Transfer Learning for Loan Recovery Prediction under Distribution Shifts with Heterogeneous Feature Spaces

Paper: arXiv 2604.02832

Authors: Christopher Gerling, Hanqiu Peng, Ying Chen, Stefan Lessmann

Abstract

Accurate forecasting of recovery rates (RR) is central to credit risk management and regulatory capital determination. In many loan portfolios, however, RR modeling is constrained by data scarcity arising from infrequent default events. Transfer learning (TL) offers a promising avenue to mitigate this challenge by exploiting information from related but richer source domains, yet its effectiveness critically depends on the presence and strength of distributional shifts, and on potential heterogeneity between source and target feature spaces. This paper introduces FT-MDN-Transformer, a mixture-density tabular Transformer architecture specifically designed for TL in RR forecasting across heterogeneous feature sets. The model produces both loan-level point estimates and portfolio-level predictive distributions, thereby supporting a wide range of practical RR forecasting applications. We evaluate the proposed approach in a controlled Monte Carlo simulation that facilitates systematic variation of covariate, conditional, and label shifts, as well as in a real-world transfer setting using the Global Credit Data (GCD) loan dataset as source and a novel bonds dataset as target. Our results show that FT-MDN-Transformer outperforms baseline models when target-domain data are limited, with particularly pronounced gains under covariate and conditional shifts, while label shift remains challenging. We also observe its probabilistic forecasts to closely track empirical recovery distributions, providing richer information than conventional point-prediction metrics alone. Overall, the findings highlight the potential of distribution-aware TL architectures to improve RR forecasting in data-scarce credit portfolios and offer practical insights for risk managers operating under heterogeneous data environments.

Complexity vs Empirical Score

• Math Complexity: 6.5/10
• Empirical Rigor: 7.5/10
• Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces an innovative architecture (FT-MDN-Transformer) to address transfer learning under heterogeneous feature spaces and distribution shifts in credit risk. It backs up its theoretical design with both controlled Monte Carlo simulations and real-world empirical validation on loan and bond datasets.

Research Flowchart

  flowchart TD
    A["Research Goal: Transfer Learning for Loan Recovery Rate (RR) Forecasting under Distribution Shifts & Heterogeneous Features"] --> B["Data Inputs: Simulated Monte Carlo Data & Real-World Datasets (GCD Loans + Bonds Target)"]
    B --> C["Methodology: FT-MDN-Transformer Architecture (Tabular Transformer + Mixture Density Network)"]
    C --> D["Computational Process: Evaluate under Covariate, Conditional, & Label Shifts"]
    D --> E["Key Findings: Outperforms baselines in data-scarce targets; robust to covariate/conditional shifts; provides accurate portfolio probabilistic distributions"]

Cited References (40)

1. Modelling Recovery Rates for Non-Performing Loans (2019) — link
2. Revisiting Deep Learning Models for Tabular Data (2021)
3. Predicting bank loan recovery rates with a mixed continuous-discrete model (2014) — link
4. Heterogeneities among credit risk parameter distributions: the modality defines the best estimation method (2023) — link
5. Weighted Quantile Regression Forests for Bimodal Distribution Modeling: A Loss Given Default Case (2020) — link
6. Local logit regression for loan recovery rate (2021) — link
7. Recovery rates: Uncertainty certainly matters (2019) — link
8. Credit Risk Modeling Using Transfer Learning and Domain Adaptation (2022) — link

0 of 40 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.