Shapley Computations Using Surrogate Model-Based Trees

TL;DR

This paper introduces a surrogate model-based tree method for efficient and accurate computation of Shapley and SHAP values, unifying global and local interpretability with adjustable trade-offs between speed and precision.

Contribution

It proposes a novel surrogate model-based tree approach for computing Shapley values, addressing computational challenges and unifying interpretation methods.

Findings

Improves accuracy of Shapley value computation

Unifies global and local interpretability

Offers a trade-off between runtime and accuracy

Abstract

Shapley-related techniques have gained attention as both global and local interpretation tools because of their desirable properties. However, their computation using conditional expectations is computationally expensive. Approximation methods suggested in the literature have limitations. This paper proposes the use of a surrogate model-based tree to compute Shapley and SHAP values based on conditional expectation. Simulation studies show that the proposed algorithm provides improvements in accuracy, unifies global Shapley and SHAP interpretation, and the thresholding method provides a way to trade-off running time and accuracy.