In this example, PSO
may not capture the true relationship between the members of the community. We
are using Zachary’s network of karate club members, a well-known dataset used
to test the performance of the communities. Fig 1(a) represents the original
community in Zachary’s network. It is separated into two distinct communities
in the network. Fig 1(b) represents the resulting communities generated by
traditional PSO techniques. They generate superfluous small communities in the
network. So, it reduced the true relationship between the members in the
networks. To improve this drawback, we use some optimization strategy and
post-processing strategy for merging the small-small communities.
to the traditional algorithm, PSO based on the global search algorithm helps to
analyze the social behavior of swarms and it convergence faster. It helps to
avoid local optima and then get a more accurate result. So, PSO can produce
better result compared with the traditional algorithms.
main advantage of this paper is
1) Discovering overlapping
communities in a social network is equal to detecting the disjoint communities
in the social network.
2) To find high-quality and
finer-grained overlapping communities from social networks we propose to
include ensemble clustering into discrete particle swarm optimization.
3) To merge finer-grained and
suboptimal overlapping communities a novel post-processing strategy is used.
The rest of the paper is organized as follows.
Related works on overlapping community detection have been discussed in section
2. Definition and corollary have been discussed in section 3. Section 4
discusses the details of our proposed work. The experimental setup has been
discussed in Section 5. Section 6 concludes the paper.
II RELATED WORK
Detection in Social Networks
social media applications, such as Digg, Flickr and YouTube have created
different multimedia social networks, eager to performing community detection
on social networks. T. Mei et al 22 presented a description of the YouTube
video-sharing virtual community. M. Wang et al 23 Proposed to deal with the
problem of disambiguates by applying community detection method to remove
communities hidden in the social networks.
J.Xie S.Killey 30 they use a multi-stage algorithm based on the
local-clustering methodology for content detection. A. Amelio and C. Pizzuti
28 Proposed to detect communities in the networks they use hypergraph
modeling which is different from LEPSO methodology; their goal to detect
meaningful communities in the networks.
2. Overlapping Communities Detection
In recent times, more
effort to defining the methods of community detection focuses on finding the
overlapping communities in the networks. A.Amelio, C. Pizzuti and J. J. Whang
28 29 proposed the link clustering method have been successfully applied to
overlapping communities’ detection. To detect overlapping communities they use
link clustering method by partitioning links instead of nodes. The advantage of
link clustering is that it produces an overlapping subgraph of the original
graph. J. B. Pereira 25 proposed to find the overlapping community modules
for protein-protein they use line graph method. Y. Y. Ahn, J. P. Bagrow, and S.
Lehmann 26 proposed a hierarchical agglomerative link clustering technique
they group the links into related clusters. J. Li and Y. Song 27 fixing the
threshold in the real world applications can easily mislead the clustering
process and they detect poor overlapping communities in the networks.
Definition and Corollary
this section, we discussed about the definitions, theorems, and corollaries in
A graph consists of nodes (N) and edges (E).
Defined as G = edge corresponds to the pair of vertices or nodes. If the vertex
n N represent a set of vertices’ incidence to n, and degree of the vertex or
node deg (n).