Volume 8, Issue 10, October – 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

A Survey on Centrality Measures of a Network

Seema Varghese1
1
Department of Mathematics, Government Engineering College, Thrissur
APJ Abdul Kalam Technological University, Kerala, India

Abstract:- A node's centrality score is an index that economics, and computer science. Centrality measures play a
identifies its importance inside a network. Depending on fundamental role in network analysis, providing insights into
the goal of the study, different writers define different the relative importance and influence of nodes within a
centrality measures. Numerous fields, including sociology, network. In this survey, we will delve into centrality
neurobiology, communication networks, electrical measures, their types, applications, and significance in the
networks, etc., find use for these measurements. The goal study of networks. Centrality measures aim to quantify the
of this work is to provide a thorough overview of degree, importance of nodes within a network. These measures are
closeness, eigen vector, and betweenness centrality. based on the fact that not all nodes are equal; some nodes
have a more significant impact on the network's structure and
Keywords:- Centrality Measure; Degree Centrality; functionality. Centrality measures provide a way to identify
Closeness Centrality; Eigen Vector Centrality; Betweenness and analyze these influential nodes.
Centrality.
Centrality measures have a wide range of applications
I. INTRODUCTION across various domains. In social networks, centrality
measures help identify key influencers, opinion leaders,
Bavelas first proposed the idea of centrality in relation and individuals who play pivotal roles in the spread of
to human communication in 1948. He put up a link between information or behaviors. Centrality measures assist in
structural centrality and influence in group dynamics since he identifying critical transportation hubs and optimizing routes
was particularly interested in how people interacted in small for efficient transportation systems. In biological networks
groups. Effective group problem-solving and perceptions of such as protein-protein interaction networks, centrality
leadership were linked to centrality. The idea of centrality, measures can identify essential proteins for understanding
however, has been applied in situations other than only disease mechanisms and drug target identification. In web
experimental studies of group problem-solving. In an effort search algorithms, centrality measures like PageRank help
to understand political unification in light of the diversity of rank web pages, improving the relevance of search results. In
Indian social life, Cohn and Marriott also used the centrality business and organizational networks, centrality measures
idea in 1958. They essentially questioned the viability of any can identify key employees or departments responsible for
form of government in a nation the size and variety of India. information flow and decision-making.
They concluded that network centers bound and twisted
different strands into a planned framework, which connected Centrality measures are essential tools in network
every facet of Indian social life.. analysis for several reasons. They help to pinpoint nodes that,
if removed or targeted, can have a significant impact on the
The idea of centrality is still applicable today and is network's structure and functionality. They provide insights
being applied in more and more situations. Everyone seems into how information, influence, or resources flow within a
to concur that centrality is an essential structural element of network, aiding in the prediction of network behavior and
social networks. But there isn't much agreement on how to vulnerabilities. Centrality measures assist in designing
measure centrality effectively, and there isn't agreement on interventions or strategies that focus on influential nodes to
what centrality actually is or what its conceptual foundations maximize desired outcomes. In network design, centrality
are either. Over the years, numerous alternative centrality measures help in optimizing the placement of important
metrics have been proposed. The establishment of metrics nodes or resources for efficiency and robustness.
should aid in idea clarification by describing a concept's
components and how they connect to one another. In the In this paper, four types of centrality measures are
instance of centrality, it appears that the opposite outcome discussed viz. Degree Centrality, Closeness Centrality,
was achieved. The complexity of many of the measurements Betweenness Centrality and Eigen vector Centrality. Degree
makes it difficult or impossible to tell what, if anything, they Centrality, Closeness Centrality, Betweenness Centrality and
are measuring. Eigen vector Centrality. The advantages and disadvantages of
these centrality measures are surveyed in detail. For all basic
Networks are pervasive in our modern world, graph theoretic terminologies, we refer to [ 17].
encompassing a wide spectrum of applications, from social
interactions on social media platforms to the structure of
transportation systems and the internet. Understanding the
structure and dynamics of these networks is crucial for
various fields, which include social science, biology,

