An Ensemble Learning and Machine Learning Approach for Predicting Diabetic Disease

Traditional diabetes prediction models often fall short in capturing complex, multifactorial patterns. This study introduces an ensemble machine learning approach combining Logistic Regression (LR), Random Forest (RF), Adaptive Boosting (AB), and XGBoost, applied across diverse age groups to enhance prediction accuracy. Using a dataset of 253,680 health records, our model tackles imbalanced data and nonlinear relationships, achieving 85.5% accuracy with AB and 87% with XGBoost. Comparative results show XGBoost's superiority with a higher F1-score (0.95) and lower error rate (13%) than alternative methods. This model's robustness provides a comprehensive framework for early diabetes prediction, offering valuable insights for healthcare interventions. Future research will explore broader health features to further refine and expand predictive capabilities.