Estimation of Marketing Live Weight of Lambs by Different Machine Learning Algorithms

Abstract

Background: Further research is needed to estimate the marketable live weight of lambs with high accuracy and reliability while minimizing contact and measurement. This study aimed to estimate the 120th-day marketing weight of Morkaraman lambs by different machine learning algorithms, considering the variables of dam age, sex, birth type, birth weight, as well as 30th day, 60th day and 90th day live weights. Methods: Artificial neural networks (ANN), classification and regression trees (CART), support vector machines with radial basis function kernel (SVMR) and Random Forest (RF) algorithms for estimation of the marketing weight were performed for training (75%) and testing (25%) datasets. Models used in this study were compared based on mean absolute error (MAE), root mean squared error (RMSE) and mean absolute per cent error (MAPE) performance metrics. The most significant predictor of the marketing live weight in all models was the 90th day live weight, whereas the birth weight, birth type and dam age were the least important predictors. The correlation coefficients between live weight values estimated by the SVMR, CART, RF and ANN models and the actual marketing live weight were determined as 0.82, 0.82, 0.82 and 0.84, respectively. Result: The best prediction for the marketing live weight of Morkaraman lambs in the 4 th month was obtained from the ANN model. Using artificial neural networks to determine the marketing weight of lambs can save time and labor because of the reduced number of weighings. It may improve decisions made in flock management.