Transforming Heart Disease Prediction with Advanced Machine Learning Techniques

Heart disease remains the leading cause of global mortality, requiring accurate early detection tools to improve patient prognosis. This study evaluates and compares the performance of multiple supervised machine learning classifiers for heart disease prediction. Using two distinct benchmark datasets from the UCI and Kaggle repositories—each comprising 14 clinical attributes—the authors analyze algorithms such as J48, Naive Bayes, Logistic Regression, Simple Cart, Bagging, Decision Stump, AdaBoost, Artificial Neural Networks, and Support Vector Machines.

The research employs a comprehensive methodology involving data preprocessing and rigorous 10-fold cross-validation. Each dataset is divided into ten subsets, rotating validation across folds to minimize overfitting and reduce evaluation bias. Performance is quantified using metrics including accuracy, precision, recall, F-measure, Mean Absolute Error, and Relative Absolute Error to identify classifiers with superior predictive power and minimal error rates.

Experimental results demonstrate that algorithm effectiveness varies by data source. Support Vector Machine achieved the highest performance metrics on the UCI repository dataset. In contrast, the Simple Cart model delivered the highest accuracy and lowest error rates on the Kaggle repository dataset. These variations underscore the importance of dataset characteristics in model selection and tuning for medical diagnostics.

The findings indicate that properly tuned and validated machine learning models can serve as reliable decision-support tools for healthcare professionals. By facilitating early and accurate identification of cardiovascular risks, these automated diagnostic frameworks can assist clinicians in formulating timely interventions and treatment strategies.