Social Network

Analysis:
Introduction to the
Course
Even Semester of Academic Year 2023-24
Instructors: 1) Sankita Patel 2) Jetashri Gandhi
Slide Credits: Teaching Material on Social Network Analysis by Tanmoy
Chakraborty, Wiley, 2021
Motivation, Syllabus,
Teaching Scheme and
Course Outcome
Books to refer
• Social Network Analysis by Tanmoy Chakraborty, Wiley (Slide credits
to this book)
• Network Science by Barabasi, Cambridge University Press
• Slide Credits: Teaching Material on Social Network Analysis by Tanmoy
Chakraborty, Wiley, 2021
What is Social Network Analysis?

Network:
An abstract representation of relations among
entities

Social Network:
A simplified representation of the social
structure characterized by actors and ties

Social Network Analysis:

The application of networks and
graph theory to analyze the relations
present in a society
 Network: Definition

 A network, also referred to as a graph, is defined as an ordered pair G(V, E), where V is a set of
nodes (also referred to as vertices or entities), and E is a set of edges (also referred to as links or
relations) joining the nodes.
 Depending on the nature of application, the above definition may be revised or augmented, as
follows:
 the nature of edges may vary – undirected (also called symmetric, or reversible) edges, directed (also called
asymmetric, or irreversible) edges, or hyperedges, etc.
 both the nodes and/or the links are associated with one or more attributes/features like weights, timestamps,
textual features, etc.
 An edge in a graph may have same node as end nodes. Such edges of a graph are called self loops
(or, simply, loops).
 A graph may have more than one edge joining a pair of nodes. Such edges are called parallel edges.
 Network: Definition (contd…)
A graph having neither self loops nor parallel edges are
called a simple graph.
A graph having directed edges (i.e. links having a
direction) is called a directed graph. Directed edges are
also referred to as arcs. A directed graph

A graph with loops A graph having no direction in its edges is called an 3

and parallel edges undirected graph. 3
A graph having weights associated with its edges are 1 -6
6 2 9 7
called a weighted graph. A weighted graph can be
4
directed as well as undirected. 7
-5
A graph having its nodes and/or edges attributed with -2
A simple graph feature values is called an attributed graph. A weighted graph
 A sample of Twitter follower-followee network
(image source: http://allthingsgraphed.com/2014/11/02/twitter-friends-network/)
Social Network Analysis:
Key Features
•Required Knowledge Domains
 Sociology  Psychology  Mathematics  Statistics  Computer Science

•Study Benefits
To know the way social interactions influence a network

To learn how the information flows inside a network

To characterize roles of the individuals in a network

To characterize communities inside a network

To characterize the evolution of a network

Key Application Areas
Healthcare Police & Military
Combating Epidemics Fighting cyber crimes
Mass Vaccination Fighting terrorism
Network-centric warfare
Social Media & E-Commerce
Scientific Research & Academic
Friend & Follow Recommendation
Collaboration
Know Your Customers Ranking scientific publications
Recommendation & viral marketing Ranking scientific authors
Web & Cyberspace Ranking publication venues
Search engine optimization Miscellaneous
Malware detection Computer-supported collaborative
Spam detection learning
Complex project management
SNA Applications: Healthcare

 Modeling the spread of infectious diseases

 Contact tracing during epidemic outbreak to identify

possible patients

 Identify and isolate super-spreaders

 Planning lockdown schedule

 Identify vulnerable population during vaccination

 Planning vaccination schedule, etc.

2003 SARS contact Network in Taiwan
https://doi.org/10.1007/978-1-4419-6892-0_15
 SNA Applications: Social Media

 Identifying friendship structures in online social media

 Recommending friends, and groups, or pages to follow

 Identifying information propagation patterns in social

networks

An Example Map of a Facebook Friendship Network

https://mathconceptions.wordpress.com/2012/01/16/application-snippet-friendship-and-influence-in-social-networks/
 SNA Applications: E-Commerce
 Customer profiling to Know the customers

 Product/Service recommendation based on customer profile

 Instigating viral marketing by pinpointing influential players

 “People like you buy”, “Frequently bought with this”, or

“Frequently browsed”, “Trending” are common buzzwords

Influential Communities in Social Network

https://towardsdatascience.com/influential-communities-in-social-network-simplified-fe5050dbe5a4
 SNA Applications:
Malware & SPAM Detection
Modeling malwares using Graph representations

