Generative adversarial networks and hyperparameter-optimized XGBoost for enhanced heart disease prediction.

Area of Science:

• Cardiology
• Artificial Intelligence
• Data Science

Background:

• Heart disease (HD) is a leading global cause of mortality, necessitating advanced diagnostic tools.
• Clinical data often suffers from bias and noise, hindering the development of reliable predictive models for early HD detection.

Purpose of the Study:

• To propose and evaluate a novel hybrid framework, GAN-XO, for early detection of heart disease.
• To address data imbalance and noise in clinical datasets for improved predictive model performance.

Main Methods:

• Utilized a Generative Adversarial Network (GAN) for synthetic oversampling to balance imbalanced datasets.
• Implemented a two-phase outlier detection and elimination process (z-score and IQR) for data cleaning.
• Employed an XGBoost classifier with hyperparameter optimization via Optuna for accurate prediction.

Main Results:

• The GAN-XO system achieved a high accuracy of 96.60% and an F1-Score of 0.9649 on a clean, balanced test dataset.
• Ablation studies confirmed the contribution of each component, highlighting the critical role of outlier elimination.
• The framework demonstrated significant performance enhancement, improving trustworthiness and model reliability.

Conclusions:

• Accurate healthcare predictions depend on integrated modeling systems and rigorous data quality management.
• The GAN-XO framework offers a promising approach to enhance clinical decision-making for early heart disease diagnosis.