UQ for Credit Risk Management: A deep evidence regression approach

Paper: arXiv 2305.04967

Authors: Ashish Dhiman

Abstract

Machine Learning has invariantly found its way into various Credit Risk applications. Due to the intrinsic nature of Credit Risk, quantifying the uncertainty of the predicted risk metrics is essential, and applying uncertainty-aware deep learning models to credit risk settings can be very helpful. In this work, we have explored the application of a scalable UQ-aware deep learning technique, Deep Evidence Regression and applied it to predicting Loss Given Default. We contribute to the literature by extending the Deep Evidence Regression methodology to learning target variables generated by a Weibull process and provide the relevant learning framework. We demonstrate the application of our approach to both simulated and real-world data.

Complexity vs Empirical Score

• Math Complexity: 6.5/10
• Empirical Rigor: 6.5/10
• Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper provides a strong theoretical contribution by extending Deep Evidential Regression to Weibull-distributed targets, specifically tailored for Loss Given Default (LGD) modeling. It balances solid mathematical formulation of epistemic and aleatoric uncertainty with practical empirical validation on both simulated and real-world credit datasets.

Research Flowchart

  flowchart TD
    A["Research Goal: UQ in Credit Risk<br>for Loss Given Default (LGD)"] --> B["Develop Deep Evidential Regression<br>extended for Weibull Process"]
    
    C["Inputs: Simulated & Real-World<br>Credit Risk Datasets"] --> D["Train Deep Learning Model<br>with Evidential Loss Function"]
    B --> D
    
    D --> E["Quantify Aleatoric & Epistemic<br>Uncertainty on LGD"]
    
    E --> F["Outcomes: Scalable UQ Framework<br>& Improved Risk Prediction"]