An evolutionary bi-clustering algorithm for community mining in complex networks

Abstract EN

A network (or formally a graph) can be described by a set of nodes and a set of edges connecting these nodes.

Networks model many real-world phenomena in various research domains, such as biology, engineering and sociology.

Community mining is discovering the groups in a network where individuals group of membership are not explicitly given.

Detecting natural divisions in such complex networks is proved to be extremely NP-hard problem that recently enjoyed a considerable interest.

Among the proposed methods, the field of evolutionary algorithms (EAs) takes a remarkable interest.

To this end, the aim of this paper is to present the general statement of community detection problem in social networks.

Then, it visits the problem as an optimization problem where a modularity-based ( ) and normalized mutual information ( ) metrics are formulated to describe the problem.

An evolutionary algorithm is then expressed in the light of its characteristic components to tackle the problem.

The presentation will highlight the possible alternative that can be adopted in this study for individual representation, fitness evaluations, and crossover and mutation operators.

The results point out that adopting as a fitness function carries out more correct solutions than adopting the modularity function .

Moreover, the strength of mutation has a background role.

When coupled with non elite selection, increasing mutation probability could results in better solutions.

However, when elitism is used, increasing mutation probability could bewilder the behavior of EA.