摘要
目的:建立糖尿病-糖尿病肾病患者数据集,基于机器学习算法,建立糖尿病肾病风险预测模型,从2型糖尿病患者中筛选出糖尿病肾病患者,辅助进行糖尿病肾病诊断。方法:收集医院内分泌科2型糖尿病患者和糖尿病肾病患者数据,建立糖尿病-糖尿病肾病患者数据集;基于Logistic回归方法进行特征筛选,选取疾病影响因素;使用K近邻、逻辑回归、决策树等8种算法建立风险预测模型,并对预测模型进行评价比较。结果:筛选出糖尿病病程、视网膜病变、尿肌酐等9个因素与糖尿病肾病的发生显著相关,纳入风险预测模型训练集中;建立模型经验证比较,发现使用随机森林算法建立模型预测准确率最高,准确率达到92.74%。结论:基于随机森林算法建立模型预测效能最好,在2型糖尿病患者人群中可以实现糖尿病肾病风险精准预测,为糖尿病肾病诊断提供帮助。
Objective:To establish a data set of patients with diabetes and diabetic nephropathy,establish a risk prediction model of diabetic nephropathy based on machine learning algorithm,screen out patients with diabetic nephropathy from patients with type 2 diabetes,and assist in the diagnosis of diabetic nephropathy.Methods:Collect the data of patients with type 2 diabetes and diabetic nephropathy in the endocrinology department of the hospital,and establish a data set of diabetes-diabetic nephropathy patients;Feature screening based on Logistic regression method to select disease influencing factors;Use K-nearest neighbors,logistic regression,decision tree and other 8 algorithms to establish risk prediction models,evaluate and compare predictive models.Results:Nine factors including diabetes duration,retinopathy and urinary creatinine were screened out to be significantly correlated with the occurrence of diabetic nephropathy,they were included in the training set of risk prediction models;The established model is verified and compared,and it is found that the prediction accuracy of the model established by the random forest algorithm is the highest,and the accuracy rate reaches 92.74%.Conclusion:The prediction performance of the model established based on the random forest algorithm is the best,which can accurately predict the risk of diabetic nephropathy in the population of type 2 diabetes mellitus,and provide help for the diagnosis of diabetic nephropathy.
作者
徐澄
彭丹
Xu Cheng;Peng Dan(The Affiliated Hospital of Xuzhou Medical University,Xuzhou 221000,China)
出处
《无线互联科技》
2022年第17期131-136,共6页
Wireless Internet Technology
关键词
糖尿病肾病
机器学习
预测模型
diabetic nephropathy
machine learning
prediction model
