Comparison of multiple linear regressions (MLR) and artificial neural network (ANN) in predicting the yield using its components in the hulless barley.

摘要

In this study 40 genotypes in a randomized complete block design with three replications for two years were planted in the region of Ardabil. The yield related data and its components over the years of the analysis of variance were combined. Results showed that there was a significant difference between genotypes and genotype interaction in the environment. MLR and ANN methods were used to predict yield in barley. The fitted model in a yield predicting linear regression method was as follows: Reg=1.75+0.883X1+0.05017X2+1.984X3 Also, yield prediction based on multi-layer neural network (ANN) using the Matlab Perceptron type software with one hidden layer including 15 neurons and using algorithm after error propagation learning method and hyperbolic tangent function was implemented, in both methods absolute values of relative error as a deviation index in order to estimate and using duad t test of mean deviation index of the two estimates was examined. Results showed that in the ANN technique the mean deviation index of estimation significantly was one-third (1/3) of its rate in the MLR, because there was a significant interaction between genotype and environment and its impact on estimation by MLR method. Therefore, when the genotype environment interaction is significant, in the yield prediction in instead of the regression is recommended of a neural network approach due to high yield and more velocity in the estimation to be used.