A Comprehensive Review on Class Imbalance Problem

Classification of imbalanced data distribution using the standard learning algorithms which assume a relatively equal misclassification costs and relatively balanced underlying class distribution has encountered a serious drawbacks. This paper presents a comprehensive review of learning from Class imbalanced data. Our aim is to provide a review of the class imbalance problem, the state-of-art techniques and the performance measurement metrics used for evaluation under class imbalance scenario. Class imbalance problem in presence of multiple classes is also discussed.