Skip to content

A Unified Approach to Interpreting Model Predictions

Authors: Scott M. Lundberg, Su-In Lee

Published: 2017 (Conference Paper)

Source: Advances in Neural Information Processing Systems

Algorithm: SHAP

arXiv: 1705.07874

DOI: 10.5555/3295222.3295230

Summary

Introduces SHAP (SHapley Additive exPlanations), a unified framework for local model explainability grounded in Shapley values from cooperative game theory. Shows that six prior explanation methods are all special cases of SHAP values, and that only SHAP satisfies a desirable set of axioms (local accuracy, missingness, consistency), making it the standard feature-attribution method.

Abstract

Understanding why a model makes a certain prediction can be as crucial as the prediction's accuracy in many applications. However, the highest accuracy for large modern datasets is often achieved by complex models that even experts struggle to interpret, such as ensemble or deep learning models, creating a tension between accuracy and interpretability. In response, various methods have recently been proposed to help users interpret the predictions of complex models, but it is often unclear how these methods are related and when one method is preferable over another. To address this problem, we present a unified framework for interpreting predictions, SHAP (SHapley Additive exPlanations). SHAP assigns each feature an importance value for a particular prediction. Its novel components include: (1) the identification of a new class of additive feature importance measures, and (2) theoretical results showing there is a unique solution in this class with a set of desirable properties. The new class unifies six existing methods, notable because several recent methods in the class lack the proposed desirable properties. Based on insights from this unification, we present new methods that show improved computational performance and/or better consistency with human intuition than previous approaches.

Primary

Paper arxiv.org

Standard

arXiv Abstract 1705.07874 arXiv PDF 1705.07874 arXiv HTML 1705.07874 DOI 10.5555/3295222.3295230

Alternate

proceedings.neurips.cc proceedings.neurips.cc GitHub: shap/shap github.com

Tags

  • Explainability

  • Feature attribution

  • SHAP

  • Shapley values

  • Machine learning interpretability

  • Model explanations