Cluster-based oversampling with area extraction from representative points for class imbalance learning

Zakarya Farou, Yizhi Wang, Tomás Horváth

Abstract: Highlights•An adaptive cluster-based oversampling approach to address class imbalance challenges.•Optimized clustering: Cophenetic Correlation & Bayesian Criteria for area identification.•Efficiently capturing the underlying data distribution for the resampling process.•An incremental k-Nearest Neighbor strategy for safe and half-safe areas extraction.•A truncated hypercube Gaussian generator for even, precise synthetic sample generation.