Scholarlink

Background <p>Live body weight is a key performance indicator in goat production, playing a crucial role in selection, feeding management, and economic efficiency. However, direct measurement of body weight under field conditions is often impractical due to limitations in equipment and labor. Therefore, reliable prediction models based on easily obtainable morphometric measurements are required. Recently, machine learning approaches have gained increasing attention in animal science due to their ability to model complex and non-linear relationships among variables.</p> Methods <p>This study evaluated the effectiveness of the Light Gradient Boosting Machine (LightGBM) algorithm for predicting live body weight in Ankara, Hair, Kilis, and Honamlı goat breeds. Relationships between live body weight and morphometric traits were examined using Pearson correlation analysis. LightGBM models were developed under different training–testing split ratios (65–35%, 70–30%, 75–25%, and 80–20%). Model performance was assessed using error metrics and variable importance analyses to determine the contribution of individual body measurements to prediction accuracy.</p> Results <p>The results showed positive and statistically significant relationships between live weight and all morphometric measurements considered (<i>p</i> &lt; 0.05). In particular, basic body measurements such as body length, chest circumference, and wither height were identified as the variables that contributed most to live weight estimation at all data splitting ratios. Variable significance analyses revealed that the LightGBM algorithm exhibited stable performance and high generalization capacity under different training-test ratios. However, the contribution of variables such as sex and birth weight to model performance was found to be limited.</p> Conclusions <p>In conclusion, the LightGBM algorithm has been determined to be an effective tool for quickly, practically, and reliably estimating live weight in different goat breeds. It is considered that this approach can contribute to the development of decision support systems for selection, herd management, and performance evaluation processes in small ruminant farming.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluation of the performance of the Light Gradient Boosting Machine (LightGBM) algorithm in predicting live weight in different goat breeds

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月8日

查看全文
  • Esra Yavuz

摘要

Background

Live body weight is a key performance indicator in goat production, playing a crucial role in selection, feeding management, and economic efficiency. However, direct measurement of body weight under field conditions is often impractical due to limitations in equipment and labor. Therefore, reliable prediction models based on easily obtainable morphometric measurements are required. Recently, machine learning approaches have gained increasing attention in animal science due to their ability to model complex and non-linear relationships among variables.

Methods

This study evaluated the effectiveness of the Light Gradient Boosting Machine (LightGBM) algorithm for predicting live body weight in Ankara, Hair, Kilis, and Honamlı goat breeds. Relationships between live body weight and morphometric traits were examined using Pearson correlation analysis. LightGBM models were developed under different training–testing split ratios (65–35%, 70–30%, 75–25%, and 80–20%). Model performance was assessed using error metrics and variable importance analyses to determine the contribution of individual body measurements to prediction accuracy.

Results

The results showed positive and statistically significant relationships between live weight and all morphometric measurements considered (p < 0.05). In particular, basic body measurements such as body length, chest circumference, and wither height were identified as the variables that contributed most to live weight estimation at all data splitting ratios. Variable significance analyses revealed that the LightGBM algorithm exhibited stable performance and high generalization capacity under different training-test ratios. However, the contribution of variables such as sex and birth weight to model performance was found to be limited.

Conclusions

In conclusion, the LightGBM algorithm has been determined to be an effective tool for quickly, practically, and reliably estimating live weight in different goat breeds. It is considered that this approach can contribute to the development of decision support systems for selection, herd management, and performance evaluation processes in small ruminant farming.