a modified weight balanced algorithm for influential users community detection in online social network (os ns)

International Journal of Advanced Engineering, Management and Science (IJAEMS) [Vol-2, Issue-6, June- 2016]
Infogain Publication (Infogainpublication.com) ISSN : 2454-1311
www.ijaems.com Page | 595
A Modified Weight Balanced Algorithm for
Influential Users Community Detection in
Online social Network (OSNs)
Bhupinder Singh1
, Manju Bala2
, Manoj Kumar3
1
Department of Computer Science, CTIEMT, IKG PTU, Jalandhar, Punjab, India
2
Professor, CTIEMT, Jalandhar, Punjab, India
3
Principal, Daviet, Jalandhar, Punjab, India
Abstract—In the modern era online users are increasing
day by day. Different users are using various social
networks in different forms. The behavior and attitude of
the users of social networking sites varies U2U (User to
User). In online social networking users join many
groups and communities as per interests and according to
the groups’/Communities’ influential user. This paper
consist of 7 sections , first section emphasis on
introduction to the community evelotion and community.
Second section signify movement between communities
,third section involve related work about the research..
Fourth section includes Problem Definition and fifth
section involve Methodology (Proposed Algorithm
Process ,Get Community Matrix, Community
detetcion).Sixth section involve Implementation.
Furthermore implementation include Datasets
,Quantitative performance, Graphical Results,
Enhancement in the existing work..Last section include
Conclusion and then references. In this paper,we are
implementing and proposing the community detection in
social media .In the proposed we have deployed a
Longest Chain Subsequence metric for finding the
number of connections to the kernel community.
Keywords—SocialMedia, OSN, Community, U2U,
Evolution, Weight Balanced, greedy approach.
I. INTRODUCTION
In online social network community analysis and
detection is a vast area of research. Many researchers are
working to find the behavior and influence behavior of
communities in OSNs’. Classification, detection and
evolution of communities in a computational sense, seeks
to assign labels to data. Given a set of features for
community, a classifier attempts to assign a label to that
community. The classifier does this by drawing upon
knowledge derived from examples of how other
communities have been labeled. These examples, referred
to as training data, serve as a source of prior knowledge
that the classifier uses to make decisions about previously
unseen communities. Categorization is a specialization of
classification.
It deals with assigning a category to community. Other
classification algorithms may simply make a yes/no
decision based upon inputs, such as a fraud detector that
indicates whether a credit card transaction is fraudulent.
Categorization algorithms place an object in one of a
small set of categories, such as categorizing cars as
coupes, sedans, SUVs, or vans. Many of the concepts we
discuss in this chapter pertain to classification as a whole,
whereas others are more closely related to categorization.
You’ll see some terms used interchangeably. Community
categorization in the sense is to find the community and
to determine its desired level of surviving.
This new type of data mandates new computational data
analysis approaches that can combine social theories with
statistical and data mining methods[14]. The pressing
demand for new techniques ushers in and entails a new
Inter disciplinary field – social media mining. Thus we
have studied the various techniques for community
evolution in the online social media and we are discussing
existing methods in this paper.
II. MOVEMENT BETWEEN COMMUNITIES
Having analyzed the membership and growth of
communities, we have a tendency to currently
intercommunicate the question of however folks and
topics move between communities. A elementary
question here is that the degree to which individuals bring
topics with them from one community to a different,
versus the degree to that topics arise in a very community
and afterward attract folks from alternative communities.
In alternative words, given a collection of overlapping
communities, do topics tend to follow folks, or do folks
tend to follow topics? we have a tendency to conjointly
investigate a connected question: once folks get into a
community square measure they additional or less
doubtless than alternative members of the community to
be participants in current and future “hot topics” of
dialogue therein community?. While these queries square
measure intuitively terribly natural, it's a challenge to
outline sufficiently precise versions of them that we are
able to build quantitative observations.
III. RELATED WORK
A substantial quantity of hardwork has been dedicated to
the task of distinctive and evaluating closely knit

