Networks Community Detection

Using Artificial Bee Colony Swarm Optimization

Ahmed Ibrahem Hafez1,5 , Hossam M. Zawbaa3,4,5,

Aboul Ella Hassanien2,5 , and Aly A. Fahmy2
1 Faculty of Computer and Information, Minia University, Egypt
ah.hafez@gmail.com
2 Faculty of Computers and Information, Cairo University, Egypt

{aboitcairo,aly.fahmy}@gmail.com
3 Faculty of Mathematics and Computer Science, Babes-Bolyai University, Romania
4 Faculty of Computers and Information, BeniSuef University, Egypt
5 Scientific Research Group in Egypt (SRGE), Cairo, Egypt

http://www.egyptscience.net

Abstract. Community structure identification in complex networks has been an

important research topic in recent years. Community detection can be viewed as
an optimization problem in which an objective quality function that captures the
intuition of a community as a group of nodes with better internal connectivity
than external connectivity is chosen to be optimized. In this work Artificial bee
colony (ABC) optimization has been used as an effective optimization technique
to solve the community detection problem with the advantage that the number of
communities is automatically determined in the process. However, the algorithm
performance is influenced directly by the quality function used in the optimiza-
tion process. A comparison is conducted between different popular communities’
quality measures when used as an objective function within ABC. Experiments
on real life networks show the capability of the ABC to successfully find an opti-
mized community structure based on the quality function used.

Keywords: Networks community detection, Community detection, Social Net-

works, Artificial bee colony, Swarm optimization.

1 Introduction

Networks are used to model and represent many real world systems. Society offers a
wide variety of possible group organizations starting from families up to nations. The
evolution of Internet has also led to the creation of online communities. Indeed, social
communities have been studied for a long time [1]. Communities also occur in many
networked systems from biology, computer science, engineering, economics, politics,
etc. In protein-protein interaction networks, communities are likely to group proteins
having the same specific function within the cell [2]. These facts have made complex
network analysis a popular research area. A common feature of complex networks is
community structure [3], i.e. groups of nodes in the network that are more densely
connected internally than with the rest of the network. Communities play special roles

P. Krömer et al. (eds.), Proceedings of the Fifth Intern. Conf. on Innov. in 229
Bio-Inspired Comput. and Appl. IBICA 2014, Advances in Intelligent Systems and Computing 303,
DOI: 10.1007/978-3-319-08156-4_23,  c Springer International Publishing Switzerland 2014
230 A.I. Hafez et al.

in the structure-function relationship, and detecting communities (or modules) can be a

way to identify substructures that may correspond to important functions in the network.
Therefore, uncovering or detecting community structure is one of the most important
problems in the field of complex network analysis.
Many methods have been developed for the community detection problem. These
methods use tools and techniques from disciplines like physics, biology, applied math-
ematics, and computer and social sciences. Results of a recent survey can be seen in [4].
One of the first community detection algorithms proposed is the Girvan-Newman (GN)
algorithm, which introduces a divisive method that iteratively removes the edge with
the greatest betweenness value [3]. Some improved algorithms have also been proposed
[5–7]. These algorithms are based on a foundational measure criterion of community,
Modularity which measures the number of within-community edges relative to a null
model. Modularity is a popular quality function that was proposed by Newman [8]. The
larger the Modularity value, the more accurate the community partition. Consequently,
community detection becomes a Modularity optimization problem. Because the search
for the optimal (largest) Modularity value is an NP-complete problem, many heuris-
tic search algorithms, such as extremal optimization, simulated annealing, and genetic
algorithms (GA), have been applied to solve the optimization problem [4].
The remainder of this paper is organized as follows. In Section 2 we define the com-
munity problem as an optimization problem and introduce the objective functions used
in the research. In Section 3 we describe our proposed algorithm. Section 4 shows our
experimental results on real life social networks and synthetic network. We then offer
conclusions and suggestions for future work in section 5.

2 The Community Detection Problem

A social network can be modeled as a graph G = (V, E), where V is a set of vertices,
and E is a set of edges that connect two elements of V. A community structure S in a
network is a set of groups of vertices having a high density of edges among the ver-
tices and a lower density of edges between different groups. The problem of detecting
k communities in a network, where the number k is unknown can be formulated as
finding a partitioning of the nodes in k subsets that best satisfy a given a quality mea-
sure of communities F(S). The problem can be viewed as an optimization problem in
which one usually wants to optimize the given quality measure F(S). A single objective
optimization problem (Ω;F) is defined as in the equation 1.
min F(S), s.t S ∈ Ω (1)
Where F(S) is an objective function that needs to be optimized, and Ω={S1 , S2 . . . ,
Sk } is the set of feasible community structures in a network. We assume that all quality
measures need to be minimized without loss of generality. A formal definition of the
optimization problem is given in [9].

2.1 Objectives
The objective function plays an important role in the optimization process; it’s the
“steering wheel” in the process that leads to good solutions. Many objective functions