System call graphs, malware similarity network,

etc. are typical examples

These graphs are large due to volume of networks

Malware detection through network analysis

SPAM detection

https://searchsecurity.techtarget.com/definition/malware
 SNA Applications:
Cybercrimes & Terrorism
 Online fraud, fake news propagation, cyber bullying/trolling,
sharing pornographic materials, etc. rising with growth of social
media

 Terrorists often use social media to communicate as well as to

brainwash innocent people

 These people often span across countries and use untraceable

communication devices

 Tracking cyber criminals in conventional methods are difficult due

to user anonymity, fake accounts, lack of cyber laws, etc.

 Social network analysis techniques help nabbing these criminals

9/11 Terrorist Network
http://www.orgnet.com/hijackers.html
 SNA Applications:
Network Centric Warfare
 Rising popularity of social network analysis
influence the military doctrines

 Various military think-tanks proposed for network-

centric warfare principle using social network
analysis methods

 It is claimed that Saddam Hussein has been

captured from his hideout exploiting network
analysis techniques

 It is also claimed US Navy Seal Team Six

assassinated Osama Bin Laden by tracking his
Network Centric Warfare secret hiding location in a similar manner
http://www.indiandefencereview.com/news/network-centric-warfare/
 SNA Applications: Scientific
Research & Academic Collaboration
 Scientific authors cite (refer) the works of other authors in their
publications to authenticate their claims

 Finding the dynamics of these citations attracted social scientists

 Various networks of scholarly articles may be formed exploiting

this relationship
 Paper-paper citation network
 Paper-paper co-citation network, etc.

 Various popular metrics are outcome of analysing these

networks.
 Publication related: H-index, i-10 index, g-index, etc.
 Venue related: impact factor, CORE rank, etc.

https://library.bu.edu/citedreferences
 SNA Applications: Scientific
Research & Academic Collaboration
 Scientific authors collaborate with one another to
improve research quality

 Various scholar networks may be formed using

these relationships
 Author collaboration network
 Author citation network
 Author co-citation network, etc.

 Information retrieved from these networks may

be used to measure authors’ research quality

https://scholarlykitchen.sspnet.org/2017/04/07/updated-figures-scal
e-nature-researchers-use-scholarly-collaboration-networks/
 SNA Applications: Computer-
supported Collaborative Learning
 Pedagogical process of observation where students learn
progressively through active group interaction using ICT

 SNA techniques used to extract relationship between

various actors (human and non-human) of CSCL

 Study insights are used to improve the students’ learning

outcome and user experience

https://www.semanticscholar.org/paper/The-Characteristics-of-the-Computer-Supported-
a-on-Hashim-Ismail/42176e6bf76dd15a2c9874e6fa8696e153a3f554
 SNA Applications:
Organizational Network Analysis
 Classic (linear) organization charts do not effectively
describe the real social network of an organization

 The informal networks, historically considered an annoying

problem and an indicator of undisciplined attitude of
workers, carry huge potential if applied suitably

 ONA provides information on how to improve

performance in the organization

 ONA represents the complete set of real relationships

between the players
 who is in touch with whom
 specific features of each player
https://www.i4cp.com/productivity-blog/what-organizational-network  Type and intensity of relationship, etc.
-analysis-is-and-how-it-benefits-companies
Network Representation:
Adjacency Matrix
An adjacency matrix for a graph is a square matrix of dimension such that each
element of A indicates the existence of an edge between the node and node
(also the weight of the corresponding edge in case of a weighted graph) in .

1 2 A(G1) = 1 8 2 A(G2) =
7
5
5 3 6 5
3 3 9
4 2
4
G1 G2
Network Representation:
Adjacency Matrix (Cont…)
advantages disadvantages
Consumes storage space, even
Easy to implement and follow if the graph is sparse

Addition, and removal of an edge Addition or removal of a node

require 𝑂(1) time require time

Query to an edge require 𝑂(1) time

Network Representation:
Adjacency List
An adjacency list for a graph is a collection of unordered lists such that each node
correspond to a list from the collection that indicates the set of neighbours of the node.
Every entry in an adjacency list for node in the graph is a node adjacent to node

