On the class overlap problem in imbalanced data classification

Class imbalance is an active research area in the machine learning community. However, existing and recent literature showed that class overlap had a higher negative impact on performance of algorithms. This paper provides detailed critical discussion objective evaluation context imbalanced data its classification accuracy. First, we present thorough experimental comparison imbalance. Unlike previous work, our experiment was carried out full scale extreme range degrees. Second, provide in-depth technical review approaches to handle datasets. Existing solutions from selective are critically reviewed categorised as distribution-based overlap-based methods. Emerging techniques latest development this also discussed detail. Experimental results consistent with show clearly algorithm deteriorates across varying degrees whereas does not always have effect. The emphasises need for further towards handling datasets effectively improve algorithms’ performance.