Sex-Specific Ensemble Models for Type 2 Diabetes Classification in the Mexican Population

Abstract

Background: Type 2 diabetes (T2D) is considered a global pandemic by the World Health Organization (WHO), with a growing prevalence, particularly in Mexico. Accurate early diagnosis remains a challenge, especially when accounting for biological sex-based differences. Purpose: This study aims to enhance the classification of T2D in the Mexican population by applying sex-specific ensemble models combined with genetic algorithm-based feature selection. Materials and Methods: A dataset of 1787 Mexican patients (895 females, 892 males) is analyzed. Data are split by sex, and feature selection is performed using GALGO, a genetic algorithm-based tool. Classification models including Random Forest, K-Nearest Neighbor, Support Vector Machine, and Logistic Regression are trained and evaluated. Ensemble stacking models are constructed separately for each sex to improve classification performance. Results: The male-specific ensemble model achieved 94% specificity and 96% sensitivity, while the female-specific model reached 96% specificity and 90% sensitivity. Both models demonstrated strong overall performance. Conclusion: The proposed sex-specific ensemble models represent a clinically valuable approach to personalized T2D diagnosis. By identifying sex-specific predictive features, this work supports the development of precision medicine tools tailored to the Mexican population. This contributes to improving diagnostic precision and supporting more equitable and personalized approaches in clinical settings.