基于XGBoost算法的2型糖尿病精准预测模型研究  

Research on Type 2 Diabetes Mellitus Precise Prediction Models Based on XGBoost Algorithm

作　　者：张洪侠[1] 郭贺[1] 王金霞 徐岩艳 吕斌[1] 闫东 常佳[1] 胡光瑞[1] 王雪[1] 李洪军[1] 刘天戟[1] 李燕林 赵志强 牛晓强 ZHANG Hong-xia;GUO He;WANG Jin-xia;et al(China-Japan Union Hospital of Jilin University ,Changchun 130033,China)

机构地区：[1]吉林大学中日联谊医院,吉林长春130033 [2]北京青梧桐健康科技有限公司

出　　处：《中国实验诊断学》2018年第3期408-412,共5页Chinese Journal of Laboratory Diagnosis

摘　　要：目的构建2型糖尿病发病风险预测模型。方法在体检人群中招募糖尿病患者53人,非糖尿病患者93人,体检的同时进行相应的基因检测,并填写健康体检自测问卷,收集全部数据采用XGBoost构建2型糖尿病预测模型。结果模型预测的准确率是86.6%,特征重要性评估结果显示,对模型贡献前三名的变量依次是血糖、甘油三酯和SLC30A8基因rs13266634-C位点的等位基因。结论 XGBoost糖尿病发病风险预测模型具有很强的预测能力。Objective To construct prediction models to estimate the risks of developing type 2 diabetes mellitus.Methods 53 persons with diabetes and 93 persons without diabetes from physical examination were Recruited for the study.To detect diabetes-related genes polymorphism and common clinical indicators of all volunteers.The volunteers also filled out self-measured questionnaires.XGBoost algorithm was used to construct diabetes prediction model to estimate risk of diabetes.Results The predictive accuracy of model is 86.6%.The first three variables that contribute to the model are fasting blood glucose,Triglyceride,SLC30 A8-rs13266634-C in the evaluation of relative importance of feature.Conclusion The prediction models we constructed have high predictability.

关 键 词：XGBoost算法 2型糖尿病 精准预测模型 

分 类 号：R587.1[医药卫生—内分泌]