Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ShapG: new feature importance method based on the Shapley value

Chi Zhao, Jing Liu, Elena Parilina

TL;DR

This work introduces ShapG, a graph-based global XAI method that computes feature importance via Shapley values defined on a feature correlation graph. By constructing a dense feature graph, pruning it to a connected, informative structure, and using local neighborhoods to approximate Shapley values, ShapG achieves accurate explanations with substantially reduced runtime compared to KernelSHAP and SamplingSHAP on two real-world datasets. The method is demonstrated across simple and complex models, including LightGBM, MLP, and various hybrid ensembles, showing robust attribution of important features like LSTAT and RM for housing prices and doctor recommendations and risk perceptions for H1N1 vaccination. ShapG thus provides a model-agnostic, scalable, and interpretable global explanation framework with practical advantages for trustworthy AI deployments.

Abstract

With wide application of Artificial Intelligence (AI), it has become particularly important to make decisions of AI systems explainable and transparent. In this paper, we proposed a new Explainable Artificial Intelligence (XAI) method called ShapG (Explanations based on Shapley value for Graphs) for measuring feature importance. ShapG is a model-agnostic global explanation method. At the first stage, it defines an undirected graph based on the dataset, where nodes represent features and edges are added based on calculation of correlation coefficients between features. At the second stage, it calculates an approximated Shapley value by sampling the data taking into account this graph structure. The sampling approach of ShapG allows to calculate the importance of features efficiently, i.e. to reduce computational complexity. Comparison of ShapG with other existing XAI methods shows that it provides more accurate explanations for two examined datasets. We also compared other XAI methods developed based on cooperative game theory with ShapG in running time, and the results show that ShapG exhibits obvious advantages in its running time, which further proves efficiency of ShapG. In addition, extensive experiments demonstrate a wide range of applicability of the ShapG method for explaining complex models. We find ShapG an important tool in improving explainability and transparency of AI systems and believe it can be widely used in various fields.

ShapG: new feature importance method based on the Shapley value

TL;DR

This work introduces ShapG, a graph-based global XAI method that computes feature importance via Shapley values defined on a feature correlation graph. By constructing a dense feature graph, pruning it to a connected, informative structure, and using local neighborhoods to approximate Shapley values, ShapG achieves accurate explanations with substantially reduced runtime compared to KernelSHAP and SamplingSHAP on two real-world datasets. The method is demonstrated across simple and complex models, including LightGBM, MLP, and various hybrid ensembles, showing robust attribution of important features like LSTAT and RM for housing prices and doctor recommendations and risk perceptions for H1N1 vaccination. ShapG thus provides a model-agnostic, scalable, and interpretable global explanation framework with practical advantages for trustworthy AI deployments.

Abstract

With wide application of Artificial Intelligence (AI), it has become particularly important to make decisions of AI systems explainable and transparent. In this paper, we proposed a new Explainable Artificial Intelligence (XAI) method called ShapG (Explanations based on Shapley value for Graphs) for measuring feature importance. ShapG is a model-agnostic global explanation method. At the first stage, it defines an undirected graph based on the dataset, where nodes represent features and edges are added based on calculation of correlation coefficients between features. At the second stage, it calculates an approximated Shapley value by sampling the data taking into account this graph structure. The sampling approach of ShapG allows to calculate the importance of features efficiently, i.e. to reduce computational complexity. Comparison of ShapG with other existing XAI methods shows that it provides more accurate explanations for two examined datasets. We also compared other XAI methods developed based on cooperative game theory with ShapG in running time, and the results show that ShapG exhibits obvious advantages in its running time, which further proves efficiency of ShapG. In addition, extensive experiments demonstrate a wide range of applicability of the ShapG method for explaining complex models. We find ShapG an important tool in improving explainability and transparency of AI systems and believe it can be widely used in various fields.
Paper Structure (19 sections, 2 equations, 10 figures, 10 tables, 3 algorithms)

This paper contains 19 sections, 2 equations, 10 figures, 10 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Explainable Artificial Intelligence (XAI) methods
  3. ShapG: a novel XAI method
  4. The Shapley value
  5. ShapG (explanations based on the Shapley value for graphs)
  6. The Shapley value for undirected weighted graphs
  7. Description and algorithm of the ShapG method
  8. Experiments
  9. Description of datasets
  10. Preprocessing data for ShapG
  11. AI prediction models
  12. Evaluation of XAI methods
  13. Results and analysis
  14. Feature importance calculated by ShapG
  15. "Housing price" dataset
  16. ...and 4 more sections

Figures (10)

  • Figure 1: Heatmap of Pearson correlation coefficients for the "housing price" dataset
  • Figure 2: Heatmap of Pearson correlation coefficients for the "H1N1" dataset
  • Figure 3: Graph connecting features in "housing price" dataset
  • Figure 4: Graph connecting features in "H1N1" dataset
  • Figure 5: Feature importance in "housing price" dataset calculated with ShapG
  • ...and 5 more figures

Theorems & Definitions (1)

  • Remark 1