ShapDBM: Exploring Decision Boundary Maps in Shapley Space

TL;DR

This paper introduces ShapDBM, a novel method for visualizing decision boundaries by transforming data into Shapley space, resulting in clearer, more compact maps that improve interpretability of complex machine learning models.

Contribution

The paper presents a new technique to compute Decision Boundary Maps by transforming data into Shapley space, enhancing map quality and interpretability over traditional methods.

Findings

ShapDBM produces decision boundary maps with comparable or better quality metrics.

ShapDBM maps are more compact and easier to explore.

The method improves visualization of complex datasets.

Abstract

Decision Boundary Maps (DBMs) are an effective tool for visualising machine learning classification boundaries. Yet, DBM quality strongly depends on the dimensionality reduction (DR) technique and high dimensional space used for the data points. For complex ML datasets, DR can create many mixed classes which, in turn, yield DBMs that are hard to use. We propose a new technique to compute DBMs by transforming data space into Shapley space and computing DR on it. Compared to standard DBMs computed directly from data, our maps have similar or higher quality metric values and visibly more compact, easier to explore, decision zones.