Through analyzing the influencing factors of congenital heart disease (CHD), it is aimed to establish CHD risk prediction model in fetus, and simultaneously provide theoretical foundation for CHD prevention. One-factor logistic regression method was used to screen the significant factors regarding CHD, and to separately adopt multiple-factor non-conditional logistic regression method and decision tree to set up model prediction fetus CHD risk and to analyze the advantages and shortcomings. Correct classification rates turned to be 80.93% and 82.79% respectively among 215 'training samples' by the two methods and the rates were 85.45 % and 89.09% respectively among 55 'testing samples'. The alliance of logistic regression and decision tree can overcome influence by co-linearity to guarantee the accuracy and perfection, as well as promoting the predictive accuracy.%通过对先天性心脏病(CHD)影响因素的分析,建立胎儿CHD危险度预测模型.采用单因素logistic回归分析筛选影响因素后用多因素非条件logistic回归和决策树法建立胎儿CHD危险度预测模型,分析比较两种预测方法 的优势与不足.实例分析表明,logistic回归模型和决策树模型对215例训练样本和55例测试样本的分类正确率分别为80.93%、82.79%和85.45%、89.09%.将logistic回归和决策树方法 联合应用,不仅能提高预测的准确率,还能克服因素间共线性的影响,从而保证分析的准确和完善.
展开▼
