Plasma and Milk Variables Classify Diet, Dry Period Length, and Lactation Week of Dairy Cows Using a Machine Learning...

Area of Science:

• Dairy science
• Animal nutrition
• Machine learning applications in agriculture

Background:

• Understanding cow physiology and management is crucial for optimizing dairy production.
• Early lactation is a critical period influencing cow health and productivity.
• Diet and dry period length significantly impact metabolic status and performance.

Purpose of the Study:

• To classify Holstein-Friesian cows based on diet, dry period (DP) length, and lactation week.
• To evaluate the effectiveness of machine learning models using body weight, milk, and plasma metabolites for classification.
• To identify key variables predictive of management strategies in early lactation.

Main Methods:

• A 3x2 factorial design experiment with 95 Holstein-Friesian cows.
• Three dry period lengths (0, 30, 60 d) and two early lactation diets (lipogenic, glucogenic).
• An XGBoost model trained on weekly body weight, milk variables, and plasma metabolites, validated with 1000 hold-out partitions.

Main Results:

• High classification accuracy (AUC > 0.9) for lactation week, independent of diet or DP length.
• Accurate classification of 0 vs. 60 d DP length (AUC > 0.8), better than other DP comparisons.
• Plasma urea and milk fat content were key for diet classification; milk yield and protein for lactation week.

Conclusions:

• Machine learning models effectively utilize milk and plasma metabolite data for retrospective classification of dairy cows.
• This approach can identify cows' management groups, including diet and dry period strategies, in early lactation.
• The findings highlight the potential of data-driven insights for dairy herd management.