Explaining predictive models using Shapley values and non-parametric vine copulas

TL;DR

This paper introduces two vine copula-based methods to improve the accuracy of Shapley value explanations for predictive models by better capturing feature dependencies, outperforming existing approaches.

Contribution

It proposes novel vine copula-based approaches for modeling feature dependence in Shapley value explanations, addressing limitations of independence assumptions.

Findings

Vine copula methods yield more accurate Shapley value approximations.

Experiments on simulated data validate the effectiveness of the proposed methods.

Real data analysis confirms improved explanation accuracy.

Abstract

The original development of Shapley values for prediction explanation relied on the assumption that the features being described were independent. If the features in reality are dependent this may lead to incorrect explanations. Hence, there have recently been attempts of appropriately modelling/estimating the dependence between the features. Although the proposed methods clearly outperform the traditional approach assuming independence, they have their weaknesses. In this paper we propose two new approaches for modelling the dependence between the features. Both approaches are based on vine copulas, which are flexible tools for modelling multivariate non-Gaussian distributions able to characterise a wide range of complex dependencies. The performance of the proposed methods is evaluated on simulated data sets and a real data set. The experiments demonstrate that the vine copula approaches give more accurate approximations to the true Shapley values than its competitors.