Measuring Name Concentrations through Deep Learning

2024/03/25 by Eva Lütkebohmert, Lütkebohmert, Eva, Julian Sester +1
Medicine · Computer Science · #Data-Driven Disease Surveillance #Anomaly Detection Techniques and Applications #Technology and Data Analysis

paper · pdf · doi:10.48550/arxiv.2403.16525

Abstract

We propose a new deep learning approach for the quantification of name concentration risk in loan portfolios. Our approach is tailored for small portfolios and allows for both an actuarial as well as a mark-to-market definition of loss. The training of our neural network relies on Monte Carlo simulations with importance sampling which we explicitly formulate for the CreditRisk+ and the ratings-based CreditMetrics model. Numerical results based on simulated as well as real data demonstrate the accuracy of our new approach and its superior performance compared to existing analytical methods for assessing name concentration risk in small and concentrated portfolios.

Related

• On the Relevance and Appropriateness of Name Concentration Risk Adjustments for Portfolios of Multilateral Development Banks
• Towards Explainable Deep Learning for Credit Lending: A Case Study
• Minimizing the Value-at-Risk of Loan Portfolio via Deep Neural Networks
• Universal approximation of credit portfolio losses using Restricted Boltzmann Machines
• DeRisk: An Effective Deep Learning Framework for Credit Risk Prediction over Real-World Financial Data