Comparative Analysis of Machine Learning Models for Predicting Heart Disease

Abstract

Heart disease is currently one among the principal causes of death worldwide, with multiple factors contributing to its development, such as genetic predisposition, health conditions, and lifestyle choices. When the narrowing of a heart valve exceeds 50%, the patient is typically considered to be at risk for cardiac arrest. Detecting early signs that could lead to heart attacks and related conditions has become a key focus of research. This study compares the performance of various classification models in predicting heart disease, including random forests, logistic regression, support vector machines, and decision trees. The results from this experimental analysis indicate that logistic regression shows higher precision, while random forests achieve higher accuracy compared to the other models. Random forests also demonstrate better functioning in terms of both accuracy and recall. Both models outperform the others when evaluated by F1 Score. Hyperparameters are optimized using grid search.