Comparing Graph Clusterings: Set partition measures vs. Graph-aware measures

TL;DR

This paper introduces graph-aware partition similarity measures that incorporate graph topology, offering a complementary perspective to traditional set partition measures for evaluating graph clusterings.

Contribution

It proposes a new family of graph-aware measures that address limitations of set partition measures in capturing graph topology during clustering comparison.

Findings

Graph-aware measures behave differently from set partition measures regarding resolution.

Both measure types provide complementary information for assessing clustering similarity.

Graph-aware measures improve understanding of partition differences in topologically complex graphs.

Abstract

In this paper, we propose a family of graph partition similarity measures that take the topology of the graph into account. These graph-aware measures are alternatives to using set partition similarity measures that are not specifically designed for graph partitions. The two types of measures, graph-aware and set partition measures, are shown to have opposite behaviors with respect to resolution issues and provide complementary information necessary to assess that two graph partitions are similar.