Heart Disease Prediction using Machine Learning and Comparative Classification

Heart disease is a major health concern, and recognizing patterns in routine clinical measurements may help identify patients who require further medical assessment. This study examines the use of supervised machine-learning methods to classify the presence or absence of heart disease using the Cleveland Heart Disease dataset. The dataset contains 303 patient records described by thirteen clinical attributes. Seven classification algorithms—Logistic Regression, K-Nearest Neighbors, Decision Tree, Random Forest, Support Vector Machine, Gaussian Naive Bayes, and Gradient Boosting—were trained and compared under a common experimental procedure. The records were divided through a stratified 80:20 training–testing split. Numerical features were standardized using statistics calculated from the training data only. The models were assessed using Accuracy, Precision, Recall, F1-score, ROC-AUC, and stratified ten-fold cross-validation. Random Forest achieved the strongest test performance, with 83.6% accuracy, an F1-score of 0.865, and a ROC-AUC of 0.916. The leading features in its importance ranking were chest-pain type, major-vessel count, thalassemia status, and maximum heart rate. The results indicate that machine learning can help identify patterns in clinical data; however, the model should be viewed as an academic decision-support experiment and not as a diagnostic tool. Validation on larger and independent patient groups is necessary before any clinical application can be considered.