Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
Predicting user behavior over social networks
P. Anusha#1, B. Gnanakoushik #2, D. Sai Kumar #3, K. Sravan Kumar #4,
A. Venkataramana #5 #1Asst. Professor, Dept of CSE, Qis College of Engineering and Technology, Ongole.
#2, #3, #4, #5 B.Tech., Scholars, Dept of CSE,Qis College of Engineering and Technology, Ongole,
ABSTRACT:
In recent years, online social networks such as Facebook, Twitter, LinkedIn are most popular
visited sites over the internet. Presently, there is a great interest in understanding and studying the
relationships among the users in social networks. Existing link prediction methods predicts the
links based on the topological structure features and the node attribute features but overlook the
benefit other features such as clicks, shares, likes, forwards and comments could bring to any
social network. To address this gap, we propose a link prediction method based on user actions
with the post which includes clicks, shares, likes, forwards and comments. In this paper, we
propose link prediction model combining user action metrics and topological structure metrics.
The proposed metric can bridge the gap between the existing methods and propose a new metric
for defining link prediction. This is a work in progress paper, further as future direction,
implementation of the proposed metrics on standard datasets by suitably training the classifiers is
the topic ofinvestigation.
Introduction:
Online social networks such as Google plus,
LinkedIn, Facebook and Twitter have huge
number of users and the number of users
isgrowing every year [1]. Such a
potentialand diverse network of users is
attracting researchers from both academia
and industry to study and understand the
utilization of these networks for economic
and social benefits [2]. Link prediction aims
at prediction of future links between theusers
[3]. Examples of link recommendation
includes “people you may know” on
Facebook, LinkedIn etc. Members of these
social networks site use these social network
platforms to communicate with others by
posting, commenting, liking, forwarding etc.
and in turn the owners of these social
networks generate huge revenue by
displaying advertisements on the candidate
nodes (profiles) based on their
preferenceorchoice (also termed as
recommender
systems in literature, a particular case of link
prediction problem) which drives ads to
reach more users [4]. In that way one can
think that the link recommendation or link
prediction provides double folded benefits for
the users as well as for the social networks.
Existing methods in link prediction mainly
focus on the link likelihood by considering
user attributes and topological structures, and
predicts the links with the highest linkage
likelihood [5]. But they overlook the benefit
of analyzing the other existing attribute
features these social networks provide such as
clicks, shares, likes and comments etc., to
predict links. This motivates the fact to
consider these already existing features and
predict links more aptly and accurately to
recommend or predict links and there is a
need for defining new metrics apart from the
topological features of the online social
networks. A social network generates huge
data dynamicallyinafastermannerwiththe
participation of millions of users worldwide
in various forms of social networks such as
follower network, followee network, citation
network, Wikipedia network, twitter,
recommendation social network etc. These
various types of networksrequire to
categorize their users, customers in different
groups to analyze, to find out popular users
for product selling statistics for upgrading
network performance, design and so on.
This categorization of users in various
groups based on different clustering criteria,
decision rules, hidden flow and finally it
leverages some general user activity
patterns, network access pattern. These
patterns are sometimes dependent on social
network platforms, flexibility and rich
features of the social network and finally this
shapes some general user behavior. Some of
these behaviors are directly computable from
the publicly available social network data
which are persistent behavior in our
conceptualization and some which may not
be revealed out from data directly. These are
regarded as non-persistent social network
user behavior. Nonpersistent behavior can be
indirectly computed from the hidden
contents such as web server access logs,
hidden work flowetc.
User Behavior Analysis
Investigation of user behavioranalysis isan
important research area in computational
social network for various reasons.