IJISRT23OCT258 www.ijisrt.com 79
Volume 8, Issue 10, October – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
II. DEGREE CENTRALITY four neighbours and two among them are high degree
vertices which can dominate the whole graph.
Some of the numerous nodes that make up a network
are essential in mediating a huge number of network III. CLOSENESS CENTRALITY
connections. These nodes play a crucial role in network
structure and are frequently recognised by numbers referred A useful metric called closeness centrality predicts how
to as centrality metrics. The concept of degree centrality was quickly data would go from one vertex to another. How short
put forward by Bavelas in 1948 to make clear how the the geodesics are from vertex i to all other vertices is a
structural position of an individual within a social network measure of closeness centrality [18]. Two vertices are
determines his or her influence in group social activities. topologically close if they are connected by a short path.
More generally, a vertex has high closeness centrality if it is
In 1979, Freeman pointed out that there is a common connected, on average, by short paths to a large number of
theme across different centrality metrics. It is that the fact other vertices in the network. Vertex efficiency and closeness
that they give the same results when applied to a star graphs. are equivalent measures.
Three fundamental properties can be ascribed to the central
vertex. “It has the maximum possible degree. It falls on the Closeness centrality is typically stated as “the inverse of
shortest possible topological path between all pairs of the normalised sum of the topological distances in the
vertices. it is located at the shortest topological distance from graph”. The farness between the vertices is another name for
all other vertices” [18]. this parameter [18]. Sometimes closeness centrality is just
written as the inverse of distance.
The first step in network analysis is to comprehend how
connection fluctuates between nodes. The number of The Closeness Centrality measure of a node i is denoted
connections each vertex has to the other vertices in the as
network, known as vertex degree, is may be the simplest
metric we can compute in this context. The distribution of CC(i) = N-1 / ∑ dij
degree values across vertices is frequently heterogeneous in
real-world networks; we frequently see that many vertices where N is the number of nodes and dij is the shortest
have few linkages while a smaller number of important distance between the ith and jth node.
vertices receive the majority of connectivity, designating
them as putative hubs that allow integration throughout the Closeness centrality is a concept in network analysis
network. and graph theory that measures the centrality or importance
of a vertex within a network based on its proximity to other
The Degree Centrality measure of a node i is denoted as vertices. It quantifies how close a vertex is to all other
vertices in the network. In essence, it identifies vertices that
CD(i) = ∑Aij can quickly reach other vertices in the network with the
shortest path.
where A is the adjacency matrix. This definition
assumes that vertices with many connections exert more Vertices with high closeness centrality are considered to
influence over network functions than others. be more central within the network because they can reach
other vertices more quickly. They are often important for
information flow and communication within the network.
The closeness centrality calculation is based on the concept
of the shortest path, which is the minimum number of edges
or steps required to reach one vertex from another. Vertices
with lower cumulative shortest path distances tend to have
higher closeness centrality. Closeness centrality values are
typically normalized by dividing them by (N-1), where N is
the total number of vertices in the network. This
normalization ensures that the centrality values are
comparable across networks of different sizes. A vertex with
a closeness centrality value close to 1 is considered highly
central, while a vertex with a value closer to 0 is less central.
Fig :1 Vertex 6 has highest degree centrality Closeness centrality is one of several centrality measures
used in network analysis. It helps identify important vertices
The limitations of this centrality measure are that all in various types of networks, including social networks,
connections are treated alike. It counts quantity but does not transportation networks, and communication networks.
consider quality. For example, in Fig :1, node 10 has more
degree than node 7. But a closer look will reveal that four of
the five neighbours of node 10 are of degree one and hence
have less influence in the network. Whereas, node 7 has only

IJISRT23OCT258 www.ijisrt.com 80
Volume 8, Issue 10, October – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
IV. BETWENNESS CENTRALITY only one in a star to obtain the maximum value taken by
CB(i). The number of these geodesics, which is n-1C2
Betweenness centrality is a crucial component of the determines the central vertex's betweenness centrality in a
analysis of social networks [1, 2], computer networks [3], star Sn (Fig:2) with n vertices. Since no pendant vertex is
and numerous other kinds of network data models [4–9]. A located between any geodesic, their betweenness centrality is
unit's closeness to other units in a communication network is zero. Once more, it is clear that in a complete network Kn, the
not its sole important property. It is more important which betweenness centrality of any vertex is 0 because there are no
units are situated on the geodesics, or shortest routes, vertex-geodesic intersections because each geodesic has a
between pairs of other units. These organizations have length of 1.
control over the network's information flow. As a measure of
a vertex's ability to influence communication, betweenness
centrality can be useful.

