An explainable hybrid framework for early detection of cardiovascular diseases using Categorical Boosting and Bees...

Area of Science:

• Cardiovascular health
• Machine learning applications
• Artificial intelligence in medicine

Background:

• Cardiovascular disease (CVD) is a leading global cause of mortality.
• Early CVD detection is crucial for effective patient management.
• Traditional machine learning (ML) models for disease prediction often lack transparency (are "black-box").

Purpose of the Study:

• To develop an interpretable ML framework for accurate CVD detection.
• To provide insights into factors influencing CVD occurrence for targeted therapies.
• To enhance clinical decision-making through transparent AI predictions.

Main Methods:

• Utilized the Framingham Cardiovascular Disease (CVD) dataset.
• Applied Random Oversampling (RO) for data balancing and Min-Max scaling for normalization.
• Developed a hybrid ML model combining Categorical Boosting (CatBoost) and BEEs algorithms.
• Implemented Explainable Artificial Intelligence (XAI) techniques (LIME, SHAP) for interpretability.

Main Results:

• The hybrid ML model achieved 98.04% accuracy, 97.09% precision, 98.96% recall, 98.02% F1-score, and 97.16% specificity.
• The model demonstrated superior performance compared to existing state-of-the-art algorithms.
• XAI techniques identified key attributes contributing to CVD occurrence.

Conclusions:

• The proposed hybrid ML model offers a highly accurate and interpretable solution for CVD detection.
• Explainable AI provides valuable insights for healthcare providers, facilitating timely and accurate treatment decisions.
• This framework has the potential to significantly improve cardiovascular disease management and patient outcomes.