PREDICTIVE MODELLING OF GUINEA FOWL GROWTH AND AGE TRAITS USING MACHINE LEARNING ALGORITHMS

Accurate prediction of body weight and age in guinea fowl (Numida meleagris) is vital for optimizing breeding, nutrition, and management practices. Traditional assessment methods are often time-consuming, invasive, and susceptible to human error. This study applies machine learning (ML) algorithms to predict body weight and age from biometric traits, aiming to enhance data-driven decision-making in poultry production. Morphometric measurements, including body length, shank length, thigh length, wing length, and chest girth, were obtained from 300 guinea fowls sampled at the Dutsin-Ma, Shanono, and Maigatari weekly markets in Katsina, Kano, and Jigawa States, respectively, 100 from each State. Three algorithms, Feedforward Neural Network (FFNN), Support Vector Machine (SVM), and Trilayered Model (TM), were trained and validated using R, R², MAE, MSE, and RMSE metrics. Results revealed that the FFNN achieved the highest prediction accuracy (R² = 0.87; RMSE = 101.7 g), outperforming SVM and TM. For age prediction, FFNN achieved R² = 0.79 and RMSE = 15.9 days, indicating high temporal precision. These findings demonstrate that ML models can accurately estimate growth traits using easily measurable parameters. These models offer a reliable, non-invasive, and scalable tool for supporting precision livestock management and improving productivity in smallholder poultry systems.