Modeling the yield of rainfed wheat, barley and alfalfa products using support vector regression and genetic programming

Climate change, the rise of global temperature, the water crisis, along with the growth of the world's population have made the world's food supply a challenge for researchers. For this reason, it is necessary to predict and simulate plant products in accordance with the climatic conditions. In this study, the relationships of meteorological parameters and standard precipitation index (SPI) and reconnaissance drought index (RDI) with yields of the rainfed wheat, barley and alfalfa plants were studied in three regions in East Azarbaijan province. For each of the temperature, rainfall, evapotranspiration and SPI and RDI parameters, the time intervals of three to nine months were considered in the period from 2004 to 2014. Then, using support vector regression (SVR) and genetic and programming (GP), the production amounts of the three studied plants were predicted. In addition, the accuracy of the mentoned methods in predicting the performance of dry crop products was evaluated using root mean squared error (RMSE) and mean absolute error (MAE) statistics. Results showed that in Tabriz for alfalfa, GP method with RMSE= 0.17 (kg ha-1), in Maragheh for the alfalfa, SVR with RMSE= 0.56 (kg ha-1) and in Sarab for barely, SVR method with RMSE=0.20 (kg ha-1) had more precise predictions. It can be stated that the use of climatic factors and drought indicators of autumn, winter and spring seasons have significant effects on increasing the accuracy of soft computing techniques in predicting the performance of rainfed products.