基于决策树方法的糖尿病并发症分类研究

Research the Chronic Diabetic Complications Classfication based on Decision Tree Approach

目的:优选辅助诊断糖尿病并发症方法。方法:分别采用决策树方法 C5.1、CART、QUEST、CHAID,对糖尿病并发症临床数据进行分类研究。结果:经3 969例确诊病例验证,4种方法都能得到诊断规则和决策树,C5.1分类精确性为:85.714%,CART、QUEST、CHAID分类精确性为:66.667%。结论:4种方法分类结果符合专家诊断思维,C5.1效果最好。

Objective：to explore the optimal method for aided diagnosis the Chronic diabetic complications.Method： classfication for the data of Chronic diabetic complications based on C5.1, CART, QUEST, CHAID.Result：the four methods can gain diagnosis rules and Decision Tree,and the classification accuracy is respectively 85.714%,66.667%,66.667%,66.667% after test for 3969 confirmed case. Conclusion：the four methods accord with experts diagnosis thinking, the best method is C5.1.