Harnessing Machine Learning for Improved Heart Disease Prediction

Heart disease is one of the most serious global health problems. Diagnosis or detection during the early stages of this condition is recommended to enhance patient outcomes. This article reports on a novel approach, referred to as Principal Component Heart Failure (PCHF), for feature selection on multiple machine learning algorithms as applied in heart disease prediction. The data sourced from Kaggle contained 1,025 patient records and were preprocessed for optimal model training. Therefore, using PCHF we found eight most essential features that maximize the prediction accuracy across nine ML algorithms, such as Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGB). Both the models of Random Forest and Decision Tree provided very high accuracy at 98 %, which were given with precision and recall values at <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 9 \%}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 8 \%}$</tex>, respectively, signifying high reliability for applications in clinical decision. Cross-validation and hyperparameter tuning further refine model performance, hence unfolding what PCHF is able to do to propel predictive models forward well in time toward early intervention for heart disease. It has the promise of use in offering real-time diagnostic support on clinical premises.