Predicting yield of wheat genotypes at different salinity by artificial neural network

For examining efficiency of Artificial Neural Network (ANN) for prediction of yield in wheat, a factorial experiment was set out in a randomized complete block design (RCBD) with three replications in a glasshouse. The treatments were included of four saline solutions and 8 wheat genotypes. This paper shows the ability of artificial neural network (ANN) technology to be used for the prediction of yield and yield components of 8 wheat genotypes for different salinity levels. Based on analysis of variance, salinity had significant effect on all traits, as salinity levels increased, yield and 1000-grain weight and K+ concentrations decreased. The results showed that a very good performance of the ANN model was achieved. Some explanation of the predicted results is given. The ANN with training algorithm of back propagation was the best one for creating of nonlinear mapping between input and output parameters. The ANN model predicted the six yields and yield components with mean R 2 and T values of 0.977 and 0.97 respectively. Furthermore, the predictions of ANN model were compared with those obtained from six multi-linear regression (MLR) models. It was found that ANN model has better predictions than the MLR models for the experimental data.