A Method for Evaluating the Interpretability of Machine Learning Models in Predicting Bond Default Risk Based on LIME and SHAP

Paper: arXiv 2502.19615

Authors: Yan Zhang, Lin Chen, Yixiang Tian

Abstract

Interpretability analysis methods for artificial intelligence models, such as LIME and SHAP, are widely used, though they primarily serve as post-model for analyzing model outputs. While it is commonly believed that the transparency and interpretability of AI models diminish as their complexity increases, currently there is no standardized method for assessing the inherent interpretability of the models themselves. This paper uses bond market default prediction as a case study, applying commonly used machine learning algorithms within AI models. First, the classification performance of these algorithms in default prediction is evaluated. Then, leveraging LIME and SHAP to assess the contribution of sample features to prediction outcomes, the paper proposes a novel method for evaluating the interpretability of the models themselves. The results of this analysis are consistent with the intuitive understanding and logical expectations regarding the interpretability of these models.

Complexity vs Empirical Score

• Math Complexity: 3.0/10
• Empirical Rigor: 5.5/10
• Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper proposes a pragmatic heuristic metric (MIAI) by computing the cosine similarity between LIME and SHAP feature attribution vectors to assess intrinsic model interpretability. While the theoretical math is light and relies on standard definitions of LIME and SHAP, the application to real-world bond default data offers a practical empirical framework.

Research Flowchart

  flowchart TD
    A["Research Goal:<br>Evaluate inherent interpretability of ML models"] --> B["Data Input:<br>Bond Market Default Data"]
    B --> C["Train ML Classifiers<br>(Logistic Regression, RF, XGBoost, etc.)"]
    C --> D["Evaluate Classification Performance"]
    C --> E["Apply LIME & SHAP to calculate<br>Feature Contributions"]
    E --> F["Propose Novel Model-Level<br>Interpretability Evaluation Method"]
    D --> G["Key Outcomes:<br>Quantified interpretability aligns with<br>intuitive model complexity"]
    F --> G