Prediction of Air Quality Index using Regression Models- A Case Study on Hyderabad, India

Abstract

Clean air is the minimum required to sustain healthy lives of human beings and other creatures on the earth.Unfortunately, the air is getting polluted with organic and inorganic substances entering the atmosphere and creating a health hazard to humans and the ecosystem. Air quality monitoring stations must be used to notify the public about the air quality in their environment in advance.This paper presents the performance evaluation of machine learning algorithms, linear regression, lasso regression, and random forest regression. Four air pollutants, Sulphur Dioxide (SO₂), Ammonia (NH₃), Particulate Matter (PM10), and Oxides of Nitrogen as (NO_x)levels for the years 2019 and 2020 in Hyderabad, India, were used to train the regression models to predict the AQI. Among the three, random forest regressor with hyperparameter tuning using random search cross validation got more accuracy with MAE = 1.32, MSE = 3.12, RMSE = 1.76, and correlation coefficient (R²) = 0.995.