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

Kamesh Korangi, Christophe Mues, Cristián Bravo

Abstract: Highlights•We predict a multi-horizon probability of default for mid-caps using deep learning.•We create a transformer-based model for panel-data multi-label classification.•We use a multimodal architecture for efficient training over different data sources.•We use a Shapley-based methodology for ranking of groups of features.