Machine learning algorithm approaches for predicting body weight in tuis Huacaya alpacas

The purpose of this study is to examine how machine learning (ML) techniques can predict the body weight of Huacaya alpacas ( Vicugna pacos ) based on body measurements. Eighteen body measurements (BM) and six different ML models were used. Body weight (BW) and BM: head length (HL), ear length (EL), head width (HW), interorbital distance (ID), head height (HH), neck length (NL), upper neck perimeter (UNP), lower neck perimeter (LNP), wither height (WH), back height (BH), rump height (RH), dorsal length (DL), distance between ischial tips (DBI), tail length (TL), Thoracic Perimeter (TP), Abdominal Perimeter (AP), anterior fore-shank perimeter (AFP) and hoof length (HoL), were collected from 153 10 month old Huacaya alpacas of both sexes, from the Quimsachata annex of the Instituto Nacional de Innovación Agraria in Puno Department, Peru. The machine learning algorithms used to estimate body weight were Artificial Neural Networks (ANN), eXtreme Gradient Boosting (XGBoost), Multivariate Adaptive Regression Splines (MARS), Support Vector Machines for Regression (SVM), Random Forest (RF) and Classification and Regression Trees (CART). The performance of the models was evaluated by the mean absolute percentage error (MAPE), mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R 2 ). Predictors with a high correlation (r ≥ 0.75) were removed from the dataset. The remaining predictors were then processed through variable selection procedures using the Boruta algorithm. The results from the Boruta algorithm confirmed that HW, UNP, BH, DBI, TL, TP, AP and HoL are important predictors of alpaca weight. The ML models were then trained on those selected predictors. RF had the highest R 2 values and lowest values of MAE, RMSE, and MAPE. In conclusion, the RF algorithm can be recommended for accurately estimating BW from body measurements of Huacaya alpacas of both sexes.