communities in giant social networks, most of that is
predicated on the premise that it's a matter of common
expertise that communities exist in these networks. Akrati
Saxena et al. [1] analysed the impact of the topology of a
social network, specifically its meso scale properties-
community structure and core-periphery structure, on a
acculturation traversing over it. we have a tendency to
propose a acculturation propagation model for artificial
scale free graphs .David Burth et al. [2] represented
however process researches have approached this
subject and therefore the strategies accustomed analyse
such systems. supported on a good tho' non-exaustive
review of the literature, a taxonomy is projected to
classify and describe totally different classes of analysis.
Chang Su et al. [3] projected a community detection rule
supported influential nodes. First, we tend to introduce a
way to notice cogent nodes supported stochastic process.
Then we tend to mix the rule with order statistics theory
to seek out community structure. we tend to apply our
rule in 3 classical information sets and compare to
alternative algorithms. Anna Zygmunt et al. [4]
conferred a proper model of social structure with evolving
groups. a brand new rule decisive a life cycle of social
groups and their cores is conferred. supported the
knowledge derived from them, one will - victimization
Social Network Analysis strategies - build a network of
connections between users, that then may be analysed to
seek out influential users and groups that square measure
shaped around them.
Weishu Hu et al. [5] outlined and study a completely
unique community detection drawback that's to find the
hidden community structure in giant social networks
supported their common interests. we tend to observe that
the users usually pay a lot of attention to those users who
share similar interests, that change how to partition the
users into totally different communities per their common
interests. Liaoruo Wang et al. [6] define and explore a
novel problem called community kernel detection in order
to uncover the hidden community structure in large social
networks. They discover that influential users pay closer
attention to those who are moresimilar to them.
Beiming Sun et al. [7] described that abundant endeavor
has been conducted to investigate info from social
networks, as well as finding the influential users. during
this paper, we tend to propose a graph model to represent
the relationships between on-line posts of 1 topic, so as to
spot the influential users . Michael Trusov et al. [8]
mentioned that the success of web social networking sites
depends on the amount and activity levels of their user
members. though users generally have varied connections
to alternative web site members (i.e., “friends”), solely a
fraction of these alleged friends may very well influence a
member’s web site usage. Emilio Ferrara [9] involved
with the analysis of aggregation patterns and social
dynamics occurring among users of the most important
OSN because the date: Facebook. In detail, we tend to
discuss the mesoscopic options of the community
structure of this network, considering the angle of the
communities, that has not nevertheless been studied on
such an oversized scale.
Suhani Grabowicz P et al. [10] mentioned that in trendy
days social network services are popularly utilized in
electronic commerce systems. so as to market the
business, it's fascinating to explore hidden relationships
among users developed on the social network. during this
paper, we tend to outline and study a completely unique
community detection downside that's to get the hidden
community structure in massive social networks
supported their common interests. David R Hunter et al.
[11] present a scientific examination of a true network
information set victimization most chance estimation for
exponential random graph models moreover as new
procedures to judge however well the models match the
determined networks.
Salvatore Catanese et al. [12] given access to information
regarding Facebook users and their friendly relationship
relations. For this they noninheritable the mandatory
info directly from the side of the web site, so as to
reconstruct a sub graph representing anonymous
interconnections among a big set of users. Author
delineate impromptu, privacy-compliant crawler for
Facebook information extraction. Pasquale American
state Meo et al. [13] explained to work out whether or not
2 distinct users may be thought of similar. during this the
authors planned AN approach so as to estimate the
similarity of 2 users supported the information of social
ties (i.e., common friends and teams of users) existing
among users, and therefore the analysis of activities (i.e.,
social events) during which users area unit concerned.
IV. PROBLEM DEFINITION
The basic structure of social area network in community
detection is categorized in two tasks. Initial step begins
with the identification and extraction of influential kernel
members. The second step is to identify the
corresponding structure of the community kernels. The
problem arises in detection of the exact structure among
the influential kernel members and auxiliary
communities. Most of the conventional algorithms ignore
the link among the auxiliary communities and community
kernels, resulting in failure to distinguish among them.
The overlapping of the communities drops the actual
structure of the community. Although, in order to
distinguish the kernel members, the Page rank and
conductance based algorithm are designed that utilizes the
link information. However, the link information among
the auxiliary and kernel members are ignored finding the
community kernels. Most importantly, the growing
vertices of the network must be addressed developing
robust algorithm with large scalability.
The problem of community detection has many smart
applications, yet as representative user finding, friend
recommendation, network visual image, and marketing.
However, this drawback is non-trivial and poses a group
of challenges. First, it's hard to identify the extremely
potent users.

