Abstract:
In recent years, machine learning techniques have been widely applied for credit rating. To make a rational comparison of performance of different learning-based credit rating models, we focused on those models that are constructed and validated on the two mostly used Australian and German credit approval data sets. Based on a systematic review of literatures, we further compare and discuss about the performance of existing models. In addition, we identified and illustrated the limitations of existing works and discuss about some open issues that could benefit future research in this area.