Class Imbalance Problems in Machine Learning: A Review of Methods And Future Challenges

Nowadays, class imbalance problem is one of the most important affairs among machine learning and data mining researchers. In this problem, majority of the sample data are labeled as one class and only a few samples are labeled as another class. This creates an imbalance in the dataset, and it is hard to solve because common classification algorithms are not designed to face these sorts of data. It means, the machine shows a result that is more influenced by the majority class, while the minority class is neglected. In this review paper we explore the main issues that come with the class imbalance problem, we also discuss different techniques to handle this problem. A fair knowledge about the multiclass imbalance problem is also given. In the end, we indicate some opportunities and challenges for future research in this domain.