A Comparative Evaluation Of Machine Learning Algorithms For Early Detection Of Type 2 Diabetes

This paper presents a comparative evaluation of seven supervised ML classifiers Logistic Regression, Naive Bayes, k-Nearest Neighbors, Decision Tree, Random Forest, XGBoost, and SVM for early Type 2 diabetes (T2DM) prediction. Using the Pima Indians Diabetes Database (PIDD, n=768) and Frankfurt Hospital Diabetes Dataset (FHDD, n=2000), we apply standardized preprocessing with SMOTE-based class imbalance correction and stratified 10-fold cross-validation[1]. Models are evaluated on Accuracy, Precision, Recall, F1-Score, ROC-AUC, and MCC. Results show XGBoost consistently outperforms all classifiers (AUC: 0.901 PIDD, 0.951 FHDD), while Logistic Regression retains interpretability advantages for clinical deployment. Feature importance analysis identifies fasting plasma glucose, BMI, and HbA1c as top predictors, aligning with clinical guidelines.