基于异质性模型的低体重出生儿判别分析研究  

Research on the Discriminant Analysis of Newborns with Low Body Weight Based on Heterogeneous Model

作　　者：谢名阳 殷雨晨 王源昌[1] 李兴平[1] XIE Ming-yang;YIN Yu-chen;WANG Yuan-chang(School of Mathematics,Yunnan Normal University,Kunming 650500,Yunnan Province,P.R.C.)

机构地区：[1]云南师范大学数学学院,650500

出　　处：《中国数字医学》2020年第11期93-98,共6页China Digital Medicine

摘　　要：在低体重出生儿预测判别分析中,母亲行为习惯和身体特征是关键性因素,但相似的受试者母亲特征所生婴儿表现型差异较大,给低体重出生儿预测判别造成较大困惑。为提高低体重出生儿预测判别精准度,引入广义线性混合模型拟合个体异质性,构建异质性机器学习模型,通过异质性模型的分类数值模拟,其中异质性随机森林模型判别准确率达100%。数值模拟结果显示异质性模型在判别分析上有极大的优势,可以辅助医生有针对性地对受试者母亲进行诊断,实现低体重出生儿的预测精准智能化。In predictive discriminant analysis of low birth weight infants,the behavior habits and physical characteristics of mothers are the key factors,but the phenotypes of infants born with similar characteristics of mothers of subjects are quite different,which causes great confusion for predictive discrimination of low birth weight infants.In order to improve the accuracy of predicting and discriminating low birth weight infants,a generalized linear mixed model was introduced to fit individual heterogeneity,and a heterogeneous machine learning model was constructed.The classification and numerical simulation of heterogeneous model showed that the discriminating accuracy of heterogeneous random forest model was 100%.The results of numerical simulation show that the heterogeneity model has great advantages in discriminant analysis.It can assist doctors to diagnose the mothers of the subjects and realize the accurate and intelligent prediction of low birth weight infants.

关 键 词：异质性 异质性模型 婴儿出生体重 机器学习 

分 类 号：R319[医药卫生—基础医学]