利用机器学习构建和优化早期体尺性状对成年母鸡腹脂沉积的预测模型  

Construction and Optimization of Prediction Models for Abdominal Fat Deposition in Adult Hens based on Early Body Size Traits using Machine Learning

作　　者：杨苗苗 谢莉 简宝怡 罗超维 谢卓君 朱飘 周天日 李华[1] 向海 YANG Miaomiao;XIE Li;JIAN Baoyi;LUO Chaowei;XIE Zhuojun;ZHU Piao;ZHOU Tianri;LI Hua;XIANG Hai(Guangdong Provincial Key Laboratory of Animal Molecular Design and Precise Breeding,School of Animal Science and Technology,Foshan University,Foshan 528225,China)

机构地区：[1]佛山大学动物科技学院,广东省动物分子设计与精准育种重点实验室,佛山528225

出　　处：《畜牧兽医学报》2025年第2期548-558,共11页ACTA VETERINARIA ET ZOOTECHNICA SINICA

基　　金：国家科技创新2030-重大项目(2023ZD04064);国家自然科学基金(32102538);广东省基础与应用基础研究基金(2022A1515012014);2023年度广东省大学生创新创业训练计划项目(202311847014)。

摘　　要：为探究活体、无创、简单、高效的母鸡腹脂沉积性状测定和选育技术方法,本研究以清远麻鸡为研究对象,将多体尺性状选择法与8种机器学习模型相结合,分别构建不同日龄体尺性状对母鸡腹脂含量的回归预测模型和分类预测模型。利用58~136日龄间各个日龄的多个早期体尺性状结合机器学习方法,体尺测定日龄对预测成年清远麻母鸡腹脂含量的准确性未表现出明显差异;进行回归预测时,RF模型的预测效果最好,拟合效果R 2为0.821~0.861,预测误差MAE为6.32~7.27;进行分类预测时,Bagging模型在二分类、三分类中均具有更高的预测准确度,二分类准确度ACC可达94.54%~100%,三分类准确度ACC可达99.58%~100%。本研究基于机器学习建立并优化了优质鸡腹脂沉积活体预测模型,能够为优质鸡腹脂早期活体选育等奠定技术基础,也为腹脂含量预测模型构建的相关技术探索提供参考。In order to investigate a live,non-invasive,simple,and efficient method for the assessment of abdominal fat deposition traits and selection of hens,the present study took Qingyuan partridge chickens as research object and combined the multiple body size trait selection method with eight machine learning models to construct regression prediction models and classification prediction models for the abdominal fat deposition of hens at different days of age.Using multiple early body size traits at various ages between 58 and 136 days of age combined with machine learning methods,the accuracy of body size traits at different ages in predicting abdominal fat content of adult Qingyuan partridge hens did not show significant differences.The RF model had the best prediction effect for regression prediction,with a fitting effect of R 2 of 0.821-0.861 and a prediction error MAE of 6.32-7.27.In terms of classification prediction,the Bagging model exhibited superior performance in both binary and tertiary classification.The binary classification accuracy ACC reached 94.54%to 100%,while the tertiary classification accuracy ACC reached 99.58%to 100%.In this study,live prediction models for abdominal fat deposition in high-quality chickens were established and optimized based on machine learning.These models can serve as a technical foundation for early live selection on abdominal fat deposition and breeding of high-quality chickens as well as for the exploration of the related technology for the construction of prediction models for abdominal fat content.

关 键 词：鸡 腹脂沉积 早期体尺性状 机器学习 预测模型 

分 类 号：S831.2[农业科学—畜牧学]