Machine Learning-Based Heart Disease Prediction: Accuracy Comparison of Four Classification Algorithms

Heart disease continues to be one of the major causes of death worldwide, making early detection and accurate prediction essential for improving patient care and survival rates. This study compares the performance of four machine learning classification algorithms for heart disease prediction, namely Logistic Regression, Decision Tree, Random Forest, and K-Nearest Neighbors (KNN). The research uses a publicly available heart disease dataset from Kaggle, with several preprocessing techniques applied, including data cleaning, normalization, categorical encoding, feature selection, and class imbalance handling using the Synthetic Minority Oversampling Technique (SMOTE) to improve model effectiveness. The models are evaluated using accuracy, precision, recall, and F1-score, with particular focus on reducing false negatives because of their critical impact in medical diagnosis. The findings show that Logistic Regression achieves the highest accuracy (91.61%) and precision (53.72%), indicating strong overall classification capability. Meanwhile, the Decision Tree model demonstrates higher recall, making it more effective in identifying positive heart disease cases. The results of this study emphasize that selecting machine learning models for healthcare applications should not depend only on accuracy, but should also consider recall and clinical reliability to support early detection and better patient outcomes.