By measuring how much a vertex is situated on the

geodesics joining pairs of other vertices, betweenness
centrality [10–14] measures the betweenness of a vertex in a
network. In many situations in the actual world, it is quite
important. When there is only a single geodesic between any
pair of vertices determining betweenness is simple and
straightforward and the internal vertices of the geodesic have
complete control over communication between pairs of
others. When there are several geodesics connecting a pair of
vertices, the situation worsens and the control of the internal
vertices gets fragmented.

Bavelas initially established the idea of betweenness Fig: 2 The Star Graph S6
centrality in 1948 [15]. The potential of a vertex to regulate
information flow in the network is the significance of the V. EIGEN VECTOR CENTRALITY
vertex centrality idea. The degree to which they stand out
from others and have the potential to assist, obstruct, or Eigenvector centrality is a concept in network analysis
influence the flow of signals makes positions considered to and graph theory that measures the centrality or importance
be structurally central. Freeman divided betweenness of a vertex within a network based on the idea that the
centrality into three categories in his papers [5, 16]. Two importance of a vertex depends on the importance of its
vertex centrality indices—one based on counts and one based neighbors. In other words, a vertex is considered central if it
on proportions—as well as a measure of network or graph is connected to other central vertices.
centralization overall are included in the three measurements.
“Mathematically, eigenvector centrality [18] is
“Betweenness centrality, 𝐶𝐵(V) for a vertex i is defined calculated using the following formula:
as x(i) = 1 / λ * Σ(j) A(i, j) * x(j)

𝐶𝐵 (i) = Σ 𝜎𝑠𝑡 (i) / 𝜎𝑠𝑡 where: x(i) is the eigenvector centrality of node i, λ is
the largest eigenvalue of the adjacency matrix of the network.
where 𝜎𝑠𝑡 is the number of shortest paths with vertices 𝑠 A(i, j) is the element in the adjacency matrix that represents
and 𝑡 as their end vertices, while 𝜎𝑠𝑡(i) is the number of those the connection between node i and node j. Σ(j) represents the
shortest paths that include vertex i” [16]. High centrality sum over all neighbouring nodes j of node i.”
value is an indication that a node lies on a large fraction of
geodesics connecting pairs of nodes [19]. Each pair of nodes Eigenvector centrality considers not just the number of
in a connected network provides a value between 0 and 1 to connections a vertex has (as in degree centrality) but also the
the betweenness centrality of all other nodes. If there is only importance of the vertices to which it is connected. In other
a single shortest path joining a particular pair of vertices, then words, it values connections to vertices that are themselves
that pair provides a betweenness centrality 1 to each of its central. Eigenvector centrality is calculated iteratively. The
internal nodes and zero to all other nodes. For example, in a centrality of a vertex depends on the centrality of its
path graph, a pair of nodes gives a betweenness centrality neighbours, and this process continues until it converges to a
value 1 to each of its internal nodes and zero to the external stable solution.
nodes. If there are 𝑘 geodesics of length 2 joining a pair of
nodes, then that pair of nodes provides a betweenness The largest eigenvalue (λ) of the adjacency matrix plays
centrality 1/𝑘 to each of the internal nodes. a crucial role in the calculation. The centrality values are
scaled by this eigenvalue, which ensures that all centrality
As the universal vertex of a star is located on the values are positive and allows for comparisons among nodes.
geodesic (which is unique) connecting every pair of other Nodes with higher eigenvector centrality values are
vertices, Freeman [16] shown that the central vertex is the considered more central or influential within the network. A
node with a high eigenvector centrality value is not only

