基于深度神经网络的糖尿病发病风险评估

打开文本图片集

关键词： 糖尿病; 深度神经网络; 风险评估

中图分类号：R699;R587.2

文献标识码：A

DOI：10.3969/j.issn.1006-1959.2026.09.010

文章编号：1006-1959（2026）09-0066-06

Diabetes Mellitus Onset Risk Assessment Based on Deep Neural Networks

LI Zilu 1 ， SONG Hao 2 ， LIU Yanmei 3

（1. Faculty of Information Engineering and Automation， Kunming University of Science and Technology，

Kunming 650000， Yunnan， China;

School of New Energy and Mining， Xinjiang University of Technology， Hetian 840000， Xinjiang， China;

3.The Sixth Clinical College， Qingyuan Hospital Affiliated to Guangzhou Medical University， Qingyuan 511500， Guangzhou， China） Abstract： This study constructed a diabetes mellitus risk assessment model based on deep neural network. The experimental data were from the NAGALA （1994–2016） cohort， including 19 characteristics such as demographic information and biochemical indicators. An adaptive sampling method based on clustering is adopted to solve the problem of class imbalance. Then， a three-layer deep neural network model is constructed， and the SHAP analysis method is used to divide the high and low risk groups. Finally， the performance of the model is verified by comparative analysis and risk assessment effect analysis. The results showed that HbA1c， FPG， age and BMI were the main influencing factors of diabetes mellitus. The deep neural network model can effectively distinguish whether the patient is sick in each follow-up period， and its performance is better than other machine learning models. There are significant differences in the overall and specific indicators between the high and low risk groups. The diabetes mellitus risk assessment model can accurately predict the incidence and risk of diabetes mellitus， and provides a powerful tool for early screening and risk assessment of diabetes mellitus.

Key words： Diabetes mellitus; Deep neural networks; Risk assessment

糖尿病（diabetes mellitus）是全球发病率逐年上升的慢性疾病，根据世界卫生组织的数据显示，糖尿病在全球范围内已成为导致心血管疾病、肾脏病等并发症的重要因素[1]。（剩余9065字）