Comparative Analysis of Prediction Models for Particulate Matter (PM2.5) Prediction

One of the key challenges with respect to the environment is the rise of concentration of pollutants in air, which needs to be addressed. For the prediction of pollutants, researchers have used a variety of cutting-edge machine learning prediction models. This study presents a comparative analysis of five alternative algorithms namely Multivariate LSTM, ET (Extra Tree), RF (Random Forest), LightGBM (Light Gradient Boosting Machine) and KNN (K-Nearest Neighbor) for prediction of pollutant PM2.5. In this work, the models are evaluated using the performance evaluation metrics such as MAE (Mean Absolute Error), RMSE (Root Mean Square Error), and MAPE (Mean Absolute Percent Error). The results show that all the models give less error but Multivariate LSTM model outperforms the other models. It is also observed that RH (Relative Humidity), PM10 and CO (Carbon Monoxide) are the important features in prediction of PM2.5 concentrations.