Related Experiment Videos
Heart Disease Prediction Using Statistical Feature Selection and Interpretable Machine Learning
Ramadan Babers1, Ahmed M M Madbouly2
1Faculty of Computer Studies, Arab Open University.
Journal of Visualized Experiments : Jove
|June 22, 2026
Summary
This study introduces an integrated framework for predicting heart disease, tackling data scarcity with generative adversarial networks (GANs) and enhancing accuracy through advanced feature selection and explainable AI. The model offers robust and interpretable predictions for clinical use.
Area of Science:
- Cardiology
- Computational Biology
- Artificial Intelligence
Background:
- Heart disease is a leading global cause of mortality, necessitating advanced predictive tools.
- Existing research often addresses data scarcity, feature selection, and interpretability in isolation.
- An integrated approach is needed to synergistically overcome these predictive challenges.
Purpose of the Study:
- To develop a comprehensive computational framework for early heart disease prediction.
- To integrate generative adversarial networks (GANs), hybrid feature selection, and explainable AI (XAI).
- To enhance predictive accuracy and clinical interpretability in heart disease diagnosis.
Main Methods:
- Utilized a generative adversarial network (GAN) to mitigate class imbalance and data scarcity.
- Implemented a hybrid feature selection combining Welch's t-test, Cohen's d, and Harris Hawk Optimization.
- Employed explainable AI methods (SHAP, PDP, Odds Ratios) for model interpretability.
Main Results:
- The framework demonstrated superior performance on the Cleveland and Statlog heart disease datasets.
- Achieved high accuracy, F1-scores, and ROC-AUC values compared to baseline and existing methods.
- The integrated model provided robust and clinically interpretable predictions.
Conclusions:
- The developed framework offers a synergistic solution for heart disease prediction challenges.
- Combines advanced machine learning techniques with clinical interpretability for practical application.
- Represents a significant advancement in computational approaches to cardiovascular disease diagnosis.