<![CDATA[Heart disease is one of the leading causes of mortality worldwide, emphasizing the need for accurate early detection systems. Machine learning models such as XGBoost have demonstrated strong performance in medical classification tasks; however, their effectiveness is highly dependent on optimal hyperparameter configurations. This study aims to improve the performance of XGBoost for heart disease classification by applying hyperparameter optimization using the Optuna framework with the Tree-structured Parzen Estimator (TPE) algorithm. The UCI Heart Disease dataset, consisting of 918 records, is used in this study. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) is applied to the training data. Model performance is evaluated using accuracy, precision, recall, F1-score, and ROC-AUC metrics. The experimental results show that the optimized XGBoost model achieves an accuracy of 89.13%, outperforming the baseline model with 87.50%, and improves recall from 87.50% to 89.10%. In addition, the optimized model attains a higher ROC-AUC value of 0.9319, indicating improved classification stability. These findings demonstrate that Optuna-based hyperparameter optimization effectively enhances the performance and reliability of XGBoost, making it suitable for supporting early heart disease diagnosis in medical decision support systems.]]>
Copyrights © 2026
