Imbalanced Learning with Oversampling based on Classification Contribution Degree

• DOI: 10.1002/adts.202100031
• Publication Date: 2021-03-26T13:21:04Z
• Authors: Zhenhao Jiang; Jie Yang; Yan Liu
• Keywords: Oversampling; Degree (music); Benchmark (surveying); Statistical classification

Abstract Imbalanced datasets exist commonly in the real world, which leads to poor performance of general machine learning models because skewed class distribution. To address data‐imbalance problem, a novel oversampling method based on classification contribution degree, called OS‐CCD is presented. First new concept, established micro and macro information extracted from raw datasets. With enables positive samples near boundary located an area with high density generate more synthetic than others. Furthermore, neighbor selection for no longer random but light selected probability. Experimental results 12 benchmark substantiate that four used classifiers outperform those six popular methods terms accuracy, F1‐score AUC.