Stacked Ensemble Model With Explainable AI for Early Detection of Heart Disease.

Area of Science:

• Cardiology
• Artificial Intelligence
• Machine Learning

Background:

• Heart disease (HD) remains a leading global cause of mortality.
• Early detection of HD is challenging due to non-specific symptoms and unclear predictive models.

Purpose of the Study:

• To develop a highly accurate and interpretable machine learning model for early heart disease detection.
• To enhance clinical decision support systems for heart disease risk assessment.

Main Methods:

• A two-layer stacked ensemble model was developed, combining Support Vector Machine, K-Nearest Neighbors, Naïve Bayes, and Decision Tree base learners with Logistic Regression as a meta-learner.
• The ensemble was trained on a balanced sample from five public heart disease datasets.
• Global and local explainable artificial intelligence (XAI) techniques, including SHAP and LIME, were employed for model interpretability.

Main Results:

• The proposed ensemble model achieved high performance with 93.69% accuracy, 93% F1-score, and 94.7% AUC.
• Explainability methods identified ST slope and chest pain type as key predictors of heart disease.
• The model demonstrated strong predictive power and provided clinically relevant insights through various XAI visualizations.

Conclusions:

• The developed ensemble model offers a robust and interpretable approach to heart disease risk assessment.
• Integrating advanced machine learning with XAI can significantly improve the accuracy and clinical utility of diagnostic tools.
• This framework supports better patient-centered care and auditability in healthcare decision-making.