<![CDATA[ Diabetes Mellitus is a chronic metabolic disease that has a significant impact on public health due to the risk of serious complications, such as heart disease and kidney failure. Early detection is crucial to prevent these complications. The application of machine learning has proven effective in improving the accuracy of diabetes classification. This study aims to evaluate the effectiveness of the Stacking Ensemble technique compared to individual models, XGBoost and LightGBM, in classifying diabetes. The dataset used is the Diabetes Health Indicators from the CDC, consisting of 253,680 samples and 21 features. The preprocessing stages include normalization, class balancing using SMOTE, and an 80:20 train-test data split. The models were evaluated using accuracy, precision, recall, F1-score, and ROC-AUC metrics. The results show that the Stacking Ensemble achieved the highest accuracy (91.79%), followed by LightGBM (91.29%) and XGBoost (90.78%). The highest precision was achieved by the Stacking Ensemble (96.97%), while the highest recall was obtained by LightGBM (87.04%). These findings indicate that the ensemble learning method can enhance the accuracy of diabetes prediction, thereby supporting more accurate medical decision-making. ]]>
Copyrights © 2025
