrialized cities such as Delhi, Bangalore, Chennai, Ahmedabad, and Lucknow have been collected for the past five years (2015–2019) and predictive models are built. The prediction accuracy of traditional ML–models such as decision tree (DT), linear regression (LR), random forest (RF), and gradient boosting (GB) is compared with new hybrid models such as LR + DT and GB + DT. The predicted data are compared with the actual data of the subsequent years. The performance of the regression models is evaluated through mean absolute error (MAE), root mean square error (RMSE), and correlation coefficient (r). The hybrid models have proven efficiency while compared to the traditional methods in prediction of air quality of the subsequent years. Consequently, the health statistics (death rate) released by Institute of Health Metrics and Evaluation (IHME) is correlated with air pollution-based diseases. The presented ML-based work can be used in real-time to predict the future air pollution for the better planning and control of the air pollution-based diseases around the globe.KeywordsAir quality indexMachine learning (ML)PollutantsCardio vascular disease (CVD)Chronic respiratory disease (CRD)