EXPLAINABLE PREDICTION OF HEART DISEASE USING CLINICAL INDICATORS

Abstract

Heart disease remains a leading cause of morbidity and mortality worldwide, highlighting the need for accurate and transparent prediction models that can support timely clinical decision-making. Although machine learning has significantly improved the prediction of cardiovascular disease, the limited interpretability of many predictive models restricts their clinical adoption. This study proposes an explainable machine learning framework for predicting heart disease using routinely collected clinical indicators from the UCI Heart Disease dataset comprising 303 patient records and 13 predictor variables. Following data preprocessing, five supervised machine learning algorithms, including Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, and Extreme Gradient Boosting (XGBoost), were developed and comparatively evaluated. Model performance was assessed using accuracy, precision, recall, F1-score, and receiver operating characteristic area under the curve (ROC–AUC). Explainability was incorporated through SHapley Additive exPlanations (SHAP) to quantify the contribution of individual clinical variables to prediction outcomes. Among the evaluated models, XGBoost achieved the highest predictive performance with an accuracy of 92% and an ROC–AUC of 0.96. SHAP analysis identified chest pain type, number of major vessels, ST depression, maximum heart rate, age, and thalassemia as the most influential predictors of heart disease, providing clinically meaningful explanations for model decisions. The findings demonstrate that integrating explainable artificial intelligence with machine learning can improve both predictive performance and model transparency, thereby supporting trustworthy clinical decision support and facilitating early cardiovascular risk assessment. Future research should validate the proposed framework using larger multicentre datasets and real-world clinical data.