摘要
建立集约化牧场奶牛亚临床乳腺炎的预测模型,实现对该病的早期准确预警,对保障奶牛福利、繁殖与生产性能及牧场经济效益具有重要意义。首先,基于自动监控设备与榨奶系统采集的多维数据(泌乳量、活动量、反刍时间、电导率等),分析发现患病牛与健康牛在各监测变量上差异显著,且患病牛组变量值从d-3至d-0普遍呈下降或上升趋势;其次,分别采用人工神经网络与正交偏最小二乘判别分析两种机器学习算法构建奶牛亚临床乳腺炎预测模型;最后,通过提取并排序变量重要性特征,对比发现6个关键变量(ecsdd-2、ecpd-1、ecvpd-2、ecvpd-1、ecvpd-3、ecsdd-3)在两种算法中表现高度一致,说明这些变量可用于奶牛亚临床乳腺炎的预测。
To establish a prediction model for subclinical mastitis of dairy cows in intensive pasture and realize timely and accurate prediction of the disease is of great significance for protecting the welfare of dairy cows,ensuring reproductive performance,production performance and economic benefits of pasture.Firstly,the multi-dimensional data(milk yield,activity,rumination time,electrical conductivity,etc.)monitored by automatic monitoring equipment and milking system were analyzed.It was found that there were significant differences in various variables between the sick cattle and healthy cattle groups,and the variables in the sick cattle group showed a decreasing/increasing trend from d-3 to d-0.Secondly,the prediction model of subclinical mastitis of dairy cows was established by using artificial neural networks and orthogonal partial least squares algorithm.Finally,the important features of the variables were extracted and sorted,and the comparative analysis showed that six important variables(ecsdd-2,ecpd-1,ecvpd-2,ecvpd-1,ecvpd-3,ecsdd-3)showed high consistency between the two algorithms,indicating that these variables could be used to predict subclinical mastitis of dairy cows.
作者
郎迪
周晓晶
汪鹏森
刘丰羽
Lang Di;Zhou Xiaojing;Wang Pengsen;Liu Fengyu(Heilongjiang Bayi Agricultural University,Daqing 163319)
出处
《黑龙江八一农垦大学学报》
2025年第6期109-117,共9页
journal of heilongjiang bayi agricultural university
基金
黑龙江省生态环境保护科研项目(黑龙江省规模化牧场围产期奶牛常见病碳足迹评估)
北大荒集团农业科技进步贡献率测算、结构演变及提升对策研究(BDH2024-11)
黑龙江省农业农村厅科研项目(基于牧场大数据奶牛健康管理规范〔2020〕)
黑龙江八一农垦大学博士科研启动项目(XDB202305)。
关键词
预测
亚临床乳腺炎
人工神经网络
正交偏最小二乘判别
prediction
subclinical mastitis
artificial neural networks
orthogonal partial least squares discrimination
