<![CDATA[Hypertension is a major global health problem that significantly contributes to cardiovascular disease and mortality. This study evaluates the performance of Random Forest, XGBoost, and Support Vector Machine (SVM) algorithms integrated with Correlation-Based Feature Selection (CFS) for hypertension classification using hospital clinical data. The dataset comprises 500 clinical records containing demographic and physiological variables. CFS was applied to reduce irrelevant and redundant attributes before model training. Model performance was assessed using accuracy, precision, recall, F1-score, and AUC-ROC through 10-fold cross-validation. Statistical significance was examined using the Friedman test followed by the Wilcoxon signed-rank test with Bonferroni correction. The results show that CFS improved classification performance across all models by approximately 5–6%. XGBoost achieved the best performance with 93.5% accuracy and 0.96 AUC, followed by Random Forest and SVM. However, systolic and diastolic blood pressure, which define the hypertension label, were retained as predictors, indicating a diagnostic classification design rather than independent risk prediction. Therefore, the findings should be interpreted as dataset-based hypertension classification, not future hypertension risk prediction.]]>
Copyrights © 2026
