2014 Fourth International Conference on Communication Systems and Network Technologies

Social Network Analysis Tools

Nadeem Akhtar
Department of Computer Engineering, Zakir Husain College of Engineering & Technology
Aligarh Muslim University
Aligarh, India
[email protected]

Abstract. Social networks like Facebook, Twitter, and Google+ In this paper, some graph analysis tools for the analysis of
are most visited domains on the Internet. They contain huge data large online social networks are discussed and compared. We
about the users and the relationships among them. To analyze have considered four tools namely Gephi, Networkx, IGraph
and mine useful information from these huge social network and Pajek. Such comparative study of social network analysis
data, special graph based mining tools are required that can
tools has already been done earlier [4][5]. We have added
easily model the structure of the social networks. A number of
more comparative results concentrating on efficiency,
such analysis tools are available with their own features and
benefits. Choosing an appropriate tool for a particular task is visualization and graph features.
difficult to decide. This work present a comparative analysis of II. SOCIAL NETWORK ANALYSIS
four social network analysis tools- Networkx, Gephi, Pajek,
IGraph based on platform, execution time, Graph types, Social network analysis (SNA) is the methodical analysis
algorithms complexity, input file format and graph features. of social networks. Social network analysis views social
(Abstract) relationships in terms of network theory, consisting
of nodes (representing individual actors within the network)
Keywords: social networks analysis; SNA tools; Gephi; Pajek; and ties (which represent relationships between the
Networkx; IGraph (keywords) individuals, such as friendship, kinship, organizational
position, sexual relationships, etc.) [6].
I. INTRODUCTION
Analysis tasks of social networks includes following:
A Social network is defined as a network of relationships x Discovering the structure of social network
or interactions, where the nodes consist of people or actor, and x Finding various attribute values for the network- Ex.
the edges or archs consist of the relationships or interactions radius, diameter, centrality, betweenness, shortest
between these actors [1]. Social networks and the techniques paths, density etc
to analyse them existed since decades [2]. There can be x Finding communities in the social network
several type of social networks like email network, telephone x Visualizing the whole or part of the social network
network, collaboration network. But recently online social Several works has been done on various social networks to
networks like Facebook, Twitter, LinkedIn, MySpace etc have analyse and discover various kinds of relationships and
been developed which gained popularity within very short information [7][8][9][10].
amount of time and gathered large number of users. Facebook
A. Kinds of Network Analysis
is said to have more than 500 million users in 2010 [3].
The field of social networks and their analysis has evolved There are two basic kinds of social network analysis, ego
from graph theory, statistics and sociology and it is used in network analysis, and complete network analysis.
several other fields like information science, business Ego network analysis is concerned with analysis of
application, communication, economy etc. Analysing a social individual nodes. A network can have as many egos as nodes
network is similar to the analysis of a graph because social in the graph. Egos can be persons, organizations or whole
networks form the topology of a graph. Graph analysis tools society. In ego network analysis, individual behaviour and its
have been there for decades. But they are not designed for variation is mined and described.
Complete network analysis is concerned with the analysis
analysing a social network graph which has complex
of all the relationships among a set of nodes. Techniques such
properties. An online social network graph may be very large.
as subgroup analysis, equivalence analysis and measures like
It may contain millions of nodes and edges. Social networks
centrality (closeness, degree, and betweenness) all require
are dynamic i.e. there is continuous evolution and expansion. complete networks [11].
A node in social network usually has several attributes. There
are small and large communities within the social graph. Old
graph analysis tools are not designed to manage such large
and complex social network graph.

