COMPARISON OF FIVE MACHINE LEARNING ALGORITHMS FOR PREDICTION OF BODY WEIGHTS OF CATTLE

Abstract

This study compares five machine-learning algorithms for estimating cattle body weights using the linear body measurements of the animals. Bunaji, Rahaji, and Sokoto Gudali breeds of cattle were used to gather information on body weight and linear body measurements. One hundred of each breed were used to sample the animals in Adamawa State's Yola North Local Government Area. Approved reference marks for cattle by FAO were used for the linear body measurements. All the variables of body weight and linear body measurements were entered into the Python programming language. Five machine learning algorithms were compared to the model's performance. The techniques are K-Nearest Neighbor Regressor, Support Vector Regressor, Extreme Gradient Boosting (XGB), Random Forest (RF), Neural Network (NN), and Support Vector Regressor (SVR) (KNN). The proportion of variation in the data set was explained by the predictor variables up to 95.7% and 95.4%, respectively, according to the coefficient of determination (R2 ) for the two algorithms. The Support Vector Regressor's (SVR) performance was the least effective, with the lowest R2 and greatest MSE, RMSE, and MAE values. Extreme Gradient Boosting (XGB), a machine learning algorithm, is the most effective at predicting the body weight of the three Nigerian cattle breeds studied. This research should lead to the creation of computer applications that will make determining cattle body weight easier. More data from Nigerian cattle breeds should be added to the model.