Classification of imbalanced data using support vector machine and rough set theory: A review

Abstract

The performance of machine learning classifier such as support vector machine (SVM) degraded by the nature and structural construct of real-world data which is in most cases are imbalanced. The accuracy and decision making typically biased towards majority class and this significantly affect the result of the classification of minority class. Nevertheless, dataset does not always comprise of significant attributes even with large number of points in certain class, but rather it could potentially lead to redundancy and irrelevant features. Rough set (RS) theory is a mathematical tool for tackling ambiguity and removing redundancy in the dataset. This can further help the classification system in improving its accuracy of the prediction for both majority and minority class. Commonly, RS theory was utilised as a preprocessing method to bring about the knowledge, association rules, or potential patterns in the data. The output of RS theory is a reduced set of attributes which contains same indiscernibility as the original dataset. Hence, the focus of this paper is a review of literature and findings on the classification strategy which employs SVM and RS as a combined system to solve the problem of imbalanced data.