基于机器学习的2型糖尿病并发症预测模型研究

目的：利用统计学和机器学习方法，探究2型糖尿病并发症的影响因素，构建2型糖尿病并发症预测模型，对并发症的发生进行预测，为2型糖尿病并发症的预防和早期筛查提供理论参考。方法：以国家人口健康科学数据中心2009年的2型糖尿病患者数据为研究对象，对性别、年龄、尿常规检查和生化检查等信息进行单因素和多因素logistic回归分析，并构建XGBoost模型，进行2型糖尿病并发症的预测。结果：单因素和多因素Logistic回归模型与XGBoost模型显示，2型糖尿病并发症的发生与14项影响因素具有相关关系，模型预测准确率为82.85%。结论：模型预测的效果良好，具有一定的参考价值。

Machine learning-based prediction model of type 2 diabetes mellitus complications

Objective To study the influencing factors of type 2 diabetes mellitus complications using statistical and machine learning methods, and to establish the prediction model of type 2 diabetes mellitus complications for predicting their incidence in order to provide the theoretical reference for their early prevention and screening. Methods The gender, age, routine urine examination and laboratory testing parameters were analyzed by univariate logistic regression analysis and multivariate logistic regression analysis respectively with the data of diabetes mellitus patients in 2009 covered in the National Population Health Science Data Center as the study object. The XGBoost model was established for predicting the type 2 diabetes mellitus complications. Results Univariate and multivariate logistic regression models combined with XGBoost model showed that the incidence of type 2 diabetes mellitus complications was related with 14 influencing factors with a prediction accuracy of 82.85%. Conclusion The predicting efficacy of univariate and multivariate logistic regression models combined with XGBoost model is good with a certain reference value.