Validation of linear, nonlinear, and hybrid models for predicting particulate matter concentration in Tehran, Iran

摘要: Information on particulate matter forecast is significant as it allows residents to manage its undesirable effects. For the purpose of predicting PM10 concentration in air Tehran, various models were used, including (i) a linear model (multiple liner regression, MLR), (ii) two hybrid (Adaptive Neuro-Fuzzy Inference System, ANFIS well ensemble empirical mode decomposition and general regression neural network, EEMD-GRNN), (iii) nonlinear (multi-layer perceptron, MLP). The output variable these was measure suspended particles while predictor variables information quality which consisted CO, NO2, O3, previous day, PM2.5, SO2 meteorological data included average atmospheric pressure (AP), maximum temperature (Max T), minimum (Min daily relative humidity level (RH), total precipitation (TP), wind speed (WS) for year 2016 Tehran. Analysis revealed that comparison with results MLR MLP, obtained most accurate (R2 = 0.97, root mean square error (RMSE) = 1.0713, absolute (MAE) = 0.6111) training phase (R2 = 0.89, RMSE = 3.6165, MAE = 2.8993) testing phase. However, used current study had almost similar prediction results. As can be concluded, models, turn out have higher accuracy concentration.