Predictive Model for Analyzing PM 2.5 Level Of Air

Abstract

It is often noticed that we tend to ignore or accept the harmful effects due to the acute levels of air pollution. This works aims at predicting the PM2.5 level of air in Delhi NCR region. Data mining & learning algorithms have proven to be useful techniques for predicting the various aspects of the air quality [1]. In this work we have worked on three learning models namely LSTM, Auto-regression and SVM to understand which of the models will be more suitable for prediction of PM2.5 level of air. We have performed pollutant forecasting on the data available at CPCB website from 01-jan-2016 upto 16-mar-2019. Our proposed model can predict the concentration of PM 2.5 level of air quality index by applying the three learning models on past data of air quality index dataset of Delhi NCR region. Motivation behind this work lies in the fact that alarming levels of PM2.5 has become a very serious problem particularly in Delhi NCR region. In the proposed work we have predicted the value of PM2.5 by using all other pollutants including Nitrogen Oxide, Sulphur Dioxide, Benzene, Ozone, Toluene, Carbon Monoxide. Our area of study is Anand Vihar of Delhi which is one of the pollution driven areas of Delhi. Our experimentation shows that Support Vector Regression model provides better prediction results in comparison to LSTM & auto-regression based learning.