Identification of criminal activities, anomaly
detection, information diffusion in short
times, sentiment study of larger social
groups, popularity estimation, product rating
etc. belongs to behavior analysis in social
networks. Here data are collected froma
social network to analyze user behavior both
quantitatively as well as qualitatively. Our
proposed behavior analysis architecture has
been shown in figure 1 below. A. Locus
Standy of the Survey Our survey differs
from the existing surveys in the following
ways. Like [35], we discuss sources, causes
and aspects of social network user behavior,
and also include a detailed discussion of
behavior analysis. In addition, we include a
large collection of up- to-date social network
user behavior analysis methods under the
categories of statistical, classification-based,
knowledge based, soft computing,
clustering- based etc, rather than restricting
ourselves to only statistical approaches. We
also include several important research
issues and open challenges. Like [7] and [5]
we attempt to provide a classification of
various social network user behavior
analysis methods, systems and software
tools introduced till date in addition to a
classification of social network platforms
and their characteristics. In addition, we
perform detailed comparisons among these
methods. Furthermore, like [6], we have
provided a list of research issues and open
challenges. Unlike [8] our survey is not
restricted to only social network activity
classification and analysis. It includes a large
number of up-to-date methods and analysis
approaches. A detailed discussion on types
of social network datasets, gathering and
preprocessing has been included in our
survey. However, unlike [19], we include
ideas for analysis of user behavior, in
addition to providing a list of practical
research issues and open challenges. Unlike
[23], our survey is focused on socialnetwork
user behaviour, their characterization,
recognition and detection approaches,
methods and systems, and comparisons
among them.
The Problem of User behavior Analysis
The behavior is a qualitative term which
itself is a complex concept to express in
terms of mathematical formulae. This in our
understanding may be defined as frequent
social network activities performed by a user.
More clearly, this is the most frequent pattern
of user activity in a social network. In the
real-world, communication interactions
among social network users include
interaction patterns leading to which patterns
are most influential in the whole network for
users. It is important to determine these
influential patterns in the network. These
patterns indicate the overall behavior of the
social network. This is essential for social
network in which, the interaction patterns of
densely connected actors are involved.
Therefore, we are interested to find the most
influential interaction structures in the social
network community. For a given graph G
with vertices V and edges E. We need to
identify the most frequent social network
user interaction patterns P which are
embedded in G. Each unique pattern in the
social network is considered one behavior of
users in the social network. After finding the
frequent interaction patterns of the users in
the network, we need to correlate these
patterns of users to the social network
platform to qualitatively analyze the user
behaviour in that particular social network
platform.
Our Contributions
This paper provides a structured and broad
overview of the extensive research on social
network user behavior analysis. The
majorcontributions of this survey are listed
below:
(a) User behavior in online social networks
has been well studied, but the correlation of
this behavior across these networks has not
yet been addressed clearly. We tried to
correlate our behavior analysis of social
network users in a simple manner with
examples. (b) We have categorized user
behavior analysis in two broad perspectives:
persistent & non-persistent. This is a new
way of looking at social network user
behavior analysis research. (c) Most
existing surveys do not cover all the
approaches till date, for social network user
behavior analysis, but we have covered a
wide range of approaches available in
literature up to 2016. (d) Most existing
surveys avoid characterization and
representation of social network user
behavior, which are crucial in the
behavior analysis task. We present several
techniques to characterize the behavior from
available datasets and comparethem.
(e) In addition to discussing behavioranalysis
methods, we present several wayshow to
characterize online social network user
behavior, and also present a comparison
among the various analysis. (f) We
summarize tools used in various stepsfor
social network user behavioranalysis.
(g) We also provide a description of the
datasets used for analysis. (h) Finally, we
highlight several important research issues
and challenges from both theoretical and
practical.
Behavior Characterization
Social network user behaviour
characterization is to define behaviour
mathematically in an appropriate manner.
This mathematical behaviour representation
will help in finding out the appropriate
behaviour pattern from huge social network
data sets. This is a very complex
representation as human behaviour greatly
differs from social network user behaviour
which again varies from one social network
platform to another. Same user in different
social network platforms exhibit various
activity patterns which again varies fromuser
to user that shapes the general user
behaviour. Moreover, characterization of
behaviour in a computable form to model
the general behaviour poses difficulty in
terms of processing complexity, exact data
representation for the model input and so on.
People so far have tried to define the
composition of user behaviour in terms of
set of activities, patterns, way of
participation in the social network etc.
Various methods were employed in
characterization of user behaviour in various
social network platforms. Maria Kihl et. al.
2010, in their article, “Traffic analysis and
characterization of Internet user Behaviour”,
[21] analyzed traffic measurements from a
Swedish municipal broadband access
network and derived corresponding user
behaviour models. The paper focuses on
Internet usage in terms of traffic
patterns,volumes and applications. Also,
user activity characteristics, as session
lengths and traffic rate distributions, are
analyzed and modelled. Bruno Goncalves et
al. proposed a simple model for user
behaviour that includes finite priority
queuingandtimeresourcesthatreproduces
the observed social behaviour in their
research paper “Modeling Users Activity on
Twitter Networks: Validation of Dunbars
Number”[10] in 2011. They analyzed a
dataset of Twitter conversations collected
across six months involving 1.7million
individuals and test the theoretical cognitive
limit on the number of stable social
relationships. Radoslaw Brendel and Henryk
Krawczyk in 2011, proposed a behavioural
model based on user roles in social network
in their paper, “Primary role identification in
dynamic social networks”[4]. Social
network often forms very complex
structures that additionally change over time.
