A Generalized and Adaptive Method for Community Detection

TL;DR

This paper introduces a flexible, generalized Louvain method for community detection in complex networks, addressing limitations of traditional modularity and maintaining high performance across various quality functions.

Contribution

It presents a generic Louvain algorithm adaptable to different quality functions, with a sufficient condition for integration and comparable performance to the classical method.

Findings

The generalized Louvain method performs similarly to the classical version.

It can incorporate various quality functions beyond modularity.

The approach maintains high efficiency on large graphs.

Abstract

Complex networks represent interactions between entities. They appear in various contexts such as sociology, biology, etc., and they generally contain highly connected subgroups called communities. Community detection is a well-studied problem and most of the algorithms aim to maximize the Newman-Girvan modularity function, the most popular being the Louvain method (it is well-suited on very large graphs). However, the classical modularity has many drawbacks: we can find partitions of high quality in graphs without community structure, e.g., on random graphs; it promotes large communities. Then, we have adapted the Louvain method to other quality functions. In this paper, we describe a generic version of the Louvain method. In particular, we give a sufficient condition to plug a quality function into it. We also show that global performance of this new version is similar to the classical Louvain algorithm, that promotes it to the best rank of the community detection algorithms.