Explanations of model predictions with live and breakDown packages

TL;DR

This paper introduces R packages 'live' and 'breakDown' for explaining complex model predictions, compares them with existing methods like lime and ShapleyR, and enhances interpretability of black box models.

Contribution

The paper presents two new R packages for model explanation, offering novel approaches and benchmarking them against existing state-of-the-art solutions.

Findings

live and breakDown provide effective attribution of predictions to input features

Comparison shows strengths and limitations relative to lime and ShapleyR

Enhances interpretability of complex models in R environment

Abstract

Complex models are commonly used in predictive modeling. In this paper we present R packages that can be used to explain predictions from complex black box models and attribute parts of these predictions to input features. We introduce two new approaches and corresponding packages for such attribution, namely live and breakDown. We also compare their results with existing implementations of state of the art solutions, namely lime that implements Locally Interpretable Model-agnostic Explanations and ShapleyR that implements Shapley values.