Penguins Search Optimization Algorithm for Community Detection in Complex Networks

Abstract

In the last decade, the problem of community detection in complex networks has attracted the attention of many researchers in many domains, several methods and algorithms have been proposed to deal with this problem, many of them consider it as an optimization problem and various bio-inspired algorithms have been applied to solve it. In this work, the authors propose a new method for community detection in complex networks using the Penguins Search Optimization Algorithm (PeSOA), the authors use the modularity density evaluation measure as a function to maximize and they propose also to enhance the algorithm by using a new initialization strategy. The proposed algorithm has been tested on four popular real-world networks; experimental results compared with other known algorithms show the effectiveness of using this method for community detection in social networks.

A Distributed Hybrid Community Detection Methodology for Social Networks

“…The Genetic Algorithms [2, [14][15][16][17], which are particularly repetitive meta-heuristics inspired by the theory of natural evolution; •…”

“…Handle the network connectivity information, expressed in vectors, as part of each corresponding user node information [22,23,31]; • Construct content edges by selecting the top K neighbors of each vertex using the node attribute information contained [10,13]; • Optimize a unified objective function [24,32]; and • Leverage the well-known Swarm Intelligence methods such as BA (Bat Algorithm) [14,17], FA (Firefly Algorithm) [14,16] and PeSOA (Penguins Search Optimization) [15] to define bio-inspired metaheuristics schemes that use the evolution mechanisms to proceed to the detection of underlying communities.…”

“…Those hybrid approaches have been pragmatically proven [10,13,15,17,24] to generate more qualitative community hierarchy structures to the classic approaches' outcome. However, they have shown strong limitations performance-wise, regarding the size of the networks that are capable of handling, and sensitivity regarding the processing type of the information, especially regarding the binary and the categorical data types [7].…”

“…In particular, based on the homophily concept, which is the natural human tendency for interacting with peers of similar interests and characteristics [9], it is more than obvious that a valuable community should be the group of associated vertices that besides being more densely intra-connected, should also have various distinctive node attributes in common [3]. Thus, plentiful hybrid methods [3,7,[10][11][12][13][14][15][16][17] have been published to uncover communities, leveraging not only the vertices associations but also the inherent node attribute information.…”

“…On the contrary, the more sophisticated hybrid methods [7,[13][14][15][16][17][18][22][23][24][25][26][27][28] try to combine all the intrinsic information modalities by defining either global probabilistic inference-based models or global heuristic measure-based models to unveil the underlying communities. Even if their generated outcome is considered more accurate and further qualitative comparing to the respective of classic algorithms, since all the different information aspects are taken into account, their application is limited to particularly small networks due to the exhaustive statistical analysis required.…”