A transformer-based model for default prediction in mid-cap corporate markets

In this paper, we study mid-cap companies, i.e. publicly traded companies with less than US$10 billion in market capitalisation. Using a large dataset of US mid-cap companies observed over 30 years, we look to predict the default probability term structure over the short to medium term and understand which data sources (i.e. fundamental, market or pricing data) contribute most to the default risk. Whereas existing methods typically require that data from different time periods are first aggregated and turned into cross-sectional features, we frame the problem as a multi-label panel data classification problem. To tackle it, we then employ transformer models, a state-of-the-art deep learning model emanating from the natural language processing domain. To make this approach suitable to the given credit risk setting, we use a loss function for multi-label classification, to deal with the term structure, and propose a multi-channel architecture with differential training that allows the model to use all input data efficiently. Our results show that the proposed deep learning architecture produces superior performance, resulting in a sizeable improvement in AUC (Area Under the receiver operating characteristic Curve) over traditional models. In order to interpret the model, we also demonstrate how to produce an importance ranking for the different data sources and their temporal relationships, using a Shapley approach for feature groups.