Modelling and prediction of air pollution using hybrid tree–LASSO approach

Abstract

This study employs hybrid tree–Least Absolute Shrinkage and Selection Operator approach to forecast pollutant concentrations (PM2.5, PM10, NO2, and CO) in Skopje, using data from 2018 to 2022, which includes meteorological variables and pollution measurements from three sensor nodes. Models were trained on pre-COVID-19 data and then tested on post- COVID-19 observations to assess the pandemic’s impact on air quality. Results show that models consistently overpredicted pollution levels during the pandemic, suggesting a positive effect of COVID-19 restrictions on air quality. Applications and research directions of the models in the context of metallurgy, mining, and mineral processing are discussed.