A cardiovascular disease prediction method based on cross-combination strategy and dynamic weighted stacking ensemble

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Computational Biology

Background:

• Cardiovascular disease (CVD) remains a leading global cause of mortality.
• Early and accurate CVD prediction is critical for effective intervention and mortality reduction.
• Existing models face challenges in feature association mining and generalization.

Purpose of the Study:

• To develop an advanced Cardiovascular Disease (CVD) prediction method.
• To address limitations in feature association and model generalization in current CVD prediction tools.
• To enhance the accuracy and reliability of CVD risk assessment.

Main Methods:

• Proposes a Cardiovascular Disease (CVD) prediction method using an ensemble of cross-combination strategy and dynamic weighted stacking ensemble (CCS-DWSE).
• Constructs heterogeneous base models via multi-feature selection and multi-paradigm classifiers.
• Employs a dynamic weighted stacking ensemble framework with real-time weight adjustment and a K-Nearest Neighbors (KNN) meta-model for adaptive fusion.
• Utilizes SHAP for feature interpretability analysis.

Main Results:

• Achieved high prediction accuracies of 98.32%, 92.33%, and 82.10% across three public datasets.
• Attained an Area Under the Curve (AUC) of up to 0.9992.
• Demonstrated a 2.76% performance improvement over existing state-of-the-art models.

Conclusions:

• The CCS-DWSE method offers an efficient and scalable solution for Cardiovascular Disease (CVD) prediction.
• Presents a novel approach for collaborative multi-model learning on complex medical data.
• Highlights the potential of dynamic ensemble methods for improving diagnostic accuracy in healthcare.