An Integrated Ensemble Learning Model for Multi Class Imbalanced Datasets Classification

Abstract

Due to arrival of big data era, many real world datasets generated from various data sources are not evenly distributed which create class imbalance problem. In spite of many years of research on imbalanced data, it is still one of the most challenging research problems in big data analytics, machine learning and data mining. Different techniques for binary class classification are relatively more and advanced nowadays, yet multi class classification problem is still an open research problem. This paper addresses multi class imbalance learning by proposing a hybrid ensemble algorithm which uses fuzzy relation strategy along with diversified error correcting output code to improve performance of the algorithm. Additionally this paper presents analysis of various algorithms along with proposed method on various datasets and we claim that proposed algorithm is a superior and reliable approach in terms of F-measure, G-mean and accuracy and AUC.