Integrating Machine Learning with Morphometric Indicators to Predict Body Weight in Female Indigenous Matebele Goats Across Tooth-Age Classes in Zimbabwe

Accurate estimation of body weight (BWT) using linear body measurements (LBMs) is critical in smallholder goat systems, where scales are often unavailable. This study evaluated four machine learning (ML) models—Random Forest (RF), Decision Tree (CART), Support Vector Regression (SVR), and Multi-Layer Perceptron (MLP)—to predict BWT from heart girth (HG), withers height (WTH), body length (BL), and rump height (RH). Using 132 female goat observations, RF achieved the highest predictive performance (R² = 0.759; RMSE = 2.425 kg), with HG identified as the dominant predictor, corroborating previous findings of its strong correlation with BWT. CART models offer interpretable decision rules suitable for field applications, highlighting their practical utility for smallholder farmers. These results demonstrate that ML approaches can provide accurate and accessible tools for predicting BWT and informing feeding strategies, health interventions, selection, and market decisions in resource-limited goat production systems.