Description of actor roles in such structures
requires to take into account this dynamics
reflecting behavioural characteristics of the
actors. A role can be defined as a sequence
of different types of activities. Various types
of activities are modeled by pattern sub-
graphs, whereas sequences of these activities
are modelled by sequence diagrams. For
such defined roles, a role identification
procedure is proposed to assign each actor
primary role played in a dynamic social
network. Erheng Zhong et. al. in 2012 has
investigated user behaviour in composite
manners thereby meaning behaviour of one
user in various social network platforms.
Many previous works modelled user
behaviour from only one historical user log
[4]. They observed that many people are
members of several social networks at the
same time, such as Facebook, Twitter etc.
Importantly, their behaviour and interests in
different networks influence one another.
This produces an opportunity to leverage the
knowledge of user behaviours in different
networks, in order to alleviate thedata
sparsity problem, and enhance the predictive
performance of user modeling. They
formulated the problem to model multiple
networks as composite network knowledge
transfer. They first selected the most suitable
networks inside a composite social network
via a hierarchical Bayesian model,
parameterized for individual users. Then
they build topic models for user behaviour
prediction using both the relationships in the
selected networks and related behaviour data.
Francis T. Odonovan et. al. in 2013, in their
research work, “Characterizing User
Behaviour and Information Propagation on a
Social Multimedia Network”[9] studied
users’ online activities. One such activity is
the sharing of multimedia, the popularity of
which can vary dramatically. They discussed
initial analysis of anonymized, scraped data
from consenting facebook users, together
with associated demographic and
psychological profiles. They presented five
clusters of users with common observed
online behaviours, where these users also
show correlated profile characteristics.Ryan
A. Rossi studied dynamic behaviour in
social network graphs in their work in 2013
and published a research paper “Modeling
Dynamic Behaviour in Large Evolving
Graphs”[17]. They conducted Studies to
model and characterize the temporal
behaviours of individual nodes from a given
large time-evolving graph. They have also
investigated how to model the behavioural
transition patterns of nodes. They proposed
a temporal behaviour model that captures
the roles of nodes in the graph and how they
evolve over time. According to the model,
interpretable behavioural roles are
generalizable and computationallyefficient.
Their model is for (a) identifying patterns
and trends of nodes and network states
based on the temporal behaviour, (b)
predicting future structural changes, and (c)
detecting unusual temporal behaviour
transitions. Vasanthan Raghavan et. al
developed probabilistic models for user
activity in social networks by incorporating
the social network influence as perceived by
the user and published “Modeling Temporal
Activity Patterns in Dynamic Social
Networks”[24] research paper in 2013. Qiang
Yan et. al. published “Social network based
microblog user behaviour analysis”[23] in
2013. The influence of microblog on
information transmission is becoming more
and more obvious. By characterizing the
behaviour of following and being followed
as out-degree and in-degree respectively, a
microblog social network was built in this
paper. It was found to have short diameter of
connected graphs, short average path length
and high average clustering coefficient. The
distributions of out-degree, in-degree and
total number of microblogs posted present
power-law characters. The exponent of total
number distribution of microblogs is
negatively correlated with the degree of each
user. With the increase of degree, the
exponent decreases much slower. Based on
empirical analysis, they proposed a social
network based human dynamics model in
this paper, and pointed out that inducing
drive and spontaneous drive lead to the
behaviour of posting microblogs. Zhenhua
Wang et. al. in 2014 in their research paper
“Analysis of user behaviours by mining
large network data sets”[7] studied user
behaviour model. Understanding the
intelligence of human behaviours bymining
petabytes of network data, represents the
tendency in social behaviours research and
shows great significance on Internet
application designing and service expansion.
Meanwhile, the running mobile networks
that generate huge data can be the best social
sensor for these studies. This paper
investigates a practical case of mobile
network aided social sensing which
uncovers some features of users behaviours
in mobile networks by intelligently
processing the big data. The paper studies
the users behaviours with regard to
communication, movement, and
consumption based on large user data sets.
