DEVELOPMENT OF AN ARTIFICIAL NEURAL NETWORK TO PREDICT ANIMAL AND FORAGE PRODUCTION

This study focuses on training the mathematical models for prediction of forage and animal production in Brazilian beef cattle system. In this study, two functions are trained to find the most optimal prediction of herbage mass besides leaf and stem percentages, and average daily gain. We aimed to compare artificial neural networks (ANNs) and multiple linear regression (MLR) to predict both forage and animal production. Two datasets were used in each evaluation. The multivariable results showed that there was no formation of groups in each dataset, so all inputs were used in analyses. These analyses to determine the best model ANN or MLR results, considering the correlation between the predicted value and the observed value. Other evaluations were performed for ANNs, more specifically for structures. The inputs and the number of hidden layers was analyzed to define the best structure for prediction of future results. Significance level was considered by P-value < 0.05. It was found that ANN is better than MLR predicting results for both datasets. For the inputs used in each ANN, there were differences only for animal production, with the higher prediction values 0.72 using ANN. In other words, the number of hidden layers for both datasets were not different. Hence, ANN, with a specific structure for each evaluation, is a potential tool for prediction of results for forage and animal production.