基于EMVS-logistic模型的关键因素筛选及风险预测--以糖尿病合并症高血压数据为例  

Screening of Key Factors andRisk Prediction Based on EMVS-Logistic Model:Take Diabetes Comorbidity Hypertension Data as an Example

作　　者：单苗慧 朱春华[1] 倪其伟 SHAN Miao-hui;ZHU Chun-hua;NI Qi-wei(School of Statistics and Data Science,Nanjing Audit University,Nanjing 210000,China;School of Mathematics and Statistics,Anhui Normal University,Wuhu 241000,Anhui,China)

机构地区：[1]南京审计大学统计与数据科学学院,江苏南京210000 [2]安徽师范大学数学与统计学院,安徽芜湖241000

出　　处：《兰州文理学院学报（自然科学版）》2022年第4期25-30,共6页Journal of Lanzhou University of Arts and Science(Natural Sciences)

基　　金：国家社科基金一般项目(21BTJ030)。

摘　　要：糖尿病是一种慢性疾病,已经成为危害我国人民健康的三大慢性疾病之一.深入对患者糖尿病危险因素与合并症的研究对于早期的糖尿病预防具有重要的影响意义.为解决实际数据中的“维数灾难”问题,利用EM算法贝叶斯变量选择(EMVS)、Lasso变量选择与Logistic回归结合,构建患病风险识别模型,对糖尿病合并症高血压数据集进行关键因素筛选及风险预测.Diabetes Mellitus is a chronic disease,which has become one of the three major chronic diseases that endanger people’s health in China.In-depth research on the risk factors and comorbidities of diabetes in patients has important implications for early diabetes prevention.In order to solve the“curse of dimensionality”problem in actual data,this paper uses the EM algorithm Bayesian variable selection(EMVS),Lasso variable selection and Logistic regression to build a disease risk identification model.Screening of key factors and risk prediction of diabetic hypertension data set are carried out.

关 键 词：贝叶斯变量选择 Lasso变量选择 糖尿病合并症高血压 EM算法 

分 类 号：O212.8[理学—概率论与数理统计]