Time Series Analysis and Forecasting of Air Quality in India

Abstract

This paper aims to analyze the air quality in India and the effects of seasons and COVID-19 on the concentration of pollutants in the air and thereby their effect on the air quality index (AQI). The analysis is performed on a full scale, taking into consideration different levels of granularities such as daily, weekly and monthly data. This study performs extensive preprocessing of the time series data for air quality to make it output the best results. The results evidenced that particulate matter i.e., PM 2.5 and PM 10 have the greatest impact on air quality. Analysis of the effect of change in seasons on the overall air quality has been carried out, along with the impact of the nationwide lockdown due to COVID-19, which led to a substantial improvement in the AQI levels. Furthermore, we also use the state-of-the-art forecasting algorithm Prophet to predict the monthly average air quality index and compare it with the actual recorded values, giving us a highly accurate prediction. We also performed a comparative analysis of AQI for the cities of Delhi and Bengaluru, having different seasons and climates, which results in valuable insights on to what extent the environmental factors affect the air quality measures of that location.