1 2 3 1 2,8 3,3
1 2 1 8 2
2 7 2 1,8 3,5 4,6 5,7
1 3 4 5 5
5 3 6 5
3 3 1 2 5 3 1,3 2,5 5,9
3 9
4 2
4 2 5 4 4 2,6 5,2
G1 G2
5 2 3 4 5 2,7 3,9 4,2
Network Representation:
Adjacency List (Cont…)

Removal of edge require 𝑂(|

advantages disadvantages

Require 𝑂(|𝑉|+|𝐸|) space 𝐸|) time

Query to an edge require 𝑂(|

require 𝑂(1) time 𝑉|) time
Insertion of vertex and edge

Removal of vertex require 𝑂(|𝑉|

+|𝐸|) time
Network Types: Link-centric View
Unipartite Network
Consists of a vertex set V and an edge set E. There is no restriction on the
formation of edges between nodes of the network

Example: An organizational LAN, where nodes are the devices, and edges
are the local area links.

Used to model the situation when links can join any pair of nodes of the
network
Network Types: Link-centric View
Bipartite Network
Consists of a vertex set V that is divided into two sets V1 and V2 that
are disjoint and independent. Each edge of the network connects a
vertex in V1 to another vertex in V2

Example: An e-commerce user-product network. One part consists of

the users, the other part consists of the products, the links are based
on the basis of who bought what.

Generalization of Bipartite network is n-partite networks, where the

vertex set is partitioned into n number of part, an edges join a node
from one part with a node from another part.
Network Types: Link-centric View
Signed Networks
Consists of a set of nodes V, a set of edges E, and a function f : E → {+, −}
that assigns each edge a positive or a negative sign

+ Example: Consider a social media website that allows users to tag other
+
+ - +
users as friends or foes. The positive edges are friendship links and
+ negative links are between foes.
-
studied specifically in the context of balance and status theory which
determines the stability or existence of certain types of structural patterns
in a network.
Network Types: Node and Link-
centric View: Homogeneous Network
Consists of a set of nodes V, all of which are of same type, and a set of
edges E , all of which are of same type

Example: Follower-Followee network of any kind.

Nodes are the users of the platform (all nodes are of same type),
Links are the follower-followee link between these nodes (all links are
of same type)
Network Types: Node and Link-centric
View: Heterogeneous Network
Consists of a set of nodes V and a set of edges E, and two
associated mapping functions, fv and fe, for nodes and edges,
respectively. fv maps a node to a node type and fe maps an edge
to an edge type

 Example: We consider a specific instance of Twitter network like the figure.

 Two types of nodes:
 Twitter Users
 Tweet Posts
 Four types of edges representing four types of relations between these nodes:
 .Posted-by (User – Post: Directed links)
 Followed-by (User – User: Directed links)
 Similar (Post – Post: Undirected links)
 Retweet (User – Post: Directed links)
Network Types: Node and Link-
centric View: Attributed Network
Consists of a set of nodes V and a set of edges E, and two
associated mapping functions, fv and fe, for nodes and edges,
respectively. fv maps a node to a node attribute vector and and fe CSE, IIT

maps an edge to an edge attribute vector

ECE, IIT CSE, NIT

Example: We consider a specific instance of Facebook network like the figure ECE, NIT
Nodes are some Facebook users CSE, IIT
CSE, NIT
Edges are given by Facebook friendship relationship between these
users
Node attributes are the users’ academic affiliations CSE, IIT ECE, IIT

There is no edge attribute in this network

The example is a node-attributed network

Network Types: Node and Link-centric
View: Multidimensional Network
A special type of multilayer network where each layer
represents a particular type of relationship among nodes

Example 1: A special instance of Twitter network:

Nodes are Twitter users in both layer
Layer 1 edges: user – user similarity links (based on mutual
interests) – Undirected links
Layer 2 edges: user – user follower-followee links – Directed
links

Note: In this example, each layer is node homogeneous.

