Assessing the Loss Given Default of Bank Loans Using the Hybrid Algorithms Multi-Stage Model

Abstract

The loss given default (LGD) is an important credit risk parameter in the regulatory system for financial institutions. Due to the complex structure of the LGD distribution, we propose a new approach, called the hybrid algorithms multi-stage (HMS) model, to construct a multi-stage LGD prediction model and test it on the US Small Business Administration (SBA)’s small business credit dataset. We then compare the model’s performance under four routes by different evaluation metrics. Finally, pertinent business information and macroeconomic features datasets are added for robustness validation. The results show that HMS performs well and stably for predicting LGD, confirming the superiority of the proposed hybrid unsupervised and supervised machine learning algorithm. Financial institutions can apply the approach to make default predictions based on other credit datasets.

Keywords

• loss given default prediction
• credit risk
• unsupervised machine learning
• supervised machine learning
• multi-stage model