<![CDATA[Hyperparameter tuning plays a crucial role in determining the accuracy and reliability of machine learning models for medical diagnosis, particularly in the identification of critical diseases such as heart disease, which involve complex clinical data characteristics. However, standard machine learning models such as Support Vector Machine (SVM) often suffer from classification bias when hyperparameters are not properly configured. To address this limitation, this study aims to improve heart disease prediction performance through SVM optimization using the Grid Search method. The experiments were conducted using the Cleveland heart disease dataset obtained from the UCI Machine Learning Repository, consisting of 303 patient records and 13 predictive features. The target variable was transformed into a binary classification problem (healthy versus diseased). To prevent data leakage, preprocessing procedures—including missing value imputation and feature normalization—were strictly integrated within a closed machine learning pipeline. The performance of the optimized SVM model was then compared with a baseline SVM model using an independent hold-out test set and a nested cross-validation framework evaluated through ROC-AUC metrics. The experimental results demonstrate that systematic hyperparameter optimization significantly improves predictive performance. The findings indicate that the combination of Grid Search and a rigorous validation strategy produce a robust heart disease prediction model with strong generalization capability when applied to unseen clinical data.]]>
Copyrights © 2026
