Comparison of Logistic Regression and Support Vector Machine Algorithm Performance in Heart Failure Prediction

General Background: Heart failure remains a global health concern due to its high prevalence and mortality rates. Specific Background: In Indonesia, heart disease ranks as a leading cause of death, emphasizing the need for reliable predictive tools. Knowledge Gap: Despite the availability of machine learning algorithms, limited studies provide a comparative evaluation using updated clinical datasets in the Indonesian context. Aims: This study aims to compare the performance of Logistic Regression and Support Vector Machine (SVM) in predicting heart failure. Results: Using a dataset of 918 samples and rigorous preprocessing, SVM achieved 90% accuracy with an AUC of 0.93, outperforming Logistic Regression, which scored 88% accuracy and an AUC of 0.9304. SVM demonstrated superior sensitivity and robustness in handling non-linear data, while Logistic Regression offered better calibration for risk interpretation. Novelty: The study’s novelty lies in its integrated open-source pipeline, use of biomedical signal features, and statistical validation via McNemar’s test. Implications: These findings support the implementation of SVM in automated clinical decision systems for early heart failure detection, while highlighting Logistic Regression's value in interpretability-focused clinical settings.Highlight : SVM has higher accuracy and F1-score, indicating superior classification performance. Logistic Regression remains superior in probability calibration, important for clinical risk interpretation. Both models effectively support the development of machine learning-based health prediction systems.