Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Perspective on Explainable Artificial Intelligence Methods: SHAP and LIME

Ahmed Salih, Zahra Raisi-Estabragh, Ilaria Boscolo Galazzo, Petia Radeva, Steffen E. Petersen, Gloria Menegaz, Karim Lekadir

TL;DR

This perspective assesses SHAP and LIME as two dominant XAI methods for tabular data, examining how their explanations are influenced by the underlying ML model and feature collinearity. Through case studies in biomedical contexts and airline satisfaction data, the authors demonstrate that explanations can vary across models and that collinearity can distort attributions. They propose robustness tools such as Normalized Movement Rate (NMR) and Modified Index Position (MIP), along with local surrogates like shapr and alternatives like GraphLIME, to enhance reliability and interpretability. The work provides practical guidance for deploying XAI in high-stakes domains, emphasizing transparent communication of assumptions and cross-model validation to improve trust and utility of explanations.

Abstract

eXplainable artificial intelligence (XAI) methods have emerged to convert the black box of machine learning (ML) models into a more digestible form. These methods help to communicate how the model works with the aim of making ML models more transparent and increasing the trust of end-users into their output. SHapley Additive exPlanations (SHAP) and Local Interpretable Model Agnostic Explanation (LIME) are two widely used XAI methods, particularly with tabular data. In this perspective piece, we discuss the way the explainability metrics of these two methods are generated and propose a framework for interpretation of their outputs, highlighting their weaknesses and strengths. Specifically, we discuss their outcomes in terms of model-dependency and in the presence of collinearity among the features, relying on a case study from the biomedical domain (classification of individuals with or without myocardial infarction). The results indicate that SHAP and LIME are highly affected by the adopted ML model and feature collinearity, raising a note of caution on their usage and interpretation.

A Perspective on Explainable Artificial Intelligence Methods: SHAP and LIME

TL;DR

This perspective assesses SHAP and LIME as two dominant XAI methods for tabular data, examining how their explanations are influenced by the underlying ML model and feature collinearity. Through case studies in biomedical contexts and airline satisfaction data, the authors demonstrate that explanations can vary across models and that collinearity can distort attributions. They propose robustness tools such as Normalized Movement Rate (NMR) and Modified Index Position (MIP), along with local surrogates like shapr and alternatives like GraphLIME, to enhance reliability and interpretability. The work provides practical guidance for deploying XAI in high-stakes domains, emphasizing transparent communication of assumptions and cross-model validation to improve trust and utility of explanations.

Abstract

eXplainable artificial intelligence (XAI) methods have emerged to convert the black box of machine learning (ML) models into a more digestible form. These methods help to communicate how the model works with the aim of making ML models more transparent and increasing the trust of end-users into their output. SHapley Additive exPlanations (SHAP) and Local Interpretable Model Agnostic Explanation (LIME) are two widely used XAI methods, particularly with tabular data. In this perspective piece, we discuss the way the explainability metrics of these two methods are generated and propose a framework for interpretation of their outputs, highlighting their weaknesses and strengths. Specifically, we discuss their outcomes in terms of model-dependency and in the presence of collinearity among the features, relying on a case study from the biomedical domain (classification of individuals with or without myocardial infarction). The results indicate that SHAP and LIME are highly affected by the adopted ML model and feature collinearity, raising a note of caution on their usage and interpretation.
Paper Structure (8 sections, 4 figures, 3 tables)

This paper contains 8 sections, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. eXplainable artificial intelligence
  3. SHAP
  4. LIME
  5. A case study
  6. Recommendations
  7. Conclusions
  8. Funding

Figures (4)

  • Figure 1: GitHub Star for 10 common XAI methods.
  • Figure 2: SHAP output to explain the four models globally. DT: decision tree; LGBM: light gradient-boosting machine; LR: logistic regression; SVC: support vector machines classifier; ACC: accuracy; MI: myocardial infarction.
  • Figure 3: LIME output to explain the model locally for the same instance using four classifiers. DT: decision tree; LGBM: light gradient-boosting machine; LR: logistic regression; SVC: support vector classifier; MI: myocardial infarction.
  • Figure 4: Correlation heatmap.