Probabilistic Rule Models as Diagnostic Layers: Interpreting Structural Concept Drift in Post-Crisis Finance

2025-10-30

AI Summary

Concept drift in credit risk prediction degrades model performance over time, particularly during economic crises. Method: This paper proposes an interpretable correction layer architecture based on Probabilistic Rule Models (PRMs), integrating Markov Logic Networks into the model backend to explicitly encode symbolic rules that capture systematic shifts in predicted risk scores—enabling precise diagnosis and attribution of evolving borrower risk structures. Contribution/Results: Unlike opaque adaptive methods, the architecture ensures full auditability and transparency, facilitating post-crisis identification of risk evolution mechanisms for specific demographic cohorts. Experiments on Fannie Mae mortgage data—spanning pre- and post-2008 financial crisis periods—demonstrate that the extracted rules clearly expose heterogeneous risk sensitivities across borrower groups and their structural transformations. The approach significantly enhances model interpretability, explainability, and regulatory compliance—especially in high-risk scenarios.

Technology Category

• Machine Learning: Transparent, Interpretable, Explainable ML
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Machine learning models used for high-stakes predictions in domains like credit risk face critical degradation due to concept drift, requiring robust and transparent adaptation mechanisms. We propose an architecture, where a dedicated correction layer is employed to efficiently capture systematic shifts in predictive scores when a model becomes outdated. The key element of this architecture is the design of a correction layer using Probabilistic Rule Models (PRMs) based on Markov Logic Networks, which guarantees intrinsic interpretability through symbolic, auditable rules. This structure transforms the correction layer from a simple scoring mechanism into a powerful diagnostic tool capable of isolating and explaining the fundamental changes in borrower riskiness. We illustrate this diagnostic capability using Fannie Mae mortgage data, demonstrating how the interpretable rules extracted by the correction layer successfully explain the structural impact of the 2008 financial crisis on specific population segments, providing essential insights for portfolio risk management and regulatory compliance.

Problem

• Detecting and interpreting concept drift in financial risk prediction models
• Providing transparent adaptation mechanisms for outdated machine learning models
• Explaining structural changes in borrower riskiness using interpretable rules

Innovation

• Correction layer using Probabilistic Rule Models
• Markov Logic Networks for interpretable rules
• Diagnostic tool explaining structural concept drift