Drop the mask! GAMM-A Taxonomy for Graph Attributes Missing Mechanisms

TL;DR

This paper introduces GAMM, a taxonomy for understanding missing data mechanisms in attributed graphs, highlighting the limitations of current imputation methods under realistic graph-aware missingness scenarios.

Contribution

It extends existing missing data taxonomies to attributed graphs by incorporating graph-specific dependencies, providing a systematic framework for analyzing missingness.

Findings

State-of-the-art imputation methods perform poorly on graph-aware missingness scenarios

GAMM framework links missingness to node attributes and graph structure

Empirical results highlight the need for graph-specific missing data handling

Abstract

Exploring missing data in attributed graphs introduces unique challenges beyond those found in tabular datasets. In this work, we extend the taxonomy for missing data mechanisms to attributed graphs by proposing GAMM (Graph Attributes Missing Mechanisms), a framework that systematically links missingness probability to both node attributes and the underlying graph structure. Our taxonomy enriches the conventional definitions of masking mechanisms by introducing graph-specific dependencies. We empirically demonstrate that state-of-the-art imputation methods, while effective on traditional masks, significantly struggle when confronted with these more realistic graph-aware missingness scenarios.