Performance Evaluation of Logistic Regression, Random Forest, and SVM Models in Heart Disease Prediction

Abstract

Early identification of high-risk patients for cardiovascular disease is critical for reducing morbidity and improving treatment outcomes. This study applies supervised machine learning techniques to predict heart disease using the publicly available Kaggle heart failure dataset, which comprises 918 observations with demographic, clinical, and laboratory attributes, including age, resting blood pressure, cholesterol level, fasting blood sugar, maximum heart rate achieved, ST depression induced by exercise (Oldpeak), and electrocardiographic and chest pain characteristics. The dataset was pre-processed using a unified pipeline that standardized numerical features and encoded categorical variables via one-hot encoding. The data were split into training and testing sets using an 80/20 stratified approach. Three classification algorithms like Logistic Regression, Random Forest, and Support Vector Machine (SVM) with a radial basis function kernel were evaluated using accuracy, precision, recall, F1-score, and ROC–AUC metrics, complemented by confusion matrices and ROC curves. All models demonstrated strong predictive performance, achieving test accuracies of approximately 0.88. The SVM model exhibited the highest discriminative capability, with a ROC–AUC of approximately 0.95, while Logistic Regression achieved the highest recall (≈ 0.93), making it particularly suitable for applications where minimizing false negatives is critical. Correlation analysis identified Oldpeak, maximum heart rate, age, and fasting blood sugar as key factors associated with heart disease. These findings suggest that relatively simple machine learning models, when combined with appropriate preprocessing, can serve as effective decision-support tools for heart disease risk stratification in clinical settings.