Use of new methods to determine the inputs effective in estimating soil temperature

In this research, an estimate of the depth of 10 cm in the soil of the Synoptic Station of Tabriz in East Azerbaijan province was carried out using artificial neural network (ANN) and backward vector machine (SVM). Two main component analysis (PCA) and gamma (GT) tests were used for pre-processing data and input data. According to the results, for Tabriz station, 3 input variables were selected by gamma test. In the main components analysis method, four main components for the synoptic station of Tabriz were selected. The results of modeling indicate that the gamma-based gamma-ray machine (GT-SVM) model with a mean square error of 2.48 ° C can be selected as the selected model for the station. The most important variables known to estimate the temperature of the soil were the average temperature, sunshine, wind speed and relative humidity, respectively, by gamma test. Finally, according to the results, it can be concluded that the methods used for pre-processing the data in this study do not differ significantly in soil temperature prediction, and both methods have worked well. Also, the SVM model in all estimations has a more acceptable performance than the ANN model.