International Journal of Advanced Engineering, Management and Science (IJAEMS)
Infogain Publication (Infogainpublication.com
www.ijaems.com
V. METHODOLGY
Proposed Algorithm Process
Fig 1: shows flowchart of Proposed Algorithm Process
Compute size of the data
Set initial communities with one node per community
Initialize best community to current value
Compute initial community matrix
Lines and columns sum (speed optimisation)
Compute Best variation
For all the pairs of nodes that could be merged
If they share
edges
If best variation, then keep track of
the pair
Compute the
variation
if i==1 && j==1
dQ = 2 * (communityMat(i,j) - ls(i)*cs(j));
Merge the pair of clusters maximising proximity
Update community matrix
Update lines/colums sum
Update value
If new value is better, save current partition
International Journal of Advanced Engineering, Management and Science (IJAEMS)
Infogainpublication.com)
ETHODOLGY
Proposed Algorithm Process
1: shows flowchart of Proposed Algorithm Process
Get Community
Fig 2: shows flowchart of Adjacent Matrix
Community Detection
Fig 3: shows flowchart of community detection
VI. IMPLEMENTATION
Datasets[15] :- Our experiments are conducted on two
facebook datasets. Facebook data was collected from
survey participants using third party facebook app such as
social circles. The dataset includes node features
(profiles), circles, and ego networks.
Quantitative performance.
and F1-score to evaluate existing techniqu
proposed Algorithm Technique.
Fig 4: Feature construction
We need to compute the size of data.
communities with one node per community
community to current value. Compute initial community
Set initial communities with one node per community
e best community to current value
Compute initial community matrix
Lines and columns sum (speed optimisation)
of nodes that could be merged
If best variation, then keep track of
proximity
If new value is better, save current partition
Number of communities
Initialise adjacency matrix list
Create edges and normalise values by dividing by the sum of
all edges.
Input Data set
Applying existing approach for community Detection
Applying proposed approach for community Detection
Extract Community
[Vol-2, Issue-6, June- 2016]
ISSN : 2454-1311
Page | 597
Community Matrix
2: shows flowchart of Adjacent Matrix
Community Detection
3: shows flowchart of community detection
IMPLEMENTATION
Our experiments are conducted on two
Facebook data was collected from
survey participants using third party facebook app such as
social circles. The dataset includes node features
(profiles), circles, and ego networks.
Quantitative performance. We use Precision, Recall,
score to evaluate existing techniques with our
proposed Algorithm Technique.
: Feature construction[6]
We need to compute the size of data. Set initial
communities with one node per community. Initialise best
community to current value. Compute initial community
Number of communities
Initialise adjacency matrix list
Create edges and normalise values by dividing by the sum of
Input Data set
Applying existing approach for community Detection
Applying proposed approach for community Detection
Extract Community

matrix. Lines and columns sum (speed optimisation) .
Initialise best known and current Q values.
curQVal=trace(communityMat)-
sum(sum(communityMat^2));
Loop until no more aggregation is possible.Compute Best
Q variation. For all the pairs of nodes that could be
merged. If they share edges,then Compute the variation in
Q. If best variation, then keep track of the pair. Then
Merge the pair of clusters maximising Q. Update
community matrix. Update lines/colums sum. Further
Update the value. If new Q is better, save current
partition.
Graphical Results:-
Fig. 5: Detection of Community 1 (Facebook Data) :
Existing
Fig. 5: Detection of Community 2 (Facebook Data) :
Existing
Fig. 6: Detection of Community (Facebook Data) :
Proposed
Enhancements in Existing Work
The existing method works on two algorithms Greedy
Algorithm and Weight-Balanced Algorithm (WEBA). The
greedy algorithm works on finding the maximum distance
from a given set of vertices to other connections. The
process is repeated for every vertex from the vertex set in
the graph. The vertex having maximum number of degree
is selected as community head. However, the greedy
algorithm does not consider the link among auxiliary
members and kernel members. In the WEBA algorithm, a
weight measure is associated with each kernel community
followed by optimization for finding the exact kernel
structure and auxiliary kernels. However, in such
algorithm, the small communities’ having less number of
connections has a lower priority value and never gets
identified.
In the proposed we have deployed a Longest Chain
Subsequence metric for finding the number of
connections to the kernel community. The method
performs grouping the communities based on LCSS
metric. Despite of considering only the maximum value,
the algorithm considers the new connections as well as
the overlapped connections for a particular community.
This results in better identification of the kernel
communities as well as the auxiliary communities intact.
The proposed method performs the distance measure with
the Longest Chain Sub Sequences (LCSS) metric defined
as.
( ) ( )
( ) ( ) ji
ji
jiji
jijiij
yifx
yifx
andjifi
YXLCSYXLCS
YXLCSYXLCSd
≠
=
==





