Prediction of Particulate Matter (PM2.5) Across India Using Machine Learning Methods

Air pollution is one of the global issues that has been a major concern in many countries due to its dangerous effect on human health as well as the environment. Air pollution in India is also a national threat because many big cities already have poor air quality, severely affecting human health and nature. Among the pollutants, particulate matter 2.5 (PM2.5) is the dominant factor causing cancers and other diseases in humans. Hence, predicting this type of air pollutant is vital for effective pollution control measures. In this paper, prediction models have been developed to forecast PM2.5 concentrations in India from air pollutant time series data taken at different locations in India using four machine learning algorithms. The algorithms used here are Support Vector Regression, Decision Tree Regression, K-Nearest Neighbors Regression, and Bayesian Ridge Regression. The data used in this study consists of hourly PM2.5 data from 7 areas in India for the year 2016. We have analyzed the univariate time series data by mapping it into a supervised problem. Experimental results show that the proposed models have effectively predicted PM2.5 concentrations, with Bayesian Ridge Regression-based model showing better performance than other models.