Predicting dry matter intake in Pelibuey sheep using machine learning methods

This study determined to predict the dry matter intake (DMI) in growing male Pelibuey sheep by using 3 different machine learning methods. Individual data was obtained from 130 animals whose average body weight (ABW) was 23 ± 6 kg and the DMI was 1.04 ± 0.27 kg/d from an experiment conducted under tropical conditions. To create the database, the following data were recorded: % concentrate in the diet (CON), initial body weight (IBW, kg), final BW (FBW, kg), mean metabolic BW (MBW^0.75, kg^0.75), daily weight gain (ADG, g/d), crude protein (CP) and neutral detergent fibre (NDF). Multivariate Adaptive Regression Splines (MARS), Classification and Regression Tree (CART), and Support Vector Regression (SVR) were used for the development of a predictive algorithm. The determination coefficient was determined over 0.90 for the MARS algorithm. Overall, the MARS algorithm was a reliable predictive model for DMI prediction in the Pelibuey sheep.