Air Quality Forecasting using decision trees algorithms

Forecasting air pollution is crucial as it not only affects the physical health of people but also provides guidance for pollution control. Particulate Matter with a diameter of less than
$2.5\mu m$
(PM
2.5
) is one of the major contributors to air pollution which can cause acute and chronic effects on human health. Forecasting PM
2.5
involves various meteorological factors as well as the influence of historical data. Therefore, the prediction of the surface PM
2.5
concentration is of great importance for the protection of human health. This study uses machine learning models, namely Decision Tree (DT), Random Forest (RF), and Gradient Boosting Regressor (GBR) to predict PM
2.5
hourly scale concentrations, using meteorological data and PM
2.5
concentrations from adjacent stations. The dataset was collected in the city of Beijing in China as a study region. The experiments have shown that the gradient boosting regressor model achieves higher predictive precision than the other models proposed for estimating hourly PM
2.5
concentrations with an
$R^{2}$
value varying between 0.9 and 0.97. This work offers a promising and affordable approach to robustly predict PM
2.5
concentrations.