<![CDATA[Diabetes mellitus is a noncommunicable disease with a steadily increasing prevalence, necessitating accurate methods for early detection. This study aims to analyze the performance of the Support Vector Machine (SVM) method using three kernels—Linear, Polynomial, and Radial Basis Function (RBF)—in classifying diabetes risk. The data used is the Pima Indians Diabetes dataset, consisting of 768 observations with 8 predictor variables and one target variable. This study employs a quantitative approach involving data exploration, preprocessing, and model evaluation using 10-fold cross-validation. Model performance was evaluated based on the metrics Accuracy, Sensitivity, Specificity, Kappa, and ROC-AUC. The results showed that all three kernels demonstrated good classification capabilities; however, the Polynomial kernel outperformed both the Linear and RBF kernels. The Polynomial kernel achieved the highest Accuracy value of 77.60%, as well as superior Kappa and ROC-AUC values. Furthermore, the Polynomial kernel also demonstrated better balance in classifying diabetic and non-diabetic patients. These results indicate that kernel selection in the SVM method significantly affects classification performance, and the Polynomial kernel is the most optimal model for the dataset used in this study.]]>
Copyrights © 2026