Reza Motamedi et. al. published
“Characterizing Group- Level User
Behaviour in Major Online Social
Networks”[15] in 2014. They conducted a
detailed measurement study to characterize
and compare the group level behaviour of
users in Facebook, Twitter and Google+.
They focused to capture user behaviour with
the following metrics: user connectivity,
user activity and user reactions. Furthermore,
they conducted temporal analysis on
different aspects of user behaviour for all
groups over a two-year period. This analysis
leads to a set of useful insights including: (i)
Facebook and Google+ users express their
opinion quickly whereas Twitter users tend
to relay a received post to other users and
thus facilitate its propagation. Despite the
culture of re-share among Twitter users, a
post by a Popular Facebook user receives
more re-shares than a post by a Popular
Twitter user. (ii) Added features in an
Online Social Network (OSN) can
signicantly boost the rate of action and
reaction among its users.Vasanthan
Raghavan et. al studied temporal user
behaviour in 2014 and published “Modeling
Temporal Activity Patterns in Dynamic
Social Networks”[14]. The focus of this
work is on developing probabilistic models
for temporal activity of users in
socialnetworks (e.g., posting and tweeting)
by incorporating the social network
influence as perceived by the user. Although
prior work in this area has developed
sophisticated models for user activity, these
models either ignore social network
influence completely or incorporate it in an
implicit manner. They overcome the non-
transparency of the network in the model at
the individual scale by proposing a coupled
hidden Markov model (HMM), where each
user activity evolves according to a Markov
chain with a hidden state that is influenced
by the collective activity of the friends of the
user.
Behaviour Recognition
From various observation of social network
user data, activities, web usage content etc a
perceived underlying user behaviour can be
identified. Many people tried to reveal user
behaviour in different social network
platforms. Marcelo Maia et. al in
theirresearch paper “Identifying User
Behaviour in Online Social Networks”[20] in
2009 has described about human behaviour.
They explained how to characterize the
different classes of user behaviour.
Traditionally, user behaviour
characterization methods, based on user
individual features, are not appropriate for
online networking sites. In these
environments, users interact with the
siteandwithotherusersthroughaseriesof
multiple interfaces that let them to upload
and view content, choose friends, rank
favourite content, subscribe to users and do
many other interactions. Different
interaction patterns can be observed for
different groups of users. In this paper, they
proposed a methodology for characterizing
and identifying user behaviours in online
social networks. They used a clustering
algorithm to group users that share similar
behavioural pattern. Next, they have shown
that attributes that stem from the user social
interaction, are good discriminators and
allow the identification of relevant user
behaviours. K. R. Suneetha R.
Krishnamoorthi, in 2009 in their research
paper “Identifying User Behaviour by
Analyzing Web Server Access Log
File”[1]applied data mining techniques on
web server log data to find out web access
pattern behaviour. They explained that
Personalization for a user can be achieved
by keeping track of previously accessed
pages. These pages can be used to identify
the typical browsing behaviour of a user and
subsequently to predict desired pages. By
determining access behaviour of users,
important links can be identified to improve
the overall performance of future accesses.
The information gathered through Web
mining is evaluated by using traditional data
mining parameters such as clustering and
classification, association, and examination
to find out sequential user activity pattern.
Sofia Angeletou, Matthew Rowe Harith
Alani studied online user behaviour in
online communities in 2012 published the
article “Modelling and Analysis of User
Behaviour in Online Communities”[23].
They used statistical analysis,combined
with a semantic model and rules for
representing and computing behaviour in
online communities. They applied this
model on a number of forum communities
from Boards.ie to categorise behaviour of
community members over time, and report
on how different behaviour compositions
correlate with positive and negative
community growth in these forums. Aoying
Zhou, Weinign Qian Haixin Ma in 2012
studied collective behaviour in social
networks and published “Social Media Data
Analysis for Revealing Collective
Behaviours”[3]. They established that given
sufficient social media data, users collective
behaviours could be sensed, studied, and
even predicted in a certain circumstance.
They conducted experiments on data from
two social network platforms, i.e. Twitter,
and Sina Weibo. Collective behaviours are
actions of a large amount of various people,
which are neither conforming nor deviant.
