摘要
马匹体重是反映与衡量其健康状况的重要指标之一,并在马匹选育、肉质评价、饲养管理、马匹鉴定等方面具有重要参考意义。传统马体重估测模型的特征值之间存在共线性问题。故文中利用85匹一至三岁伊犁马的胸围、体高、体长信息作为特征值,采用K均值聚类算法确定隐含层中心点位置,并构建了基于径向基函数(RBF)的神经网络体重估测模型。模型采用平均绝对离差与线性拟合优度作为评价指标。线性伊犁马体重估测模型的平均绝对离差为15.45 kg,决定系数R 2为0.688,基于RBF神经网络的伊犁马体重估测模型的平均绝对离差为7.75 kg,决定系数R 2为0.917。研究结果表明:RBF神经网络模型能有效去除特征值之间的共线性问题,提高伊犁马体重估测准确度。基于RBF神经网络的伊犁马体重估测模型效果优于线性回归、通用性马体重估测模型,为准确估测伊犁马体重提供了新思路。
A horse’s weight is one of the most important indexes to reflect and measure its health condition,and provides crucial reference for several aspects such as horse breeding,evaluation of meat quality,feeding and management,horse identification,etc.There are multicollinearity problems between the characteristic values of traditional horse weight estimation model.As a result,we make the data including chest circumference,height at withers and body length of 85 Yili horses between one-year-old and three-year-old as the characteristic value,use the K-means clustering algorithm to identify the center point position of the hidden layer and build a neural network estimation model of weight basing on the radial basis function(RBF)that adopts mean absolute deviation and linear goodness of fit to be the evaluation index.The mean absolute deviation of the linear Yili horse weight estimation model is 15.45 kg,and the determination coefficient R 2 is 0.688.The mean absolute deviation of the Yili horse weight estimation model based on RBF neural network is 7.75 kg,and the determination coefficient R 2 is 0.917.The research shows that the RBF neural network model can efficiently remove those multicollinearity problems between characteristic values and improve the accuracy of the weight estimation for Yili horses.The neural network estimation model of weight basing on RBF is more effective than that of the linear regression and generality model,which has provided a new thinking way for precisely estimating the weight of Yili horses.
作者
朱让东
张太红
郭斌
ZHU Rang-dong;ZHANG Tai-hong;GUO Bin(School of Computer and Information Engineering,Xinjiang Agricultural University,Urumqi 830001,China)
出处
《计算机技术与发展》
2020年第3期198-203,共6页
Computer Technology and Development
基金
新疆维吾尔自治区重大科技专项(2017A01002-5)。
关键词
伊犁马
体重
估测
径向基函数
神经网络
Yili horse
weight
estimation
radial basis function
neural network