==
−−
−−
00
),,,max(
,,
11
11
φ
Fig. 7: show Longest Chain Sub Sequences (LCSS) metric
VI. CONCLUSION
Now it s clear that communities in OSN are playing a
vital role and the users of communities are changing their
role as per community behavior. If the community is
demising then its directly meaning that its users are
leaving that particular community and if the users are
joining the community then it is clear that particular
community will grow in future. Various methods are user
to find the community structure and the analytical
behavior. In future work can be extended to find the
spatial information of the user as well as to the
community in social network. For future work, we would
like to explore the dynamic behavior of community
kernels and their auxiliary communities.We are interested
in how community kernels take shape and evolve over
time. In addition, we would like to combine link and
content information in our problem definition and
algorithm design for practical applications, such as query
dependent community detection. Inspite of assuming the

utmost price, the algorithmic program considers the new
connections yet because the overlapped connections for a
selected community. This ends up in higher identification
of the kernel communities yet because the auxiliary
communities intact.
REFERENCES
[1] Akrati Saxena,S.R.S.Iyengar,YayatiGupta,
“Understanding Spreading Patterns on Social
Networks Based on Network Topology”,
arXiv:1505.00457v1 [cs.SI] , May 2015
[2] David Burth Kurka, Alan Godoy, Fernando J. Von
Zuben, “Online Social Network Analysis: A Survey
of Research Applications in Computer Science”,
arXiv:1504.05655v1 [cs.SI] , Apr 2015
[3] Chang Su, Yukun Wang , Lan Zhang “Community
Detection in Social Networks Based on Influential
Nodes ”, Journal of Software, VOL. 9, NO. 9,
September 2014
[4] Anna Zygmunt, Jarosław Ko´zlak, Edward
Nawarecki, Adam Mika, “Determining life cycles of
evolving groups”, AGH University of Science and
Technology, Al. Mickiewicza VOL.35, PP.1102-
1111 Krak´ow, Poland August ,2014
[5] Weishu Hu, Zhiguo Gong, Leong Hou U and Jingzhi
Guo, “Identifying influential user communities on
the social network ”,Department of Computer and
Information Science, Faculty of Science and
Technology, University of Macau, Macao, China,
2013 Taylor & Francis.
[6] Liaoruo Wang, Tiancheng Lou, Jie Tang and John
E. Hopcroft, “Detecting Community Kernels in
Large Social Networks”, ICDM '11 Proceedings of
the 2011 IEEE 11th International Conference on
Data Mining PP. 784-793 ISBN: 978-0-7695-4408-3
[7] Beiming Sun, Vincent TY Ng, “Identifying
Influential Users by Their Postings in Social
Networks”, MSM’12, June 25, 2012, Milwaukee,
Wisconsin, USA. Copyright 2012 ACM 978-1-
4503-1402-2/12/06
[8] Michael Trusov, Anand V.Bodapati, and Randolph
E.Bulkin, “Determining Influential Users in Internet
Social Networks”, Journal of Marketing Research
VOL. XLVII (August 2010), 643–658
[9] Emilio Ferrara, “A large-scale community structure
analysis in Facebook”,Ferrara EPJ Data Science
2012, 1:9
[10]Suhani Grabowicz P, Ramasco J, José J.
Ramasco ,Esteban Moro,Josep M. Pujol, Victor M.
Eguiluz, “Social features of online networks: the
strength of intermediary ties in online social media”,
PLoS ONE 7(1):
e29358.doi:10.1371/journal.pone.0029358,2012
[11]Hunter D, Goodreau S, Handcock M, “Goodness of
fit of social network models”, J Am Stat Assoc
103(481):248-258, 2008.
[12]Salvatore A. Catanese, Pasquale De Meo , Emilio
Ferrara, Giacomo Fiumara, Alessandro
Provetti,“Crawling Facebook for Social Network
Analysis Purposes”, WIMS’11, May 25-27, 2011
Sogndal, Norway
[13]Pasquale De Meo a, Emilio Ferrara b,a, Giacomo
Fiumara a, Alessandro Provetti a,c, “Mixing local
and global information for community detection in
large networks,”, Journal of Computer and System
Sciences VOL. 80 (1), PP.72-87
[14]Manju Bala, Manoj Kumar, Bhupinder Singh,
“Socia Media Evolution and Detection of influential
Community Users in online Social Networks
(OSNs)”, ICACSET 2016 - Singapore on April 14 -
15, 2016
[15]https://snap.stanford.edu/data/

a modified weight balanced algorithm for influential users community detection in online social network (os ns)

Recommended

More Related Content

What's hot (20)

Similar to a modified weight balanced algorithm for influential users community detection in online social network (os ns) (20)

Recently uploaded (20)

a modified weight balanced algorithm for influential users community detection in online social network (os ns)