支持向量机和Logistic回归在GDM风险预测中的应用

Application of support vector machine and logistic regression in risk prediction of GDM

目的探讨两种不同机器学习算法在妊娠期糖尿病(gestational diabetes mellitus,GDM)风险预测中的应用。方法选取2019年7月—2020年8月在广州市妇女儿童医疗中心及广东省计划生育专科医院进行产前检查的孕早期妇女520例,其中妊娠期糖尿病孕妇200例,随机抽取同期正常孕妇320例,收集孕妇的一般资料和孕早期(8~12周)的生化指标、血常规和凝血功能等检测资料。利用这些分析变量建立支持向量机(SVM)和Logistic回归(LR)预测模型。根据模型预测能力和模型实用性,如准确率、精确率、真阳性(TP)率、假阳性(FP)率、召回率、F测度、受试者工作特征曲线(ROC)进行效果评价。结果两种预测模型的分类准确率总体为86%。SVM模型在真阳性(TP)率、假阳性(FP)率、召回率、F测度、受试者工作特征曲线(ROC)方面优于LR模型。结论在分类与预测方面,支持向量机算法比Logistic回归模型更具有实用价值。

Objective To explore the application of two different machine learning algorithms in the risk prediction of gestational diabetes mellitus(GDM).Methods A total of 520 pregnant women with gestational diabetes mellitus were selected from Women and Children s Medical Center and Guangdong Family Planning Hospital from July 2019 to August 2020,including 200 cases of gestational diabetes mellitus,and 320 normal pregnant women in the same period.The general information of pregnant women and the detection data of biochemical indexes,blood routine test and coagulation function in early pregnancy(8~12 weeks)were collected.Support vector machine(SVM)and logistic regression(LR)prediction models were established by using these analysis variables.According to the predictive ability and practicability of the model,something like accuracy rate,precision ratio,true positive(TP)rate,false positive(FP)rate,recall rate,F-measure and receiver operating characteristic curve(ROC)were evaluated.Results The classification accuracy of the two models was 86%.SVM model is better than LR model in TPrate,FPrate,recall rate,F measure and ROC.Conclusion Support vector machine is more practical than logistic regression model in classification and prediction.