Collaboratrices et collaborateurs

Predicting dry matter intake (DMI) is crucial for farm management. The study aimed to develop a model based on readily available variables on a dairy farm. The data set collected in barn, between November and March for respectively 2015 to 2021, included 6202 weekly averaged daily observations in 413 lactations from 273 Holstein cows. Data were analyzed using R Statistical Software (v4.1.2; R Core Team 2022). A correlation analysis was performed to identify the best predictive variables. We identified five animal-related and two diet-related variables as the most correlated to DMI. Afterwards, a model was developed using backward regression from a full model consisting 7 variables. After removing non-significant variables (days in gestation, metabolic body weight, and dietary net energy of lactation), the selected predictive variables (mean ± SD) were the following: parity 2.5 ± 1.6; WOL 14.9 ± 10.9; energy corrected milk ECM 33.7 ± 7.6 kg/d and neutral detergent fiber (NDF) 354 ± 59 g/kg DM). The metrics used for the evaluation were the root mean squared prediction error (RMSPE), its decomposition, and a concordance correlation coefficients (CCC) analysis was performed. The resulting prediction equation was DMI(kg/d) = 16.86 - (0.18 × Parity) + (0.24 × ECM(kg/d)) - (0.02 × NDF(g/kg DM))) + (0.29 × WOL) - (0.005 × WOL2) (RMSPE = 1.45 kg, 7.16% and CCC = 0.90). Where parity is equal to 1 for multiparous and 0 for primiparous. In conclusion, our study successfully developed a predictive model for dry matter intake (DMI) in lactating dairy cows using easily accessible variables in dairy farms.