Detecting Community Structures in Social Networks with Particle Swarm Optimization - Frontiers in Internet Technologies - page 269

Clustering Algorithm

assign all nodes as unclassified;
for each unclassified node ν
if (

CORE ) then
generate new cluster_ID;
insert

( ν )

ε , μ

()

DirREACH ν

into queue Q;

ε , μ

while(Q != 0) do
y = first node in Q;
assign cluster_ID to y;
if ( ) then CORE ( y )

ε , μ

()

x ∈ DirREACH y

for each do

if x is unclassified then
insert x into queue Q;
if x is NOISE then
assign cluster_ID to x;
remove y from Q;
else
label v as NOISE;

ε , μ

3 Constructing New Weighted Network

After clustering process, now the original network consists of some core areas and
isolated nodes. Meanwhile, in order to reduce the network scale, a new compressed
weighted network will be constructed by abstracting each core area and isolated node
in the original network as a super node in the new weighted network. For edges that are
in the same core area of the original network, a self-join edge will be added to the
corresponding super node in the new network. And for edges that are between core
areas or isolated nodes in the original network, an edge between the corresponding
super nodes will be added. Finally, a new weighted network is constructed. Fig.1
shows the conversion. Then the optimization based on modularity will be performed
on the new constructed network which has a smaller scale than the original one.

Fig. 1. Conversion from the original network to a new weighted network