Recognition of lactation pattern curves through neural networks and discriminate analysis, during the first third of lactation in dairy cows from the IX Region

Abstract

In this study the efficiency of the Artificial Neural Models and Linear Discriminant Analysis was demonstrated, to recognize and classify models or forms of lactation curves. The optimal neural model, using the Multi-level Perceptron with a hidden layer was obtained when five nodes were used (Artificial Neuron) in the inner layer. Although a higher performance of the neural models was observed compared to the linear discriminant analysis, this was slight and there was no difference in the statistical test of proportional difference (P-Value = 0.7224). This can be illustrated in the percentage of achievements in the classification for both methodologies; 97.87% for the neural model and 95.7% for the linear discriminant models, both in the training sample and validation sample.