Evaluation of Artificial neural network modeling for predicting performance traits of Arian broiler breeder

This study was conducted for predicting performance traits of the whole production period of Arian broiler breeder via its initial performances using artificial neural network (ANN). The input variables of modeling were house, number of hens in the house, weeks of production, body weight at 20-24 weeks of age and outputs of the model were body weight, egg number, egg mass, egg weight and feed intake at the 25 to 47 weeks of age. The used ANN model for prediction of body weight had 5 inputs, 5 neurons at 1st hidden layer, 2 neurons at 2nd hidden layer and 1 output, thus we write it as 5-5-3-1. Similarly, the optimized ANN model structure for feed intake, egg number, egg weight and egg mass were 7-7-4-1, 8-8-4-1, 7-7-3-1 and 7-7-3-1, respectively. R2 of adequate models for BW, FI, EN, EW and EM were 0.991, 0.998, 0.989, 0.993 and 0.996, and Root Mean Square Error were 1.55, 0.992, 0.266, 3.838 and 0.506, respectively. The results of the study shown that architecture and the specification of the neural networks such as inputs, outputs, number of neurons and number of hidden layers can affect the performance of the ANN model. The results indicated the possibility of predicting whole production period of Arian broiler breeder using early stage production records.