A Framework for Early Diabetes Prediction and Interpretation: Sex-specific Factor Analysis Approach Using Ensemble Classifiers

Diabetes mellitus is a prevalent metabolic disorder, underscoring the need for predictive strategies to enable early detection and timely intervention. This study presents a novel approach that utilizes Exploratory Factor Analysis (EFA) to enhance both interpretability and predictive accuracy, with an emphasis on sex-specific analysis. By leveraging EFA for dimensionality reduction, the model achieves significant improvements in both prediction performance and interpretability. Tailoring the analysis to sex-specific factors, this approach provides distinct insights for male and female populations, optimizing early detection strategies. Ensemble classifiers were employed, achieving optimal F1 scores and Area Under the Curve of the Receiver Operating Characteristic (AUC-ROC) scores of 97.62% and 0.9992 for males, and 98.15% and 0.9811 for females, respectively. These results underscore the efficacy of EFA-driven dimensionality reduction in clinical decision-making, marking a substantial contribution to diabetes prediction and personalized healthcare. Keywords: Diabetes Prediction, Exploratory Factor Analysis, Sex-specific Analysis, Ensemble Classifiers, Early Detection