978-1-4799-3070-8/14 $31.00 © 2014 IEEE 388

DOI 10.1109/CSNT.2014.83
Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on December 09,2024 at 17:00:36 UTC from IEEE Xplore. Restrictions apply.
 B. Applications of Social Network Analysis be "anything" (e.g. text, images), Edges can hold arbitrary
Following are some applications of social network analysis: data (e.g. weights, time-series), Standard graph algorithms,
[12][13] Network structure and analysis measures etc [15].
x Identify new scientific trends becoming commercially Gephi is an interactive visualization and
viable, e.g. RFID, Genome sequencing, tissue exploration platform for all kinds of networks, dynamic and
engineering hierarchical graphs. Runs on Windows, Linux and Mac OS X.
x Analyse expert network, Co-authorship networks, co- Gephi are a tool for people that have to explore and
citation networks, patent networks understand graphs. Like Photoshop but for data, the user
x Measurement of success interacts with the representation; manipulate the structures,
x Ranking of trends, of authors, of companies shapes and colours to reveal hidden properties [16].
commercializing trend Pajek: A widely used Software for drawing networks, Pajek
x Analysing page importance Page Rank (Related to also has analytical capabilities, and can be used to compute
recursive in-degree computation), Authorities/Hubs most centrality measures, identify structural holes, block-
x Discovering Communities: Finding near-cliques model, and so on[17].
x Analysing Trust: Propagating Trust IGraph is a free software package for creating and
x Using propagated trust to fight spam: In Email, In manipulating graphs. It includes implementations for classic
Web page ranking graph theory problems like minimum spanning trees and
network flow, and also implements algorithms like
III. SOCIAL NETWORK ANALYSIS TOOLS community structure search. The efficient implementation
Social network analysis tools are used to identify, analyze, of IGraph allows it to handle graphs with millions of nodes
visualize or simulate nodes (organizations, or knowledge) and and edges. IGraph can be installed as libraries for C, R,
edges (relationship or interaction) from various types of input Python and Ruby [18]. Above four tools are compared on the
data including mathematical models of social networks. There following six criterion- platform, Graph types, algorithm time
are several tools available for analysis of social networks. The complexity, graph layout, graph input file format, graph
International Network for Social Network Analysis (INSNA) features
maintains a large list of software packages and libraries [14]. Dataset for SNA tools comparisons: We have used Slashdot
We have selected four analysis tools for comparison dataset [19]. It contains 77317 nodes and 982787 edges
namely Gephi, Networkx, IGraph and Pajek. The selection is (Directed). Slashdot is a technology-related news website
based on several facts. All four selected software are freely features user-submitted and evaluated news stories about
available for use and they can handle large graph size. science and technology related topics.
Network analysis tools are either GUI based or IV. SOCIAL NETWORK S OFTWARE TOOLS C OMPARISONS
packages/libraries which can be used in a programming
language. Gephi and Pajek are GUI based network tools A. Comparison Based on Platform
whereas Networkx and IGraph are packages based tools. Social network analysis tools Pajek and Gephi are stand
Following are the brief detail about each of these four tools- alone software, Networkx and IGraph are libraries.
Networkx is a Python language software package for the Pajek/gephi runs on windows platforms and Networkx use
creation, manipulation and the study of structure and functions python library and IGraph use python/r/c library for social
of the complex networks. With this tool you can load and network analysis. Networkx, IGraph or Pajek can handle more
store networks in standard data formats, can generate many than one million nodes and Gephi can handle up to 150000
types of random and classic networks, analyze network nodes.
structure, build network models, draw networks, and much
more. Networkx has many features like language data
structures for graphs, dIGraphs, and multIGraphs. Nodes can
T ABLE I GENERAL C OMPARISON OF TOOLS

Software NETWORKX IGRAPH GEPHI PAJEK

TYPE LIBRARY LIBRARY STAND ALONE STAND ALONE
PLATFORM PYTHON PYTHON\R\C WINDOWS WINDOWS
LIBRARY
COMPUTATIONAL TIME FAST FAST FAST MEDIUM
NO. OF NODES 1 MILLION 1 MILLION 0.15 MILLION 1 MILLION

389

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on December 09,2024 at 17:00:36 UTC from IEEE Xplore. Restrictions apply.
 B. Comparison Based on Network Types them software have circular or spring layout. Gephi provide
In social network Analysis there are four types of network user friendly layout capability for user. Gephi provide
Graph [6]. capability like Photoshop where users have many additional
In a one-mode network, each vertex can be related to each facilities. Force layout is another famous layout algorithm
other vertex. In one mode network we have only one set of given by Gephi software.
nodes and ties are connected to these nodes. In a two-mode For user friendly visualization we can use Gephi software
network, vertices are divided into two sets and vertices can but it cannot handle large or complex graphs. For large and
only be related to vertices in the other set. complex network we can use either Networkx or IGraph.
Two mode network Graph are a particular type of networks IGraph or Networkx provide support for many other tools for
with two sets of nodes and ties are only established between visualization. For small dataset we can use standalone
nodes belonging to different sets. Techniques for analysing software (Gephi, Pajek) because we can handle standalone
one-mode networks cannot always be applied to two-mode software easily.
networks without modification or change of meaning. Special TABLE III GRAPH LAYOUT SUPPORTED BY T OOLS
techniques for two-mode networks are very complicated. We
can create two one-mode networks from a two-mode network Layout Networkx IGraph Pajek Gephi
In Multi relational network there will be multiple kinds of Circular Yes Yes Yes Yes
layout
relations between nodes. Nodes may be closely-linked in one
Random Yes Yes Yes No
relational network, but distant in another.
layout
In temporal networks (dynamic graphs) networks can Spectral Yes No No No
change over time. The lines and vertices in a temporal layout
network should satisfy the consistency condition: if a line is Spring layout Yes Yes Ye s Yes
active in time t then also its end-vertices are active in time t. Graphviz Yes No No No
For one mode or two mode network analysis we can use layout
any of software tools but for multi relational network graph Kamanda No Yes Yes No
we have only Pajek software tools .for temporarily network kawai
graph we have Networkx and Pajek tools. Fruchterman No Yes Yes No
reingold
T ABLE II NETWORK TYPES SUPPORTED BY TOOLS Force Atlas No No Yes No
layout
Graph type Networkx IGraph Gephi Pajek
D. Comparison Based on Algorithm Time Complexity,
1-Mode network Yes Yes Yes Yes Input File Formats and Graph Features
2-Mode network Yes Yes Yes Yes Social network analysis software has many algorithms for
Graph graph features. We compare these tools for each feature based
Multirelational No No No Yes on algorithms complexity.
network Graph IGraph and Networkx have algorithms for maximum
Temporarily Yes No No Yes number of features. Based on algorithms complexity we can
network Graph
say that IGraph is more useful software compare to other
softwares. IGraph provide efficient algorithms for page rank,
C. Comparison Based on Graph Layout all types of centrality, density, MST and shortest path.
In social network analysis we have many layout algorithms. Comparisons among the four social networks on the basis
IGraph or Pajek have most famous and recent layout of time complexity, input file formats and graph features are
algorithms Fruchterman Reingold or Kamanda Kawai. All of shown in Table IV, V and VI respectively.

