University Publications

Learning from Imbalanced Multi-class datasets is a challengeable problem that exists in a wide variety of real-world applications. Meanwhile, the imbalance problem for binary class datasets has been well surveyed and studied, Imbalanced Multi-class datasets have received less attention. The Imbalanced Multi-class problem belongs to supervised machine learning tasks where each instance should be assigned to one of N different classes with unequal sample sizes. It owns inherent complex characteristics that introduce more obstacles and issues to be considered during the learning process and require new principles, algorithms, and more tools. In this paper, we provide a review of the development of research in learning from Imbalanced Multi-class datasets. Our aim is at providing a critical review that involves an analysis of the problem notion, the state-of-the-art approaches, structured solutions and the current performance evaluation metrics of the Imbalanced Multi-class learning algorithms as well. Furthermore, we highlight the major challenges in this field.