On Class Imbalance Correction for Classification Algorithms in Credit Scoring

Credit scoring is often modeled as a binary classification task where defaults rarely occur and the classes generally are highly unbalanced. Although many new algorithms have been proposed in the recent past to mitigate this specific problem, the aspect of class imbalance is still underrepresented in research despite its great relevance for many business applications. Within the “Machine Learning in R” (mlr) framework methods for imbalance correction are readily available and can be integrated into a systematic classifier optimization process. Different strategies are discussed, extended and compared.