Impact of Network Topology on Efficiency of Proximity Measures for Community Detection

Abstract

Many community detection algorithms require the introduction of a measure on the set of nodes. Previously, a lot of efforts have been made to find the top-performing measures. In most cases, experiments were conducted on several datasets or random graphs. However, graphs representing real systems can be completely different in topology: the difference can be in the size of the network, the structure of clusters, the distribution of degrees, the density of edges, and so on. Therefore, it is necessary to explicitly check whether the advantage of one measure over another is preserved for different network topologies. In this paper, we consider the efficiency of several proximity measures for clustering networks with different structures. The results show that the efficiency of measures really depends on the network topology in some cases. However, it is possible to find measures that behave well for most topologies.

Dissecting graph measure performance for node clustering in LFR parameter space

“…There are many measure benchmarking studies considering node classification and clustering for both generated graphs and real-world datasets, including [23,50,51,2,32,29,28,3,4,16,39]. Despite a large number of experimental results, an exact theory is still a matter of the future.…”