ShapML

The purpose of ShapML is to compute stochastic feature-level Shapley values which can be used to (a) interpret and/or (b) assess the fairness of any machine learning model. Shapley values are an intuitive and theoretically sound model-agnostic diagnostic tool to understand both global feature importance across all instances in a data set and instance/row-level local feature importance in black-box machine learning models.

This package implements the algorithm described in Štrumbelj and Kononenko's (2014) sampling-based Shapley approximation algorithm to compute the stochastic Shapley values for a given instance and model feature.

• Flexibility:

  • Shapley values can be estimated for any machine learning model using a simple user-defined predict() wrapper function.

• Speed:

  • The speed advantage of ShapML comes in the form of giving the user the ability to select 1 or more target features of interest and avoid having to compute Shapley values for all model features (i.e., a subset of target features from a trained model will return the same feature-level Shapley values as the full model with all features). This is especially useful in high-dimensional models as the computation of a Shapley value is exponential in the number of features.

• pkg.julialang.org

using Pkg
Pkg.add("ShapML")

• Development

using Pkg
Pkg.add(PackageSpec(url = "https://github.com/nredell/ShapML.jl"))

• Docs (incuding examples from MLJ)

• Consistency with TreeSHAP

• Speed - Julia vs Python vs R