How to Explain Individual Classification Decisions

TL;DR

This paper introduces a procedure to explain individual classification decisions from any machine learning model, addressing the gap left by traditional methods like decision trees that only provide global explanations.

Contribution

It proposes a novel method that under certain assumptions can elucidate why a classifier predicts a specific label for a single data point, regardless of the model type.

Findings

Provides a general framework for local explanations

Applicable to any classification model

Enhances interpretability of machine learning predictions

Abstract

After building a classifier with modern tools of machine learning we typically have a black box at hand that is able to predict well for unseen data. Thus, we get an answer to the question what is the most likely label of a given unseen data point. However, most methods will provide no answer why the model predicted the particular label for a single instance and what features were most influential for that particular instance. The only method that is currently able to provide such explanations are decision trees. This paper proposes a procedure which (based on a set of assumptions) allows to explain the decisions of any classification method.