Evaluation of artificial neural network techniques for estimating maze, wheat and barley yields

Precision farming attempts to improve cropping efficiency by correct planning، a necessary first step in this process is the search for techniques able to identify functional relationships between measure soil and site characteristics and crop yield. Agro ecological zoning of the region was obtained by the use of climate، soil and topographical data base and determined homogen zone for this study. Each crop yield potential was calculated and finally the crops yield were determined for each zone. A variety of supervised feedforward neural network techniques، including several variations of backpropagation were selected for investigation about barley maze and wheat yield prediction. the performance of all artificial neural networks was assessed based on calculating the mean absolute error (MAE)، the root mean square error (RMSE) and the coefficient of determination (R2) of linear regression line between the predicted values from either the ANN model and the desired output was also used as a measure of performance. Input layer include soil chemical and soil physical properties as well as climatical and topography factors. In this research use of multilayer feedforward neural network with backpropagation algorithm. Artificial neural network models with two hidden layer have good accuracy between predicted and observed yield in training and validation. ANN performance show significant preciseness from land condition for predicts yield.