Short term air quality prediction and supervised Machine Learning analysis

Monitoring air quality pollutants form an important topic of atmospheric and environmental research due to the health effects caused by the pollutants present in the urban and suburban areas. The research evaluates forecasting models for predicting air pollution by exploiting various machine learning techniques. The goal is to determine the forecasting accuracy of the particulate matter 2.5 (PM2.5) concentration in air by various time-series models and further evaluate classification models based on its capability to segregate the air pollution type. In this research data is sourced from Indian government website. The time series analysis is accomplished using time series models such as Auto Regression Integrated Moving Averages (ARIMA), Generalized Autoregressive Conditional Heteroskedasticity (GARCH), TBATS model, ARIMA with multivariate regressions (ARIMAX) and Dynamic Harmonic Regression (DHR). ARIMAX performs best among all with the lowest error. For classification K nearest neighbor (KNN), Artificial Neural Network (ANN) and Ensemble model are used. Through out the research, it was found that use of ensemble model improves the performance of classifier