基于决策树算法的TIR对2型糖尿病肾病的诊断作用

Role of TIR in the Diagnosis of Type 2 Diabetic Nephropathy Based on Decision Tree Algorithm

为了解葡萄糖目标范围内时间(简称TIR)在2型糖尿病(T2DM)并发症糖尿病肾病(简称DKD)诊断预测中的作用,提出基于决策树算法的DKD诊断预测方法.该方法利用决策树算法构建一棵DKD诊断预测树,并分析各项指标和增减TIR对DKD诊断预测的不同权重,得出TIR对DKD诊断结果的重要性数据.该方法简单高效、易转化成规则、可解释性好、精度高.实验结果表明:基于决策树算法的含TIR的DKD诊断预测准确率为100%,而不含TIR的DKD诊断预测准确率为98.1%,说明基于决策树和TIR进行数据驱动的DKD诊断预测精度高;TIR在DKD诊断中具有不可替代的作用,在理论上TIR低于78.5%的2型糖尿病患者被确诊为DKD的风险可能增大,但此结论还有待临床上进一步研究证实;TIR在10项诊断指标中的重要性排名为第4,可视为重要诊断指标.

In order to understand the role of glucose time in target range(TIR)in the diagnosis and prediction of diabetic nephropathy(DKD)with type 2 diabetes mellitus(T2 DM)complications,this paper proposes a diagnosis and prediction method of DKD based on decision tree algorithm.This method uses the decision tree algorithm to constructs a DKD diagnosis and prediction tree,and analyzes the different weights of various indexes and increasing or decreasing TIR to DKD diagnosis and prediction,and obtains the importance data of TIR to DKD diagnosis results.This method is simple and efficient,easy to convert into rules,good interpretability and high precision.The experimental results show that the accuracy of DKD diagnosis and prediction based on decision tree algorithm with TIR is 100%,while the accuracy of DKD diagnosis and prediction without TIR is 98.1%,indicating that the data-driven DKD diagnosis and prediction based on decision tree and TIR has high accuracy;TIR has an irreplaceable role in the diagnosis of DKD.Patients with TIR less than 78.5%may theoretically have an increased risk of DKD diagnosis,but this conclusion remains to be further confirmed by clinical research;The importance of TIR ranks 4 th among 10 diagnostic indicators,which can be regarded as an important diagnostic indicator.