Various collective behaviours are studied in
the context of social media. Their studies
show that there are various information flow
patterns in social media, some of which are
similar to traditional media such as
newspapers, while others are embedded
deep in the social network structure. The
evolution of hotspots is highly affected by
external stimulation, the social network
structure, and individual users activities.
Furthermore, social media tends to be
immune to some repeated similar external
stimulations.
Behaviour Prediction
So far we have discussed social network
user behaviour characterization &behaviour
recognition in various platforms.Various
methods & algorithms used in above
mentioned tasks have been reported. Next
we are to report relevant research works in
the line of behaviour prediction. By
analyzing various social network data sets
and by observing previous activities, future
behaviour can be predicted as mentioned
below. M. Vasudevan and M. Tamilarasi
conducted a survey on users collective
behaviour in social network in their research
paper “Collective behaviour prediction in
social media: A survey”[19] in 2012.
Collective behaviour refers to how
individuals behave when they are exposed in
to a social network environment. This
collective behaviour gives the opportunity to
predict online behaviours of users provided
behaviour information of some actors in the
network are available. This work studies
how networks in social media can help
predict some sorts of human behaviour and
individual preference. This can help
understand the behaviour patterns presented
in social media, as well as other tasks
likesocial networking advertising and
recommendation. Adam Sadilek in
hisdoctoral thesis [1], studied “Human
behaviour at large scale in various social
networks”. The underlying theme of this
thesis is the unification and data mining of
diverse, noisy, and incomplete sensory data
over large numbers of individuals. They also
found that raw sensory data linked with the
content of users online communication,
theexplicit as well as the implicit online
social interactions, and interpersonal
relationships are rich information sources
upon which strong machine learning models
can be built. Examples where such models
apply include understanding human
activities,predicting
peoples location and social ties from their
online behaviour, and predicting the
emergence of global epidemics from day-to-
day interpersonal interactions. Jihang Ye et.
al. in 2012 in research work “Whats Your
Next Move: User Activity Prediction in
Location-based Social Networks”[12]
exploited the check-in category information
to model the underlying user movement
pattern. They proposed a framework which
uses a mixed hidden Markov model to
predict the category of user activity at the
next step and then predict the most likely
location given the estimated category
distribution. The advantages of modelling
the category level include a significantly
reduced prediction space and a precise
expression of the semantic meaning of user
activities. Michal Kosinski et. al in their
study “Manifestations of user personality in
website choice and behaviour on online
social networks”investigated the user
behaviour in social networks. This work,
based on a sample of million users,
examines how users behaviour in the online
environment, captured by their website
choices and Facebook profile features,
relates to their personality [24]. Results
show that there are psychologically
meaningful links between users personalities,
their website preferences and Facebook
profile features. They summarized that
predicting a users personality profile can be
applied to personalize content, optimize
search results, and improve online
advertising. This concept can be applied to
predict user?s behaviour in social network.
Sharad Goel and Daniel G. Goldstein
studied prediction of individual behaviour [4]
in 2013. They employed acommunications
network of over 100 million people to
forecast highly diverse behaviours from
patronizing an online department store. They
found that social network data are strongly
informative in identifying individuals who
are most likely to undertake actions, and that
in identifying such individuals, social data
generally improve the predictive accuracy of
base-line models. They published the
research article “Predicting Individual
Behaviour with Social Networks”. Tad
Hogg et. al. published “Stochastic Models
Predict User Behaviour in Social Media”in
2013. User response contributes content in
online social media depends on many factors
[2]. These include how the site lays out new
content, how frequently the user visits the
site, how many friends the user follows, how
active these friends are, as well as how
interesting or useful the content is to the user.
They presented a stochastic modeling
framework that relates a user behaviour to
details of the sites user interface and user
activity. They also described a procedure for
estimating model parameters from available
data. They applied the model to study
discussions of controversial topics on
Twitter, specifically, to predict how
followers of an advocate for a topic
responded to the advocates’posts.
Methods & Approaches For Behaviour
Analysis
Social network user behaviour analysis has
been discussed in this survey in three
subcategories such as behaviour
characterization, behaviour recognition and
behaviour prediction. They found that,
despite the relative randomness and lesser
commitment of structural relationships in
online forums, users community joining
behaviours display some strong regularity.
So here they have described that behaviour
in SN user are forum or platform dependent.
