Machine Learning-Based Prediction Model Construction for Type 2 Diabetes Mellitus: A Comparison of Algorithms and Multilevel Risk Factor Analysis - PubMed
6 hours ago
- #Type 2 diabetes prediction
- #machine learning
- #multilevel risk factors
- The study constructed a machine learning-based prediction model for Type 2 Diabetes Mellitus (T2DM), comparing seven algorithms.
- AdaBoost performed best with an AUC of 0.85, accuracy of 0.71, and F1 score of 0.71 after parameter optimization.
- Data from NHANES (2021-2023) included 6337 participants; missing values were handled via Monte Carlo imputation and collinearity reduced with PCA.
- Feature selection used random forest and recursive feature elimination; class imbalance addressed with ADASYN.
- 24 key risk factors identified across multilevel: 19 at individual traits, 3 at individual behavior, and 2 from working/living conditions.
- Innovation integrates health ecology model with machine learning for systematic cross-level risk factor identification.
- Model provides a new tool for T2DM early screening, high-risk individual identification, and targeted public health interventions.