Prediction of Body Mass of Dairy Cattle Using Machine Learning Algorithms Applied to Morphological Characteristics

Area of Science:

• Animal Science
• Agricultural Engineering
• Data Science

Background:

• Accurate body mass (BM) prediction is essential for effective cattle herd management, impacting nutritional strategies and biological efficiency assessments.
• Morphological measurements offer a non-invasive method for estimating BM in livestock.

Purpose of the Study:

• To evaluate and compare the accuracy of various models for predicting body mass in lactating Holstein cows using morphological data.
• To identify the most effective predictors and modeling techniques for on-farm application.

Main Methods:

• Spearman's correlation was used to identify key morphological predictors (thoracic perimeter, abdominal perimeter, rump width) from a dataset of 465 Holstein cows.
• Simple linear regression, multiple linear regression, artificial neural networks (ANNs), and Support Vector Regression (SVR) were employed and compared.
• A dataset split of 90% training, 5% validation, and 5% testing was utilized to assess model performance.

Main Results:

• Thoracic perimeter (TP), abdominal perimeter (AP), and rump width (RW) showed the strongest correlations with body mass.
• Multiple regression models incorporating TP, AP, and RW significantly improved prediction accuracy (R² = 0.9067, RMSE = 28.00 kg).
• Artificial neural networks (ANNs) achieved the highest accuracy (R² = 0.9125, RMSE = 25.86 kg), closely followed by SVR (R² = 0.9046, RMSE = 27.41 kg).

Conclusions:

• Advanced machine learning models like ANNs and SVR provide superior accuracy for body mass prediction in cattle compared to traditional regression methods.
• While complex models offer higher precision, simpler regression models remain practical alternatives for routine on-farm use.
• Accurate BM prediction using morphological data supports optimized herd management and efficiency in dairy cattle.