Shaik Sajeera Begum
School of Technology, Woxsen University, Hyderabad, Telangana, India.

Sanjay Dhanka
Centre for Internet of Things, MITS-DU, Gwalior, Madhya Pradesh, India.

Syed Javeed Pasha
School of Technology, Woxsen University, Hyderabad, Telangana, India.

DOI https://doi.org/10.33889/IJMEMS.2026.11.5.098

  ;

Abstract

Heart disease (HD) is a major cause of death globally. Early and accurate prediction is crucial for effective clinical decision making. However, the available datasets in the healthcare domain are prone to data noise, structural outliers, and data imbalance, which adversely affect the performance of traditional machine learning (ML) models. To overcome these problems, the current study presents a novel approach called ROHG-BXG (Robust Outlier Handling with Generative Adversarial Network and Butterfly Optimization Algorithm-Optimized XGBoost) for HD prediction problem in the Personal Key Indicators of Heart Disease (PKIOHD) dataset and Framingham dataset. This proposed scheme includes a structural outlier detection algorithm (Isolation Forest), a loss-based on-the-fly outlier detection algorithm to eliminate unreliable samples (Outlier Reduction), a class balancing algorithm (Conditional Generative Adversarial Networks (CGAN)) and an optimization algorithm for the hyperparameters of the XGBoost algorithm (Butterfly Optimization Algorithm (BOA)). The experimental results achieved 92% and 93% accuracy for both datasets. The framework was also shown to have an excellent Brier Score of 0.0655 and 0.0391and an excellent Expected Calibration Error (ECE) of 0.032 and 0.0217 for both datasets. The outcomes of this work show the proposed ROHG-BXG framework is a powerful and reliable tool to predict HD and can be utilized in intelligent clinical decision-support systems.

Keywords- Butterfly optimization algorithm, Generative adversarial network, Heart disease prediction, Isolation forest, On-the-fly outlier detection, Robust outlier handling, XGBoost.

Citation

Begum, S. S., Dhanka, S., & Pasha, S. J. (2026). ROHG-BXG: Robust Outlier Handling with GAN and BOA-Optimized XGBoost for Heart Disease Prediction. International Journal of Mathematical, Engineering and Management Sciences, 11(5), 2421-2442. https://doi.org/10.33889/IJMEMS.2026.11.5.098.