Comparative Study of Machine Learning Algorithms for Heart Attack Prediction

Heart disease remains a leading cause of global mortality, underscoring the importance of early and accurate detection methods. This study investigates the application of machine learning techniques for heart attack detection using a clinical dataset. A comprehensive methodology was adopted, beginning with data preprocessing that included missing value imputation, normalization of numerical features, and categorical encoding through one-hot transformation. Multiple models were developed and analyzed, including Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, XGBoost, K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). Model evaluation was conducted using stratified train-test splits and cross-validation, with performance measured across accuracy, precision, recall, F1-score, and ROC-AUC. The results revealed that ensemble-based models significantly outperformed traditional classifiers. Random Forest, Gradient Boosting, and XGBoost each achieved an accuracy of 98.48 %, with precision, recall, and F 1 -score values of 0.9877, and ROC-AUC values above 0.98. Notably, Random Forest achieved the highest ROC-AUC of 0.9958, indicating superior discriminative ability. In contrast, Logistic Regression attained moderate performance (79.9 % accuracy), while KNN$(66.7 \%)$and SVM ($74.6 \%)$lagged behind, reflecting limitations in handling the dataset's complexity. Comparative analysis underscores the reliability and robustness of ensemble learning approaches for clinical prediction tasks. This work highlights the potential of machine learning, particularly ensemble methods, as decision-support tools in healthcare, aiding early detection and reducing the risk of misdiagnosis. Future research may extend this work by incorporating hybrid deep learning models, larger datasets, and advanced feature selection techniques to further enhance prediction accuracy and interpretability.