Use of machine learning approaches for body weight prediction in Peruvian Corriedale Sheep

• DOI: 10.1016/j.atech.2024.100419
• Publication Date: 2024-02-20T07:55:14Z
• Authors: Ali William Canaz...; Roxana Churata-Hu...; Cihan Çakmakçı; Francisco Halley ...; Júlio Sílvio de S...; Tales Jesus Ferna...; Yhan Carlos Rojas...
• Keywords: Perimeter; Gradient boosting; Rump; Corriedale

The goal of this study was to predict the body weight Corriedale ewes using machine learning (ML) algorithms. Fourteen measurements (BM) and six different models were used. Body (BW) BM: wither height (WH), rump (RH), thoracic perimeter (TP), abdominal (AP), foreshank length (FSL), fore-shank width (FSW), (FSP), tail (TW), (TPe), hip (HW), loin (LWi), shoulder (SW), forelimb (FL), (BL), collected from 100 between 1.5 2 years old Illpa Experimental Centre National University Altiplano in Peru. algorithms used estimate Support Vector Machines for Regression (SVMR), Classification Trees (CART), Random Forest (RF), Model Average Neural Networks (MANN), Multivariate Adaptive Splines (MARS) eXtreme Gradient Boosting (XGBoost). performance evaluated by coefficient determination (R2), root mean square error (RMSE), absolute (MAE), percentage (MAPE). Highly correlated predictors (r ≥ 075) removed dataset. remaining then subjected variable selection procedures Boruta algorithm. results confirmed importance TP, LWi, BL, FSL, SW HW as ewe weight. ML trained on those selected predictors. RF had highest R2 values lowest MAE, RMSE, MAPE. In conclusion, algorithm can be recommended accurately estimating BW sheep.