There is no such common characterization
of user behaviour for all kinds of social
network forums. Ahn Dae-Yong & Randal
Watson in 2010 in their research paper, “A
Dynamic Model of User Behaviour in a
Social Network Site”[2], estimated a
dynamic model of user behaviour in a social
network site using unique data on the daily
login activity of MySpace.com members.
Using a social network site, it may have two
effects (interactions, flow of utilities over
time) of: (1)maintaining the quality of
interactions with an existing group of friends,
and (2) expanding the quantityof interactions
with the addition of new friends to the
network. Since both of these effects will
yield a higher flow of utilities in future, it is
appropriate to analyze usage of a social
network site with a model that takes into
account consumers forward-looking
behaviour. David John Robinson in his
doctoral thesis in 2010, “Cyber-Based
Behavioural Modeling”[5] proposes a noble
approach to the modelling and analysis of
behaviours based on a users’ cyber activities
(communication, workshop, play etc). They
presented a methodology to identify, extract,
and analyze cyber behaviours providing the
foundation for cyber-based behavioral
modeling. Methods are implemented to
characterize, predict, and detect change in
individual and group behaviours. Kartik
Bommepally et. al. in their research paper
“Internet Activity Analysis Through Proxy
Log”[15] in 2010 have analyzed the
behaviourofInternetusersbehindaproxy.
They also analyzed users Internet usage
pattern. User usage behaviour can be
analyzed using server log data. Laszlo
Gyarmati and Tuan Anh Trinh, in their
research paper “Measuring User Behaviour
in Online Social Networks”in 2010,
presented a large scale measurement
analysis of user behaviour in some popular
OSNs [17]. A measurement framework has
been created in order to observe user activity.
More than 500 PlanetLab nodes across the
globe have been used for measurement,
monitoring more than 80,000 users for six
weeks by downloading more than 100
million profile pages. Based on the
measurements, they addressed two key
issues of online social networks:
characterization of user activities and usage
patterns in the examined OSNs. Pietro
Panzarasa et al. in their research Patterns
and Dynamics of Users [2] “Behaviour and
Interaction: Network Analysis of an Online
Community”2010, draws on longitudinal
network data from an online community to
examine patterns of user behaviour and
social interaction, and infer the processes
underpinning dynamics of system use. The
online community represents a prototypical
example of a complex evolving social
network in which connections between users
are established over time by online messages.
Huiqi Zhang in his doctoral thesis, 2010,
“SOCIOSCOPE: Human Relationship and
Behaviour Analysis”[13] , proposed a model
for social network, relationship and human
behaviour analysis based on mobile phone
call detail records. Because of the diversities
and complexities of human social behaviour,
one technique cannot detect different
features of human socialbehaviours.
Therefore he used multiple probability and
statistical methods for quantifying
socialgroups, relationships and
communication patterns, for predicting
social tie strengths and for detecting human
behaviour changes. Mark Granovetter, State
University of New York at Stony Brook, in
his article, “Models of collective behaviour”,
developed a model for situations where
actors have two alternatives and the costs
and/or benefits of each depend on how many
other actors choose which alternative [22].
The key concept is that of threshold the
number or proportion of others who must
make one decision before a given actor does
so. Beginning with a frequency distribution
of thresholds, the models allow calculation
of the ultimate or equilibrium number
making each decision. The stability of
equilibrium resulting against various
possible changes in threshold distributions is
considered. Stress is placed on the
importance of exact distributions for
outcomes. Groups with similar average
preferences may generate very different
results; hence it is hazardous to infer
individual dispositions from aggregate
outcomes or to assume that behaviour was
directed by ultimately agreed-upon norms.
Suggested applications are riot behaviour,
innovation and rumour diffusion, strikes,
voting, and migration. Issues of
measurement, falsification, and verification
are discussed in thestudy.
Datasets For Behaviour Analysis
There are several datasets available for
social network user behaviour analysis
methods. The most commonly used datasets
are given below. Capturing and
preprocessing of high speed social network
traffic is essential for practical analysis of
social network user behaviour. Different
tools are used for capturing and analyzing of
social network traffic data for analysis.
