Artificial Neural Network Models of the Ruminal pH in Holstein Steers

─In this study four Holstein steers with rumen fistula fed 7 kg of dry matter (DM) of diets differing in concentrate to alfalfa hay ratios as 60:40, 70:30, 80:20, and 90:10 in a 4 × 4 latin square design. The pH of the ruminal fluid was measured before the morning feeding (0.0 h) to 8 h post feeding. In this study, a two-layered feed-forward neural network trained by the Levenberg-Marquardt algorithm was used for modelling of ruminal pH. The input variables of the network were time, concentrate to alfalfa hay ratios (C/F), non fiber carbohydrate (NFC) and neutral detergent fiber (NDF). The output variable was the ruminal pH. The modeling results showed that there was excellent agreement between the experimental data and predicted values, with a high determination coefficient (R >0.96). Therefore, we suggest using these model-derived biological values to summarize continuously recorded pH data. Keywords─Ruminal pH, Artificial Neural Network (ANN), Non Fiber Carbohydrate, Neutral Detergent Fiber.