Use of neural network models to estimate early egg production in broiler breeder hens through dietary nutrient intake

In this study, neural network (NN) models were constructed to predict early egg production in broiler breeder hens. By breaking down the early egg production data collected from 98 breeder houses into weekly intervals, 5 NN-based models were developed for 25 to 29 wk of age. Starting with 98 data lines for each week, the NN models were trained by 69 data lines and the remainder (n = 29) were considered as the testing set. The variables of interest for developing the models were ME (kcal/bird per day) and CP, TSAA, Lys, Ca, and available P (g/bird per day). The constructed models were subjected to an optimization algorithm. Therefore, the optimal values for the input variables to maximize early egg production in broiler breeder hens were obtained. Based on the considered criteria to evaluate the goodness of fit, the efficiency of NN-based models to estimate early egg production was confirmed. The optimization results revealed that the breeder hens consuming 407, 457, 470, 486, and 487 kcal of ME/bird per day showed the highest egg production during 25, 26, 27, 28, and 29 wk of age, respectively. Moreover, optimal performance of hens required the intake (g/bird per day) of the following during 25, 26, 27, 28, and 29 wk of age, respectively: CP: 20.3, 22.6, 25, 25.8, and 26; TSAA: 0.88, 1.02, 1.06, 1.07, and 1.07; Lys: 0.98, 1.0, 1.2, 1.3, and 1.32; Ca: 4.5, 4.6, 5.3, 5.0, and 5.4; and available P: 0.48, 0.55, 0.6, 0.61, and 0.62. Although the results showed that the energy and other nutrient requirements of broiler breeder hens during early egg production do not change in parallel with age, it seems that the company recommendations underestimated the nutrient requirements of hens during these weeks.