Comparison of Artificial Neural Network and Multiple Linear Models in Estimation of Fat-tail weight on Fat-tailed Breeds and their Crosses

All breeds of Irannians sheep except Zel has a fat tail, and despite their lower carcass fat percentage, male lambs have higher fat-tail weight. Using within breed genetic variation requires accurate and precise measuring of fat tail weight on candidates of selection. The aime of this study was comparision of artificial neural network (ANN) modeling and linear modeling methods to prediction of fat tail weight, using body weight and different tail dimensions. 32 lambs of Chal and Zandi breeds,crosses of Zandi×Chal,Zel×Zandi and Zel×Chal hybrids were used for modeling to an estimation of fat-tail weight. Inputs of the model was birth type, sex, breed, upper width , mid width and lower width of fat tail,fat tail height and body weight, output of the model was fat tail weight. body weight, genotype, and fat tail mid-width had the largest positive correlations with fat-tail weight,0.83,-0.82 and0.80,respectively. The adequacy parameters of the best artificial neural network model had a coefficient determination of 0.99 and a mean squared error(RMSE)of 70.3g. The values of these estimated parameters by the multiple linear model were 0.891 and 263.86, respectively. The results of the extension of the original study showed the complexity of the interactions between the model inputs. Present research approved to accurate and unbiased estimation of tail weight of different breeds and crosses using artificial neural network. Furthermore, present study showed that ANN model can be used for accurate and presise estimation of fat tail weight using measured traits on sheep,than linear model.