 Network Types: Node and Link-centric
View: Multidimensional Network (Cont…)
 Example 2: Customer – product relationships as a multidimensional
complex network system
 Layer 1: One type of nodes, one type of edges:
 Nodes: Customers
 Edges: Customer Social Interaction – Undirected links
 Layer 2: Two types of nodes, two types of nodes
 Nodes:
 High performance Cars,
 Fuel efficient cars
 Edges:
 Car feature association – Undirected links
 Car preference association – Directed links
Inter-layer edges: two types of edges
 Customer – Car: Purchase decision – Undirected links https://www.semanticscholar.org/paper/Modeling-customer-preferences-using-ne
twork-in-Wang-Chen/539c7f0632041903521b8cbc42eabd27b8844673
 Customer – Car: Consideration decision – Undirected links
Network Types: Local View: Ego-
centric Network
A network of the form G(V, E, u), that corresponds to a node u ∈ V,
usually known as the ‘ego’, and consists of the node u as the central
node, the nodes that are connected directly to the node u, usually
known as the ‘alters’, and the induced subgraph for the same.

 Example: A subgraph of a Facebook Friendship Network:

 Ego node corresponds to a user,
 Alter nodes are his Facebook friends from different capacities
and affiliations.
Network Types: Temporal View:
Time-varying Network
Consists of a set of nodes V and a set of edges E where each edge ei,j ∈ E is represented by a
three- tuple ei,j = {vi, vj, tij}. Here, vi and vj are two end-points, and tij indicates the persistence
duration of the edge eij

Example: Person-to-person communication network over a span of time. The visible components
are snapshots of the network at different time instances.
Network Types: Generalized View:
Hypergraph
Defined by a set of nodes V and a set of edge or hyperedges E, where each hyperedge e connects
multiple nodes of the hypergraph
Example: A special representation of Coauthoship Network:
 Nodes are authors
 Papers are hyperedges connecting the coauthors of the paper
Popular Real-world Networks

Social Network Information Network

 Telephone call network  World Wide Web (WWW)
 Email message network  Citation network
 Film actor collaboration network Technological Network
 Academic co-authorship network  Electric power grids
 Networks of airline routes
Biological Network  Network of Railway Routes
 Protein-protein interaction  Electronic circuits
networks  Delivery networks of
 Genetic regulatory networks post-office/Courier
 Neural networks  The Internet
 Metabolic networks Language Network
 Food Web  Network formed by using the persons
 Cell signalling networks speaking a particular language
Levels of Social Network Analysis:
Microscopic Level
We begin by analyzing how a pair of nodes interacts and gradually trace the interactions at the
group level or subgraph level.
Dyadic level  Interaction patterns among two nodes
Examined properties: homophily, reciprocity, social equality, mutuality, etc.
Derived global statistics: assortativity, mixing coefficient, etc.

Triadic level  Interaction patterns among three nodes

Examined properties: triadic closure
Derived network properties: Clustering Coefficient, local bridges, etc.

Ego-centric circles  Interaction pattern between ego node with its alters
Levels of Social Network Analysis:
Mesoscopic Level
 Mesoscopic analysis is an intermediary between microscopic and macroscopic analyses, which mostly
deals with a subset of the entire population.
 Communities  Formed due to frequent interactions among homogeneous nodes in a network
 Within a community, the nodes exhibit a particular kind of dynamicity
 Across communities, the dynamic behaviour differs
 Network Motifs  Subgraphs that repeat themselves frequently within or across a network
 Highly effective in capturing functional properties in a network

Undirected motifs with size 4 and their

names
Levels of Social Network Analysis:
Macroscopic Level
At macroscopic level, we deal with the entire network as a whole and try to understand the micro-
level dynamics by exploring the overall graph property.
Features of Interest :
Connectedness,
Diameter or Average path length,
Degree Distribution,
Edge Density, etc.

Example:
We find that the diameter of a network is too small network may look like a star, or a clique
We further find that overall edge density is too high network looks like a clique
Graph Visualization Tools
 Web-based tools  Standalone tools

 Pollinode: https://www.polinode.com (Non-open  NDlib-Viz:

source application) https://ndlib.readthedocs.io/en/latest/viz/ndlib-viz.h
tml
(Open source application)
 NodeGoat: https://nodegoat.net (Non-open source
application)  CytoScape: https://cytoscape.org (Open source
application)
 Linkage: https://linkage.fr (Open source application)
 Gephi: https://gephi.org (Open source application)
 EchoDemo: https://osome.iuni.iu.edu/demos/echo
 Vizster: http://vis.stanford.edu/jheer/projects/vizster
(Non-open source application)
(Open source application)

 Palladio: https://hdlab.stanford.edu/palladio (Open  SparklingGraph: https://sparkling-graph.github.io

source application) (Open source application)