<![CDATA[Diabetes Mellitus is a chronic disease with a continuously increasing prevalence, posing serious challenges to public health and contributing significantly to the global economic burden. The often non-specific nature of early symptoms increases the risk of delayed diagnosis, highlighting the need for accurate early detection approaches to support clinical decision-making. This study aims to analyze and compare the performance of three machine learning algorithms Logistic Regression, Random Forest, and XGBoost in classifying diabetes risk based on several clinical parameters, including age, body mass index (BMI), blood pressure, glucose level, and HbA1c. The dataset used in this research was obtained from the Diabetes Prediction Dataset, consisting of 100,000 records. The research process involved handling missing data, applying One-Hot Encoding to categorical variables, normalizing numerical features, and addressing class imbalance using the Synthetic Minority Over-sampling Technique (SMOTE). Model performance was evaluated using Accuracy, Precision, Recall, F1-Score, and ROC-AUC metrics to provide a comprehensive assessment. The experimental results indicate that XGBoost achieved the best performance, with an accuracy of 96.88% and a ROC-AUC value of 98.00%. Meanwhile, Random Forest attained an accuracy of 95.68% with an F1-Score of 74.76%, while Logistic Regression recorded an accuracy of 88.96% and the highest recall value of 89.12%. These findings suggest that ensemble learning methods, particularly boosting approaches, are more effective in improving the accuracy of diabetes and non-diabetes classification. The primary contribution of this study lies in providing a multi-metric comparative analysis that can serve as a reference for selecting the most effective machine learning model in the development of medical decision support systems for early diabetes detection.]]>
Copyrights © 2025