T ABLE IV T IME C OMPLEXITIES

FEATURES NETWORKX IGRAPH GEPHI PAJEK

ISOMORPHISM O(n2) EXP NA NA
CORE m=no. of lines O(M) O(M) O(M) O(M)
CLIQUES O(|V|/(log)2) O(3|V|/3) O(N)
SHORTEST PATH O(|V|.|E|) O(|V|+|E|) O(|V|+|E|) O(|V|+|E|)
CLUSTRING O(V) NA O(V) NA
ALL SIMPLE PATH O(|V|+|E|) O(|V|+|E|) NA NA
CLOSENESS O(n.|E|) O(n.|E|) NA NA
CENTRALITY
DENSITY O(n3) O(1) NA NA
MST NA O(|V|+|E|) NA NA

390

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on December 09,2024 at 17:00:36 UTC from IEEE Xplore. Restrictions apply.
 CYCLES O((|V|+|E|).c+1) NA NA NA
PAGERANK NA O(|E|) O(E) NA
BETWEENNESS NA O(|V|.|E|) NA NA
EIGENVECTOR NA O(|V|+|E|) NA NA

T ABLE V FILE FORMATS S UPPORTED BY TOOLS

INPUT FILE FORMAT NETWORKX IGRAPH GEPHI PAJEK

.NET FORMAT YES YES YES YES
.GML YES YES YES NO
.GRAPHGML YES YES YES NO
EDGESLIST(.TXT) YES YES NO NO
EDGESLIST(.CSV) Yes Yes No No
.DOT YES YES YES NO
.PAJEK NO YES NO YES
.DAT NO NO NO YES
ADJACENCY LIST YES YES NO NO
.GDF No No Yes No
GRAPH DB No Yes Yes No

T ABLE VI VARIOUS GRAPH FEATURES SUPPORTED BY TOOLS

GRAPH FEATURE NETWORKX IGRAPH PAJEK GEPHI

APPROXIMATION YES NO NO NO
ASSORTIVITY YES YES NO NO
CENTRALITY YES YES YES YES
NETWORK DIAMETER YES YES YES YES
CLUSTRING YES YES YES YES
FLOW YES YES YES NO
COMMUNITIES YES YES NO YES
COHESION NO YES NO NO
BLOCK MODELING YES NO YES NO
DENDROGAM YES YES YES NO
CLIQUE YES YES YES NO
PAGE RANK YES YES YES YES
BFS YES YES YES NO
DFS YES YES YES NO
HITS YES YES YES YES
DENSITY YES YES YES YES
CORE YES YES YES NO
ISOMORPHISM YES YES YES NO
PARTITION NO NO YES NO
POWERLAW YES YES NO NO
MST YES YES YES NO
BI PARTILE YES YES YES NO
BRIDGE YES YES YES NO
DYAD YES YES YES NO
HITS YES NO NO YES

T ABLE VII E XECUTION T IME FOR V ARIOUS FEATURES

SNA Features Networkx IGraph Gephi Pajek

Load time 54.67 sec. 3.707 sec 29 sec 3 sec
Degree centrality 58.57 sec 6.199 sec 4 sec 2 sec
Graph degree 60.87 sec 6.22 sec 4 sec 2 sec
Page rank 120.78 sec 9.81 sec 10 sec No

391

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on December 09,2024 at 17:00:36 UTC from IEEE Xplore. Restrictions apply.
 Hits 57.23 sec 15.43 8 sec No
Cliques 66.98 sec 9.35 sec Na No
Density 58.94 sec 3.302 sec 4 sec No
Modularity 81 .4 sec 9 sec 30 sec 6 sec
Network diameter 35 sec 3.51 sec 120 sec No
Core 65.84 sec 6.532 sec No 1 sec
Cohesion No 8.943 sec No No
Clustering coefficient 3303.99 sec 1800 sec 1200 sec 108 sec
Hub 76.57 5.831 sec 3 sec No
Authority Array is to big 6.783 sec 3 sec No

