Rough-Fuzzy Based Synthetic Data Generation Exploring Boundary Region of Rough Sets to Handle Class Imbalance Problem

Class imbalance is a prevalent problem that not only reduces the performance of machine learning techniques but also causes lacking inherent complex characteristics data. Though researchers have proposed various ways to deal with problem, they yet consider how select proper treatment, especially when uncertainty levels are high. Applying rough-fuzzy theory imbalanced data could be promising research direction generates synthetic and removes outliers. The work identifies positive, boundary, negative regions target set using rough objects in region as It explores positive boundary by applying fuzzy generate samples minority class remove majority class. Thus approach performs both oversampling undersampling handle problem. experimental results demonstrate novel technique allows qualitative quantitative handling.