One of the world’s most critical challenges we face today is air pollution, generating numerous health issues and negatively impacting the environment. This study compares machine learning models like LSTM, ARIMA, SARIMAX, BVAR, VAR, GRU, and Prophet for 24 hours predictions of NO2 and PM2.5 concentrations. The project’s goal was to test the performance of machine learning algorithms and develop a methodology for identifying the most accurate one aiming to integrate it with urban air monitoring station networks for live prediction. The analysis of the forecasting results of the models used MAPE, RMSE, MSE, and MAE as evaluation methods for the algorithms’ accuracy. The LSTM algorithm resulted in the best-performing model, providing an accuracy of 81,63 % for NO2 and 76.79% for PM2.5.
Brianna Alexandra Stan