Sensitive analysis of meteorological data and selecting appropriate machine learning model for estimation of reference evapotranspiration

Abstract This study applies three methods, Gene Expression Programming (GEP), M5 tree (M5T) model and optimized Artificial Neural Network by Genetic Algorithm (ANN-GA) for estimation of reference evapotranspiration in Ahvaz Dezful the southwest Iran. Comparison between results FAO Penman - Monteith (FPM) method mentioned methods shows that ANN-GA with Levenberg-Marquardt training is best M5T second appropriate evapotranspiration. In Ahvaz, R 2 RMSE are 0.996, 0.184 mm/day. For method, these values 0.997 0259 mm/day, GEP they 0.979 0.521 Dezful, 0.994, 0.235 0.992 0265 0.963 0.544 addition, sensitivity analysis maximum temperature most effective parameter, wind speed parameter. effect more than those but less Ahvaz. Because flatter (the movement freer Dezful). The third meteorological parameter average relative humidity sunny hours Dezful. reason this subject distant from Persian Gulf (it source moisture).