An Explainable Machine Learning Framework for Early Cardiovascular Disease Risk Prediction Using Routine Clinical Indicators

Abstract

Cardiovascular disease (CVD) remains one of the leading causes of morbidity and mortality worldwide, emphasizing the need for accurate, transparent, and clinically interpretable prediction models to support early diagnosis and intervention. This study developed an explainable machine learning framework for early cardiovascular disease risk prediction using routinely collected clinical indicators. A publicly available dataset comprising 1,000 patient records and 14 clinical variables was analyzed following comprehensive data quality assessment and preprocessing. Five supervised machine learning algorithms, namely Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, and XGBoost, were optimized using five-fold cross-validation with GridSearchCV and evaluated on an independent testing dataset. Model performance was assessed using accuracy, precision, recall, F1-score, and receiver operating characteristic area under the curve (ROC–AUC). XGBoost achieved the highest predictive performance, with an accuracy of 99.5%, precision of 99.15%, recall of 100.0%, F1-score of 99.57%, and ROC–AUC of 0.9994. To improve model transparency, SHAP (SHapley Additive exPlanations) analysis was employed to quantify feature contributions, identifying the slope of the ST segment, resting blood pressure, chest pain type, and oldpeak as the most influential predictors. The proposed framework combines excellent predictive accuracy with robust interpretability, demonstrating its potential to support trustworthy clinical decision-making, facilitate early cardiovascular disease screening, and promote the practical adoption of explainable artificial intelligence in precision cardiovascular medicine.