Multi-Step Ahead Ex-Ante Forecasting of Air Pollutants Using Machine Learning

In this study, a novel general multi-step ahead strategy is developed for forecasting time series of air pollutants. The values the predictors at future moments are gathered from official weather forecast sites as independent ex-ante data. They updated with new forecasted every day. Each sample used to build- separate single model that simultaneously predicts pollution levels. sought forecasts were estimated by averaging actual predictions models. was applied three pollutants—PM10, SO2, and NO2—in city Pernik, Bulgaria. Random forest (RF) arcing (Arc-x4) machine learning algorithms modeling. Although there many highly changing day-to-day predictors, proposed shows promising alternative most cases, root mean squared errors (RMSE) models (aRF aAR) last 10 horizons lower than those particular, PM10, aRF’s RMSE 13.1 vs. 13.8 micrograms per cubic meter model; NO2 model, aRF exhibits 21.5 23.8; aAR has 17.3 17.4; NO2, aAR’s 22.7 27.5, respectively. Fractional bias within same limits (?0.65, 0.7) all constructed