The authors employ signaling theory and machine learning methods to investigate loss given default predictions of commercial banks and propose a method to