Turkish Journal of Veterinary & Animal Sciences




Various machine learning algorithms have been used to model and predict the body weight of rams of the Balochi sheep breed of Pakistan. The traditional generalized linear model along with regression trees, support vector machine, and random forests methods have been used to develop models for the prediction of the body weight of animals. The independent variables (inputs) include the body (body length, heart girth, withers height) and testicular (scrotal diameter, scrotal circumference, scrotal length, and testicular length) measurements of 131 male sheep 2-36 months of age. The performance of the models is assessed based on evaluation criteria of mean absolute error, mean absolute percentage error, correlation between observed and fitted values, coefficient of determination, and root mean squared error. A 10-fold cross-validation is done on a training dataset to check the stability of the models. A separate training dataset is used to assess the predictive performance of the developed models. The random forests model was found to provide the best results for both training and testing datasets. It was concluded that machine learning methods may provide better results than the traditional models and may help practitioners and researchers choose the best predictors for body weight of farm animals.


Body weight, ram sheep, body measurements, machine learning

First Page


Last Page