REFERENCES
V. CONCLUSION [1] Measurement and Analysis of Online Social Networks by Alan Mislove,
Massimiliano Marcon,Krishna P. Gummadi, Max Planck Institute for
Stand alone software is very useful for graph Visualization Software Systems
(up to a maximum of few thousands of nodes), data format [2] For a historical overview of the development of social network analysis,
conversion. IGraph is fastest tools that provide most of graph see: Carrington, Peter J. & Scott, John (2011). "Introduction". The Sage
Handbook of Social Network Analysis. SAGE. p. 1. ISBN 978-1-84787-
features and handle large and complex network. Libraries 395-8.
(Networkx or IGraph) are more useful for tasks involving [3] http://www.facebook.com/notes/facebook/500millionstories/409753352
millions of nodes and for operations such as the union and the 130
difference between sets of nodes or for the clustering. Stand [4] David Combe, Christine Largeron, El˝od Egyed-Zsigmond and Mathias
Géry, “A comparative study of social network analysis tools”,
alone software are easy to use and easy to learn so for International Workshop on Web Intelligence and Virtual Enterprises 2
beginner Pajek and Gephi is suitable software. For complex (2010)
dataset and research purpose we can use Networkx and [5] M A huisman, MAJ van duijn, “ Software for social network analysis”,
pages 270-316, 2004
IGraph software. [6] Graph and Network Analysis Dr. Derek Greene Clique Research
For one mode or two mode network analysis we can use Cluster, University College Dublin, Web Science Doctoral Summer
any of software tools but for multi-relational network graph, School 2011
[7] Monclar, Rafael Studart, et al. "Using social networks analysis for
we have only Pajek software tools. For temporarily network collaboration and team formation identification." Computer Supported
graph we have Networkx and Pajek tools. Cooperative Work in Design (CSCWD), 2011 15th International
All of the software can handle .Net file format. But mostly Conference on. IEEE, 2011.
[8] Nadeem Akhtar, Hira Javed, Geetanjali Sengar, "Analysis of Facebook
data are present in .txt format. We can easily understand or Social Network", IEEE International Conference on Computational
handle .txt format or many websites provides data set in .txt Intelligence and Computer Networks (CICN), 27-29 September, 2013,
format so we can use IGraph or Networkx software tools. We Mathura, India
have many file format conversion software they can [9] Zelenkauskaite, Asta, et al. "Interconnectedness of complex systems of
internet of things through social network analysis for disaster
convert .txt file format into .Net file format but these software management." Intelligent Networking and Collaborative Systems
can handle only small size file. So for large size data set we (INCoS), 2012 4th International Conference on. IEEE, 2012.
use IGraph or Networkx software tools. [10] Li, Jianfeng, Yan Chen, and Yan Lin. "Research on traffic layout based
on social network analysis." Education Technology and Computer
IGraph provides mostly graph features and it also handle (ICETC), 2010 2nd International Conference on. Vol. 1. IEEE, 2010.
large and complex network. All of them software can compute [11] Online Social Networks: Measurement, Analysis, and Applications to
centrality, clustering coefficient, network diameter, page rank, Distributed Information Systems by Alan E. Mislove, Houston, Texas
[12] Social Networks Overview: Current Trends and Research Challenges”
density. But if we want to compute some specific feature we November 2010 Coordinated by the ―NextMEDIA CSA.
choose different software. If we want to compute Cohesion, [13] Business Application of Social Network Analysis BASNA-2013
we can Use Networkx or IGraph tools. If we want to compute www.basna.in
Bridge and dyad, we can use Networkx, IGraph and Pajek [14] International network of Social Network Analysis INSNA
www.insna.org
software tools. Gephi does not provide the facilities for dyad [15] Networkx http://Networkx.lanl.gov/index.html
or bridge computation. [16] Gephi https://gephi.org
IGraph and Pajek are faster software Tools compare to [17] Pajek vlado.fmf.uni-lj.si/pub/networks/pajek
[18] IGraph IGraph.sourceforge.net
others. But Pajek does not provide all graph features. So if we [19] snap.stanford.edu/data
want to analyse all graph features we can use IGraph software.
IGraph gives fastest result to almost all graph features. Load
time is minimum for Pajek software. Execution time for
centrality, page rank, graph degree and cliques is minimum in
IGraph software. So based on execution time IGraph is better
software.

392

Authorized licensed use limited to: Universidad Federal de Pernambuco. Downloaded on December 09,2024 at 17:00:36 UTC from IEEE Xplore. Restrictions apply.