MORPHOMETRICTRAITS IN NIGERIA

This study evaluated the application of machine learning models in predicting live weight of Nigerian indigenous goat breeds using morphometric traits. Data were collected from three major breeds, which include Red Sokoto, West African Dwarf, and Sahel goats. The data was collected across several markets in Benin City, Nigeria. Morphometric measurements including heart girth, body length, height at withers, and rump height were taken from each goat and analyzed using machine learning models such as Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN) and Extra Decision Tree(EDT). The Extra Decision Tree(EDT) model recorded the highest prediction accuracy (89.22%) and coefficient of determination (R² = 93.76%), while also achieving the lowest Mean Square Error (MSE = 1.23) and Mean Absolute Error (MAE=0.05). These results indicate that Extra Decision Tree provided the most reliable and stable prediction of live weight among the evaluated models. The findings demonstrate that morphometric based machine learning approaches can serve as efficient, low-cost tools for accurate live weight estimation in indigenous goat breeds, supporting improved productivity and management in Nigeria’s livestock systems.