C5.0决策树法在出生缺陷预测中的应用

目的 介绍决策树法的原理及其在出生缺陷预测中的应用,为出生缺陷研究提供一种新的思路.方法 通过1:2匹配的病例对照研究探讨湖南省前10位出生缺陷发生的影响因素;对单因素logistic回归分析中有统计学意义的变量采用C5.0决策树算法和判别分析构建预测模型.统计分析软件采用Clementine 11.0和SPSS 15.0.结果 决策树分类结果与实际类别的符合率为83.7%,灵敏度为74.1%,特异度为88.6%;判别分类与实际类别的符合率为64.7%,灵敏度为54.0%;特异度为70.3%.C5.0决策树法比判别分析法具有更好的预测效果,其判断准确率高于判别分析.结论 C5.0决策树法构建的出生缺陷预测模型,可获得比传统的判别分析更好的预测效果.通过建立孕妇资料数据库,结合专业知识选取高质量的指标,应用决策树法能够对出生缺陷的发生起到较好的预测作用.

Application of Decision Tree C5.0 in the Pre-warning of Birth Defects

Objective To introduce the principle of Decision Tree and its application in the forecasting of birth defects, and provide a new way in the study of birth defects. Methods 1:2 matched case-control study was used to explore the influencing factor of ten top birth defects in Hunan. Decision Tree C5.0 algorithm and discriminate analysis were used to construct forecast models. Clementine 11.0 and SPSS 15. 0 were used for statistical analysis. Results The coincident rate of Decision Tree categories with actual categories was 83.7% ;the sensitivity was74.1 %;the specificity was 88.6%. The coincident rate of discriminate categories with actual categories was 64.7% ;the sensitivity was 54.0% ;the specificity was 70.3%. Compared with discriminate analysis,Decision Tree methods had a better forecasting precision for its higher accuracy rate. Conclusion Compared with the conventional statistic method, Decision Tree methods had better forecasting precision. Decision Tree method is a very good forecasting model to predict birth defects.