Adaptive Ensemble Framework With Synthetic Sampling for Tackling Class Imbalance Problem

Abstract

Class imbalance is a critical challenge in heart disease prediction datasets, often leading to biased models with poor performance on the minority class. This paper proposes a novel algorithm, the Adaptive Synthetic Ensemble Balancer (ASEB), designed to address this issue through a combination of advanced synthetic data generation, adaptive class weighting, and ensemble learning techniques. ASEB integrates Adaptive Synthetic Sampling (ADASYN) and Generative Adversarial Networks (GANs) for synthetic sample generation, dynamic class weighting, and ensemble learning strategies. Experimental results demonstrate ASEB's effectiveness in improving predictive accuracy and robustness in imbalanced heart disease datasets.