Just Accepted

This article has been peer reviewed and accepted for publication. It is in production and has not been edited, so may differ from the final published form.

Predicting Weaning Weight in Hair Goat Kids: A Comparative Analysis of Tree-Based Machine Learning Approaches Using Environmental and Herd Data

Necati Esener , Yavuz Kal

Abstract

Context The study addresses the need for reliable predictive models to estimate weaning weight in Hair goat kids. Given the importance of weaning weight in livestock management and the evolving landscape of goat farming, this study explores new methodologies to enhance prediction accuracy. Aims The primary aim is to investigate the use of tree-based machine learning approaches for early prediction of weaning weight in Hair goat kids. Specifically, this study intends to assess the efficacy of various tree-based algorithms and evaluate their performance using key metrics. Methods The study employed multiple tree-based ensemble machine learning algorithms, including Bootstrap aggregating (bagging) with Random Forest (RF) and Extremely Randomized Trees (ExtraTree), as well as boosting techniques such as Extreme Gradient Boosting (XGB), Categorical Boosting (CatBoost), and Light Gradient Boosting Machine (LightGBM). The dataset comprised animal records, herd features, and external factors like weather and geographical variables. Model performance was assessed using the coefficient of determination (R2), adjusted coefficient of determination (Adj-R2), mean absolute percentage error (MAPE), mean absolute error (MAE), and mean squared error (MSE). Key Results All tree-based ensemble models performed notably, with R2 and Adj-R2 scores ranging between 0.78 and 0.80, MAPE score of 0.14, MAE scores between 2.19 and 2.13, and MSE scores between 7.80 and 8.36. A key finding was the significant role played by external data, including weather variables (temperature, dew, humidity) and grassland-related features (elevation, longitude, latitude), in predicting weaning weight. This demonstrates that incorporating diverse environmental factors into predictive models can yield accurate results. Conclusions The results highlight the success of tree-based ensemble machine learning approaches in predicting weaning weight in Hair goat kids. The inclusion of a broader range of data, particularly environmental factors, contributed to the accuracy of these models. Implications The study's findings offer valuable insights for goat farmers and breeders. The demonstrated effectiveness of machine learning applications provides a pathway for enhanced production management and decision-making, especially in response to changing climate conditions. This research marks a significant advancement in the precision and applicability of predictive models within goat production systems.

AN24154  Accepted 19 May 2025

© CSIRO 2025