IEEE Systems, Man and Cybernetics Magazine - October 2023 - 22

between network structures, such as social enhancement
and edge weights, and information dissemination
in social networks, is studied.
Influence Maximization Algorithms
This article is based on the basic principles of complex
network theory: that is, the greater the degree value of a
node in the network, the greater its influence on the web.
At the same time, there is a community effect in social
networks: that is, user nodes in social networks can freely
form social groups according to their interests and preferences,
and user nodes in the same community tend to
influence each other, while for users in different communities,
the ability to control between nodes will be significantly
weakened. Figure 1 shows the flow of the two most
influential algorithms proposed in this article to obtain
the most significant user nodes.
Incremental Index of Modularity
Detecting the authenticity of news published by influential
user nodes in social groups with similar interests and
hobbies can suppress the spread of false information
more targeted in a smaller network structure. Therefore,
we define modularity Q to divide each user node into the
corresponding community correctly. The calculation of
modularity Q is only carried out in the same node community
c, which is used to measure the quality of network
community division quantitatively. The closer the modularity
Q is to 1, means that for the current node community,
the division effect is better. By calculating
TQ , the
increment of the present node community c before and
after the iterative process is obtained, and the community
with the most extensive modularity, Q, is obtained
after iteration. The calculation formula of the modularity
Q is as follows:
Q
=- d
:
kk
2m vw
1 | Avw 2m
vwD (, ).
cc
vw
(1)
Among them, m represents the total number of connected
edges of the node community network; v and w,
respectively, represent two random nodes in the network;
if v is connected to w, then
A .1vw = Otherwise, A 0vw
= ;
kv and k ,w
respectively, represent node v and the degree
value of w: that is, the number of connected edges of the
node in the community. If the nodes v and w belong to the
same community c, then (, ),cc 1vw
d (, ).cc 0vw
d
=
LCLD Algorithm
This article proposes combining the sum of the nearest
neighbor degree of the user node and the clustering coefficient
of the user node itself as the node centrality feature.
Based on the idea of Louvain's community division [12],
this article proposes a node centrality method LCLD
based on the local properties of complex networks. The
steps are as follows.
◆ As shown in Figure 1(a), first set each node in the network
as an independent community. Then, for any
adjacent nodes i and j, add node i to the community M
where its neighbor node j is located.
◆ Then calculate the modularity increment QT before
and after joining, and compare the maximum QT from
node i and all its neighbor nodes.
◆ If Q 0T2 , add node i to the community where the corresponding
neighbor node is located. Otherwise, it
remains unchanged. Repeat iterations until the first
layer of the network's community structure is divided.
◆ Use the communities divided into (1) to build a new
network.
◆ Let the weight of the edges between nodes be the sum
of the consequences of all the boundaries between the
two communities, and repeat the division method in
(1) to (3) to get the first node of the network.
◆ The second-tier community structure, and so on, gets
the final network community structure.
◆ Calculate the CLD value. Then, score all the nodes in
the divided network:
CLD () () .diC1
! ()
=+ i
|
jN i
set of node i, Ci
Among them, N(i) represents the nearest neighbor node
represents the clustering coefficient of
node i, and d represents the degree value of node i.
First, sort the communities according to their size, and
then select the node with the highest and second highest
scores from each district until several s nodes are chosen
to form the most influential CLD node set.
(a)
(b)
Figure 1. Buzz Feed News most influential nodes.
(a) LCLD algorithm, (b) RMD algorithm.
22 IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE October 2023
RMD Algorithm
In addition to LCLD, since most complex network influence
maximization algorithms rely on obtaining the global
node characteristics of the network, which is
accompanied by high time complexity, this article proposes
another critical node selection method based on
the RMD algorithm. The algorithm only needs to obtain
the local information of the selected node and its neighbor
nodes, which has higher operating efficiency and
stronger practicability.
(2)
= otherwise
IEEE Systems, Man and Cybernetics Magazine - October 2023

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - October 2023
