A Comprehensive Analysis of Machine Learning Methods for Air Pollution Forecasting

Abstract

Air pollution is a threat that all urban municipalities across the globe are trying to tackle. In India, air pollution is the fifth major cause of death, leading to around 2 million deaths per year, according to the World Health Organization. The ability to accurately predict air pollution levels in a region would give authorities the chance to take proactive measures, preventing the exposure of citizens to toxic pollutants and avoiding accidents and damage to property caused by smog. In this paper, we forecast the level of Particulate Matter 2.5 (PM2.5) for multiple urban cities in India using various machine learning algorithms built upon historical data. This data includes meteorological parameters such as temperature, wind speed, humidity, and the pollutant levels leading up to that given date/time. Based on the performance of the forecasting models, we perform a comparative analysis of each model and derive key insights.