IJISRT23OCT258 www.ijisrt.com 81
Volume 8, Issue 10, October – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
connected to many other nodes but is also connected to nodes [11]. S. P. Borgatti, “Centrality and network flow,” Social
that are themselves central. Networks, vol.27, no.1, pp.55-71, 2005.
[12]. S. P. Borgatti and M. G. Everett, “A graph-theoretic
Eigenvector centrality is commonly used in various perspective on centrality,” Social Networks, vol. 28, no.
fields, including social network analysis, web page ranking 4, pp. 466–484, 2006.
(PageRank is a variant of eigenvector centrality), and the [13]. E. Estrada, D. J. Higham, and N. Hatano,
study of information flow in networks. It helps identify key “Communicability betweenness in complex networks,”
nodes that can exert significant influence or control over the Physica A: Statistical Mechanics and its Applications,
network due to their connections to other influential nodes. vol. 388, no. 5, pp. 764–774,2009
[14]. S. Gago, J. C, Hurajova, and T. Madaras, “On
VI. CONCLUSION betweenness-uniform graphs,” Czechoslovak
Mathematical Journal, vol. 63, no. 3, pp. 629–642,
For the analysis and comprehension of complex 2013.
networks in many different disciplines, centrality [15]. A. Bavelas, “A mathematical model for group structure.
measurements are essential tools. They enable us to discover Human organization,” Applied Anthropology, vol. 7, no.
the underlying structure, pinpoint important participants, and 3, pp. 16–30, 1948.
decide on network architecture, optimisation, and [16]. L. C. Freeman, “A set of measures of centrality based
interventions with knowledge. The study of centrality metrics on betweenness,” Sociometry, vol. 40, pp. 35–41, 1977.
continues to be essential to understanding networks and how [17]. F. Harary, Graph Theory, Addison-Wesley, USA, 1969
they affect our lives as our world gets more interconnected. [18]. A. Fornito, A. Zalesky, E. T. Bullmore, Fundamentals
of Brain Network Analysis, Academic Press, UK,2016
REFERENCES [19]. S. Kumar, R. Unnithan, B. Kannan, M. Jathavedan,”
Betweenness Centrality in some classes of graphs”,
[1]. E. Otte and R. Rousseau, “Social network analysis: a International Journal of Combinatorics,,Volume 2014,
powerful strategy, also for the information sciences,” Article ID 241723, 12 pages
Journal of Information Science, vol. 28, no. 6, pp. 441– http://dx.doi.org/10.1155/2014/241723
453, 2002.
[2]. K. Park and A. Yilmaz, “A social network an analysis
approach to analyze road networks,” in Proceedings of
the ASPRS Annual Conference, San Diego, Calif, USA,
2010.
[3]. V. Latora and M. Marchiori, “A measure of centrality
based on network efficiency,” New Journal of Physics,
vol. 9, article 188, 2007.
[4]. E. Estrada, “Virtual identification of essential proteins
within the protein interaction network of yeast,”
Proteomics, vol. 6, no.1, pp. 35–40, 2006.
[5]. L. C. Freeman, “Centrality in social networks
conceptual clarification,” Social Networks, vol. 1, no. 3,
pp. 215–239, 1979.
[6]. D. Kosch¨utzki and F. Schreiber, “Centrality analysis
methods for biological networks and their application to
gene regulatory networks,” Gene Regulation and
Systems Biology, vol. 2008, no.2, pp. 193–201, 2008.
[7]. A. M. M. Gonz´alez, B. Dalsgaard, and J. M. Olesen,
“Centrality measures and the importance of generalist
species in pollinationnetworks,” Ecological Complexity,
vol. 7, no. 1, pp. 36–43, 2010
[8]. M. E. J.Newman, “Scientific collaboration networks. II.
sho rtest paths, weighted networks, and centrality,”
Physical Review E,vol. 64, no. 1, pp. 1–7, 2001
[9]. M. Rubinov and O. Sporns, “Complex network
measures of brain connectivity: uses and
interpretations,” NeuroImage, vol.52, no. 3, pp. 1059–
1069, 2010.
[10]. P. Bonacich, “Power and centrality: a family of
measures,” American Journal of Sociology, vol. 92, no.
5, pp. 1170–1182, 1987.M. Young, The Technical
Writer’s Handbook. Mill Valley, CA: University
Science, 1989.

IJISRT23OCT258 www.ijisrt.com 82