Developing a novel framework using optimized active stacking and explainable AI for heart disease prediction.

Area of Science:

• Cardiovascular disease research
• Machine learning in healthcare
• Biomedical data analysis

Background:

• Heart disease remains a leading global cause of mortality.
• Accurate, interpretable, and efficient predictive systems are crucial for early diagnosis and intervention.
• Existing machine learning models face challenges like class imbalance, high dimensionality, and limited labeled data.

Purpose of the Study:

• To develop a robust componental framework for heart disease prediction.
• To overcome limitations of current machine learning approaches in cardiovascular risk assessment.
• To enhance classification accuracy, interpretability, and efficiency in predictive modeling.

Main Methods:

• Proximity-weighted random affine shadow sampling to address class imbalance.
• Principal Component Analysis (PCA) for feature dimensionality reduction.
• Novel models including stacking, Optimized Stacking Model with Bayesian Optimization (OSM-BO), and Entropy-based Active Learning Optimized Stacking Model (EAL-OSM).

Main Results:

• The stacking model improved key metrics including accuracy and Precision-Recall Area Under the Curve (PR-AUC).
• OSM-BO further enhanced performance with significant gains in accuracy, precision, recall, and PR-AUC.
• EAL-OSM achieved the highest improvements, demonstrating substantial increases in accuracy, precision, recall, and PR-AUC, alongside a notable reduction in Hamming loss.

Conclusions:

• The proposed componental framework offers significant improvements in classification performance for heart disease prediction.
• The framework demonstrates statistical robustness and enhanced explainability.
• This approach provides a clinically practical solution for early and accurate heart disease diagnosis.