糖尿病预测的可解释机器学习框架:整合SMOTE平衡与SHAP可解释性用于临床决策支持
Pathamakorn Netayawijit1, Wirapong Chansanam2, Kanda Sorn-In3
1Department of Information Systems, Faculty of Business Administration and Information Technology, Rajamangala University of Technology Isan, Khon Kaen Campus, Khon Kaen 40000, Thailand.
Healthcare (Basel, Switzerland)
|October 29, 2025
概括
这项研究通过将合成少数群体过量采样技术 (SMOTE) 与夏普利添加式扩展 (SHAP) 结合起来,增强了用于糖尿病预测的机器学习. 随机森林-SMOTE模型实现了高准确性,并确定了像葡萄糖和BMI这样的关键预测指标.
科学领域:
- 医疗保健中的机器学习
- 糖尿病预测模型
- 临床决策支持的AI
背景情况:
- 阶级不平衡和有限的解释性阻碍了人工智能用于糖尿病预测的临床采用.
- 现有的模型往往对高风险病例的敏感性较差,缺乏临床信任.
- 这项研究将SMOTE重新采样与SHAP可解释性相结合,以提高性能和透明度.
研究的目的:
- 开发和验证用于糖尿病预测的可解释机器学习框架.
- 通过使用先进的重新采样技术 (SMOTE) 来解决类失衡.
- 通过SHAP提供临床上有意义的解释,以加强决策支持.
主要方法:
- 一个七个阶段的管道,结合合成少数群体过量采样技术 (SMOTE) 和夏普利添加物扩张 (SHAP).
- 在1500名患者记录中对五种算法 (随机森林,梯度提升,SVM,后勤回归,XGBoost) 的比较评估.
- 在培训折叠中使用SMOTE进行5倍分层交叉验证,以防止数据泄露.
主要成果:
- 随机森林-SMOTE模型实现了96.91%的准确性,0.998 AUC,99.5%的灵敏度和97.3%的特异性.
- 在SHAP分析中,葡萄糖 (SHAP值2.34) 和BMI (SHAP值1.87) 是主要预测因素.
- 特性相互作用分析显示了葡萄糖和BMI之间的协同效应.
结论:
- 拟议的框架显示,在结合算法性能和糖尿病预测的临床适用性方面具有前景.
- 随机森林-SMOTE模型在公开可用的数据集上展示了高交叉验证的性能.
- 在临床部署之前,需要在现实世界队列中进行进一步的外部,前性验证.
相关概念视频
Diabetes Mellitus: Overview and Type I Subtype
4.8K
Diabetes mellitus is a chronic metabolic disorder characterized by high blood glucose levels due to inadequate insulin production, insulin resistance, or both. The condition affects millions worldwide and can significantly impact their health and quality of life.
Type 1 diabetes is an autoimmune disease in which the immune system mistakenly attacks and destroys the insulin-producing beta cells in the pancreas. As a result, the body is unable to produce sufficient insulin, and individuals with...
Type 1 diabetes is an autoimmune disease in which the immune system mistakenly attacks and destroys the insulin-producing beta cells in the pancreas. As a result, the body is unable to produce sufficient insulin, and individuals with...
4.8K
Diabetes: Management and Pharmacotherapy
858
The therapy for diabetes aims to alleviate hyperglycemia-related symptoms, prevent acute metabolic decompensation, and reduce chronic end-organ complications. Glycemic control is evaluated through short-term (self-monitoring, continuous glucose monitoring) and long-term (A1c, fructosamine) metrics, enabling near real-time tracking of blood glucose levels and reflecting glycemic control over specific time frames.
Insulin remains the cornerstone of treatment for most patients with type 1 and many...
Insulin remains the cornerstone of treatment for most patients with type 1 and many...
858
Diabetes Mellitus: Type 2 and Gestational
4.3K
Type 2 diabetes, characterized by insulin resistance, arises when the insulin receptors on cells lose responsiveness to insulin, diminishing the cell's capacity to take up glucose, resulting in elevated blood glucose levels. To receive a diagnosis of Type 2 diabetes, a series of blood glucose tests are necessary to assess whether the blood glucose falls within normal parameters. If the result is out of the normal range, a patient may be diagnosed as prediabetic or diabetic, depending on the...
4.3K
Pathophysiology of Diabetes
3.1K
Diabetes mellitus is a chronic metabolic disorder characterized by hyperglycemia. The four categories of diabetes are type 1 diabetes, type 2 diabetes, other specific types of diabetes, and gestational diabetes.
Type 1 diabetes is characterized by autoimmune-mediated destruction of pancreatic β cells, with environmental factors potentially triggering this process in genetically susceptible individuals. Despite many not having a family history, certain genes increase susceptibility,...
Type 1 diabetes is characterized by autoimmune-mediated destruction of pancreatic β cells, with environmental factors potentially triggering this process in genetically susceptible individuals. Despite many not having a family history, certain genes increase susceptibility,...
3.1K
