Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Explaining Predictions by Characteristic Rules

Amr Alkhatib, Henrik Boström, Michalis Vazirgiannis

TL;DR

The paper addresses the need for verifiable, compact explanations for black-box predictions by introducing CEGA, a method that aggregates local explanations into general characteristic rules via association-rule mining. CEGA is explainer-agnostic and can generate both characteristic and, with a simple orientation change, discriminative rules, enabling flexible explanations with quantified fidelity. Empirical evaluation across 20 binary datasets shows CEGA achieves fidelity comparable to Anchors while producing far fewer rules, and generally outperforms GLocalX in fidelity; its advantage grows when using SHAP or Anchors as local explainers. The work demonstrates that characteristic-rule explanations are viable and often more interpretable, with clear paths for future work in user studies, combining multiple explanation sources, and uncertainty quantification.Overall, CEGA provides a practical, scalable approach to producing faithful, compact explanations that can enhance trust and verifiability in high-stakes domains.

Abstract

Characteristic rules have been advocated for their ability to improve interpretability over discriminative rules within the area of rule learning. However, the former type of rule has not yet been used by techniques for explaining predictions. A novel explanation technique, called CEGA (Characteristic Explanatory General Association rules), is proposed, which employs association rule mining to aggregate multiple explanations generated by any standard local explanation technique into a set of characteristic rules. An empirical investigation is presented, in which CEGA is compared to two state-of-the-art methods, Anchors and GLocalX, for producing local and aggregated explanations in the form of discriminative rules. The results suggest that the proposed approach provides a better trade-off between fidelity and complexity compared to the two state-of-the-art approaches; CEGA and Anchors significantly outperform GLocalX with respect to fidelity, while CEGA and GLocalX significantly outperform Anchors with respect to the number of generated rules. The effect of changing the format of the explanations of CEGA to discriminative rules and using LIME and SHAP as local explanation techniques instead of Anchors are also investigated. The results show that the characteristic explanatory rules still compete favorably with rules in the standard discriminative format. The results also indicate that using CEGA in combination with either SHAP or Anchors consistently leads to a higher fidelity compared to using LIME as the local explanation technique.

Explaining Predictions by Characteristic Rules

TL;DR

The paper addresses the need for verifiable, compact explanations for black-box predictions by introducing CEGA, a method that aggregates local explanations into general characteristic rules via association-rule mining. CEGA is explainer-agnostic and can generate both characteristic and, with a simple orientation change, discriminative rules, enabling flexible explanations with quantified fidelity. Empirical evaluation across 20 binary datasets shows CEGA achieves fidelity comparable to Anchors while producing far fewer rules, and generally outperforms GLocalX in fidelity; its advantage grows when using SHAP or Anchors as local explainers. The work demonstrates that characteristic-rule explanations are viable and often more interpretable, with clear paths for future work in user studies, combining multiple explanation sources, and uncertainty quantification.Overall, CEGA provides a practical, scalable approach to producing faithful, compact explanations that can enhance trust and verifiability in high-stakes domains.

Abstract

Characteristic rules have been advocated for their ability to improve interpretability over discriminative rules within the area of rule learning. However, the former type of rule has not yet been used by techniques for explaining predictions. A novel explanation technique, called CEGA (Characteristic Explanatory General Association rules), is proposed, which employs association rule mining to aggregate multiple explanations generated by any standard local explanation technique into a set of characteristic rules. An empirical investigation is presented, in which CEGA is compared to two state-of-the-art methods, Anchors and GLocalX, for producing local and aggregated explanations in the form of discriminative rules. The results suggest that the proposed approach provides a better trade-off between fidelity and complexity compared to the two state-of-the-art approaches; CEGA and Anchors significantly outperform GLocalX with respect to fidelity, while CEGA and GLocalX significantly outperform Anchors with respect to the number of generated rules. The effect of changing the format of the explanations of CEGA to discriminative rules and using LIME and SHAP as local explanation techniques instead of Anchors are also investigated. The results show that the characteristic explanatory rules still compete favorably with rules in the standard discriminative format. The results also indicate that using CEGA in combination with either SHAP or Anchors consistently leads to a higher fidelity compared to using LIME as the local explanation technique.
Paper Structure (12 sections, 2 equations, 3 figures, 5 tables, 1 algorithm)

This paper contains 12 sections, 2 equations, 3 figures, 5 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. From Local Explanations to General Characteristic Rules
  4. Explanation Mining and Rules Selection
  5. Discriminative vs. Characteristic Rules
  6. Empirical Evaluation
  7. Experimental Setup
  8. Baseline Experiments
  9. Comparing Discriminative and Characteristic Rules
  10. Comparing Local Explanation Techniques
  11. Concluding remarks
  12. Acknowledgement

Figures (3)

  • Figure 1: Discriminative rules distinguish one class from the others using a few features, while characteristic rules learn all the features that characterize the class.
  • Figure 2: Comparing average ranks with respect to fidelity measured by AUC (lower rank is better) of Anchors, GLocalX, and CEGA, where the critical difference (CD) represents the largest difference that is not statistically significant.
  • Figure 3: Comparing average ranks of Anchors, GLocalX, and CEGA with respect to the number of rules