Various types of data are used for behaviour
analysis. Krishna Das et. al. in their research
paper [16] used some similar data sets as
mentioned below to study the social network
users’ group dynamic in community
detection problem. Types of data sets used
in behaviour analysis are: (i) Profile Data: A
member provides to social network
platforms for registering itself. Statistics of
the surveyed paper (ii) Posted Data: The
data which is explicitly posted by the
members such as messages, comments and
blog entries.These data are highly sparse in
nature and hence processing of these type of
datasets pose tremendous challanges. But
these are highly essential in persistent
behaviour analysis such as sentiment
analysis, product rating, reputation & trust
estimation etc. (iii) Derived Data: The data
which is derived or mined by correlating
other information. Log server data, web
traces, browsing history, time stamp data etc
belong to this category and are used in non-
persistent behaviouranalysis.
Conclusion:
we studied user behaviour in OSNs from
three perspectives- behaviour
characterization, behaviour recognition
&behaviour prediction. The study considered
users? various activities in the social
network platforms, such as connections and
interactions, traffic activities, positional
arrangements of the users in the network etc.
We reviewed the existing representative
schemes and also provided potential future
directions. We also focused on twodifferent
broad research lines such as persistent and
non-persistent behaviour of social network
users. This survey will help in understanding
chronological advancement of user
behaviour analysis research as many of the
published works till 2016 have been included
in thisreport.
References
[1] Adam Sadilek 2012, Modeling human
behavior at a large scale, Doctoral
Dissertation, University of Rochester
Rochester, NY,USA.
[2] Ahn Dae-Young and Watson Randal
2010, A Dynamic Model of User Behavior in
a Social Network Site, SSRN
ELECTRONIC JOURNAL, 1-
32,http://dx.doi.org/10.2139/ssrn.1591102.[3]
Aoying Zhou, Weinign Qian & Haixin Ma
2012, Social Media Data Analysis for
Revealing Collective Behaviors, In
Proceedings of the 18th ACM SIGKDD
international conference on Knowledge
discovery and data mining, NY, August 12 ?
16,1402-1402.
[4] Chetan Mahajan and Preeti Mulay 2015,
E3: Effective Emoticon Extractor for
Behavior Analysis from Social Media, In
2nd International Symposium on Big Data
and Cloud Computing (ISBCC?15),Chennai,
India, March 12-13, 610 -616.
[5] David John Robinson 2010, Cyber-based
behavioral modeling, Doctoral Dissertation ,
Dartmouth College Hanover, NH, USA ,
ISBN:978-1-124-27065-4.
[6] Dziczkowski G., Wegrzyn-Wolska, K. ;
Bougueroua, L. 2013, ?An Opinion Mining
Approach for Web User Identificationand
Clients’ Behaviour Analysis?, In Fifth
International Conference on Computational
Aspects of Social Networks (CASoN),
North Dakota USA, August 12-14, 79 - 84,
IEEEXplore.
[7] Edoardo Serra and V. S. Subrahmanian
2014 , A Survey of Quantitative Models of
Terror Group Behavior and an Analysis of
Strategic Disclosure of Behavioral Models,
IEEE Transactions on Computational Social
Systems, 1(1), 66-88,IEEE.
[8] Erez Shmueli, Vivek K. Singh, Bruno
Lepri, and Alex Pentland, Sensing 2014,
Understanding, and Shaping Social
Behavior , IEEE Transactions on
Computational Social Systems, 1(1), 22 -34.
[9] Francis T. O’Donovan, Connie Fournelle,
Steve Gaffigan, Oliver Brdiczka, Jianqiang
Shen, Juan Liu and Kendra E. Moore
2013, ?Characterizing User Behavior and
Information Propagation on a Social
Multimedia Network, In Int. Workshop on
Social Multimedia Research (SMMR), San
Jose, CA, USA, July 15–19, 1-6,IEEE.
[10] Gonalves B, Perra N, Vespignani A,
Modeling Users’ Activity on Twitter
Networks: Validation of Dunbar’s Number
PLOS ONE, 6(8):e22656.
[11] Homa Hosseinmardi, Amir
Ghasemianlangroodi, Richard Han, Qin Lv,
Shivakant Mishra and Amir
Ghasemianlangroodi 2014, Towards
Understanding Cyberbullying Behavior in a
Semi-Anonymous Social Network,
https://arxiv.org/pdf/1404.3839.pdf, August
21.
[12] Hong Cheng, Jihang Ye and Zhe Zhu
2013, What’s Your Next Move: User
Activity Prediction in Location-based Social
Networks, In SIAM International
Conference on Data Mining (SDM), Texas,
USA, May 2-4,171-179.
