Binary Classification: Counterbalancing Class Imbalance by Applying Regression Models in Combination with One-Sided Label Shifts

In many real-world pattern recognition scenarios, such as in medical\napplications, the corresponding classification tasks can be of an imbalanced\nnature. In the current study, we focus on binary, imbalanced classification\ntasks, i.e.~binary classification tasks in which one of the two classes is\nunder-represented (minority class) in comparison to the other class (majority\nclass). In the literature, many different approaches have been proposed, such\nas under- or oversampling, to counter class imbalance. In the current work, we\nintroduce a novel method, which addresses the issues of class imbalance. To\nthis end, we first transfer the binary classification task to an equivalent\nregression task. Subsequently, we generate a set of negative and positive\ntarget labels, such that the corresponding regression task becomes balanced,\nwith respect to the redefined target label set. We evaluate our approach on a\nnumber of publicly available data sets in combination with Support Vector\nMachines. Moreover, we compare our proposed method to one of the most popular\noversampling techniques (SMOTE). Based on the detailed discussion of the\npresented outcomes of our experimental evaluation, we provide promising ideas\nfor future research directions.\n