下一代数字医疗：基于人工智能与数据科学的糖尿病风险早期评估智能网络平台

Abstract

Diabetes mellitus is a priority public health issue due to its rising prevalence and the complications associated with sustained hyperglycemia. Early detection enables the implementation of preventive interventions and reduces the risk of cardiovascular, renal, neurological, and visual damage.

The objective of this study was to design and evaluate—through a simulated experimental exercise—an intelligent web platform for estimating diabetes risk based on clinical variables and machine learning algorithms.

A synthetic dataset comprising 1,250 records was created, and the performance of Logistic Regression, Support Vector Machines, Random Forest, and XGBoost was compared. The simulation results indicated that XGBoost achieved the best performance, with an accuracy of 96.3%, precision of 96.1%, sensitivity of 95.8%, and an F1 score of 95.9%.

Fasting glucose, glycated hemoglobin, and body mass index were the variables that contributed most to the predictions. The study concludes that integrating web-based systems with classification models can support initial screening, provided the tool undergoes clinical validation, ensures personal data protection, and remains under professional supervision.