Prediction and Analysis of Air Quality Index in South Asian countries using Deep Learning and SARIMAX

Abstract

Poor air quality can negatively affect the environment and human health. Air pollution contains unstable atoms that can attack healthy cells, damaging the body. Inhaling pollutant air affects the respiratory system, eyes, nose, throat, heart, and blood vessels. It deteriorates many breathing and lung illnesses, leading to cancer, or premature death. In South Asian countries, one of the main issues with suburban as well as urban areas is Air Quality Index (AQI). It is crucial to analyze and forecast current harmful emissions leading to poor air quality for sustainable urban development in the respective cities. Some of the gaps have been identified in existing methods, like the availability of the proper data, appropriate attributes, and other external factors like the population that can also be used to predict and forecast AQI. In this research, we introduced a hybrid Deep Learning (DL) model, including Bidirectional Long Short Term Memory (BD-LSTM), that uses SARIMAX for Time Series Forecasting to extract comprehensive features across spatiotemporal data exhibiting a 73.61% accuracy. Extensive model evaluations are done on data containing 10 different harmful pollutants that are responsible for creating a poor AQI along with the Population of the particular city which is taken as an added factor for AQI forecasting. Proposed model offers better accuracy of 6.7% than previous models