Predictive Analytics For Air Quality Classification: A MultiCity Study In India

Abstract

Pollution is one of the major causes for human health diseases. In highly populated countries like India, to meet out the job requirements of the people major cities are expanded with tremendous industrial and population growth. Predictive analysis of air quality is essential as it would caution people about the air they inhale and proper remediations to reduce the pollution could be taken. In India, including four metropolitan cities Chennai, Kolkata, Mumbai, Delhi, Bangalore is also more urbanized. This urbanization effects leads to lot of industrial hubs in those regions causing more pollution especially water and air. In recent days the air pollution is tremendously increases and lot of breathing disorders are being reported. Hence the proposed work attempts to study the air pollution trends for these cities for a decade from 2013 to 2023. A prediction and classification model is developed to predict the quality of air and classify the level of pollution. The techniques implemented are Long Short-Term Memory (LSTM), XGBoost and SARIMAX regression for prediction and XGBoost classifier for classification of various levels. XGBoost regression and classifier model outperforms with 0.9519 R² and 0.9583 classification accuracy.