[13] Huiqi Zhang, Ram Dantu, and Joao W.
Cangussu 2011, Socioscope: Human
Relationship and Behavior Analysis in
Social Network, IEEE Systems, Man, and
Cybernetics Society, 41(6), pp. 1122 -1143.
[14] Ines Brosso and Alessandro La Neve
2012, Adaptive Security Policy Using User
Behavior Analysis and Human Elements of
Information Security, Fuzzy Logic -
Emerging Technologies and Applications”,
Chapter 6, book edited by Elmer P. Dadios,
ISBN 978-953-51-0337-0, Published: March
16,2012.
[15] KartikBommepally,GlisaT.K.,Jeena
J. Prakash, Sanasam Ranbir Singh and Hema
A Murthy 2010 , Internet Activity Analysis
Through Proxy Log?, In Proceedings of the
National Conference on Communications,
Chennai ,January 29-31, pp. 1-5.
[16] Krishna Das and Smriti Kumar Sinha
2016, Social Networked Community
Analysis Using Graph Based Data Set,
International Journal of Innovations
&Advancement in Computer Science, 5(4),
13- 19, Academic SciencePublisher.
[17] Lszl Gyarmati and Tuan Anh Trinh,
Measuring User Behavior in Online Social
Networks, IEEE Network, 24( 5), 26 - 31 ,
IEEE.
[18] M. Namasivayam and N. Radhika 2014,
Diagnostic Group Socio BehaviorsUsing
Dim Extractions, International Journal of
Computer Science and Mobile Applications,
2(1), 109-114.
[19] M.Vasudevan & M.Tamilarasi
2012,Collective behavior prediction in social
media: A survey, International Journal of
Computer Networks and Wireless
Communications (IJCNWC), 2(6),676-679.
[20] Marcelo Maia, Jussara Almeida and
Virglio Almeida 2008, Identifying
userbehavior in online social networks , In
Proceedings of the 1st Workshop on Social
Network Systems, Glasgow, Scotland Uk,
April 01 - 04,Pages1-6.
[21] Maria Kihl1 , Christina Lagerstedt,
Andreas Aurelius and Per dling 2010, Traffic
analysis and characterization of Internet user
behavior, In Ultra Modern
Telecommunications and Control Systems
and Workshops (ICUMT), Moscow, 18-20
Oct., 224- 231,IEEE.
[22] Mark Granovetter 1978, Threshold
Models of Collective Behaviors, The
American Journal of Sociology, 83(6), pp.
1420-1443.
[23] Mattew Rowe, Miriam Fernandez and
Harith Alani 2013 , Modelling and Analysis
of User Behaviour in Online Communities,
IEEE Computer Society Special Technical
Community on Social Networking E-
Letter,1(2),http://stcsn.ieee.net/e-letter/vol-1-
no2.
[24] Michal Kosinski, Yoram Bachrach,
Pushmeet Kohli , David Stillwell and
ThoreGraepel 2014, Manifestations of user
personality in website choice andbehaviour
on online social networks, Machine
Learning, 95(3), pp. 357-380,ACM.
[25] Mohammad-Ali Abbasi, Sun-Ki Chai,
Huan Liu and Kiran Sagoo 2012, Real-
World behavior analysis through a social
media lens, In Proceedings of the 5th
international conference on Social
Computing, Behavioral-Cultural Modeling
and Prediction (SBP’12), Heidelberg, April 3,
18-26,Springer-Verlag.
[26] Monika Mital & Sumit Sarkar 2012,
Multihoming behavior of users in social
networking web sites: a theoretical model,
Information Technology & People, 24(4), pp.
378-392.
[27] Natalia Diaz Rodriguez, M.P Cuellar,
John Lilius and Miguel Delgado Calvo-
Flores 2014, A Survey on Ontologies for
Human Behavior Recognition, ACM
Computing Surveys, 46(4), 43-75,ACM.
[28] Natalia Daz Rodrguez, Manuel
Pegalajar Cullar, Johan Lilius, Miguel
Delgado Calvo-Flores 2014, A Fuzzy
Ontology for Semantic Modelling and
Recognition of Human Behaviour,
Knowledge-Based Systems, 66,46-60.
[29] Nhat Hai Phan, Dejing Dou, Hao Wang ,
David Kil and Brigitte Piniewski 2015,
Ontology-based Deep Learning for Human
Behavior Prediction in Health Social
Networks