In this study, neural network (NN) models were trained to predict the egg weight in broiler breeder hens. The input variables for developing the NN models were metabolizable energy (ME; kcal/bird per day), and crude protein (CP), total sulfur amino acid (TSAA), lysine (Lys), Ca (calcium), AP (available phosphorus), and linoleic acid (LA) all as g/bird per day. By breaking down the collected data from 98 breeder houses into weekly intervals, four NN-based models were developed for 25 to 28 weeks of age. From the available data set (98 data lines for each week), a training (n=69) and a testing set (n=34) were extracted. The developed models were subjected to an optimization algorithm to find the optimal values of input variables that may maximize early egg weight in broiler breeder hens. The goodness of fit statistical criteria indicated that the NN-based models could efficiently estimate egg weight in broiler breeder hens. The optimization results revealed that the maximum egg weight may be obtained with 406, 454, 466, and 487 kcal/ bird per day ME; 21.3, 24.9, 25.6, and 26 g/bird per day CP; 0.88, 0.97, 1.09, and 1.1 g/bird per day TSAA; 1.02, 1.1, 1.22, and 1.23 g/bird per day Lys; 4.13, 4.8, 5.2 and 5.27 g/bird per day Ca; 0.52, 0.57, 0.6, and 0.62 g/bird per day AP; 1.97, 2.01, 2.28, and 2.3 g/bird per day LA for 25, 26, 27, and 28 wk of age, respectively. The results showed that the energy and other nutrient requirements of broiler breeder hens for maximum egg weight do not change in parallel with age. Moreover, it seems that the Ross guideline recommendation underestimated the nutrient requirements of hens during these weeks.

Keywords