Title: Grabit: Gradient tree-boosted Tobit models for default prediction
Authors: Fabio Sigrist - Lucerne University of Applied Sciences (Switzerland) [presenting]
Abstract: A frequent problem in binary classification, and in particular in default prediction, is class imbalance between a minority and a majority class such as defaults and non-defaults. We show how this issue can be alleviated by using a tree-boosted Tobit model in cases where there is auxiliary data for the non-default events that is related to the default mechanism. For instance, such auxiliary data can consist of number of days of delay by which loans were paid back, stock returns, rating changes, or distance to default measures. We apply our proposed model for predicting defaults on loans made to Swiss small and medium-sized enterprises and obtain a large improvement in predictive performance compared to other state-of-the-art approaches.