Finding overlapping communities in networks using evolutionary method

TL;DR

This paper introduces an evolutionary algorithm for detecting overlapping communities in networks, effectively capturing nodes that belong to multiple groups by using a novel encoding schema and informativeness measure.

Contribution

It proposes a new evolutionary method with a specialized encoding schema to identify overlapping communities, addressing limitations of existing algorithms.

Findings

Successfully detects overlapping communities in complex networks

Handles nodes with multiple community memberships

Provides a new measure for node informativeness

Abstract

Community structure is a typical property of many real-world networks, and has become a key to understand the dynamics of the networked systems. In these networks most nodes apparently lie in a community while there often exists a few nodes straddling several communities. An ideal algorithm for community detection is preferable which can identify the overlapping communities in such networks. To represent an overlapping division we develop a encoding schema composed of two segments, the first one represents a disjoint partition and the second one represents a extension of the partition that allows of multiple memberships. We give a measure for the informativeness of a node, and present an evolutionary method for detecting the overlapping communities in a network.