=Paper= {{Paper |id=Vol-1988/LPKM2017_paper_4 |storemode=property |title=Sentiment Classification Method for Identification of Influential Learners in Social Networks Communities |pdfUrl=https://ceur-ws.org/Vol-1988/LPKM2017_paper_4.pdf |volume=Vol-1988 |authors=Radhia Toujani,Jalel Akaichi |dblpUrl=https://dblp.org/rec/conf/lpkm/ToujeniA17 }} ==Sentiment Classification Method for Identification of Influential Learners in Social Networks Communities== https://ceur-ws.org/Vol-1988/LPKM2017_paper_4.pdf

Sentiment classication Method for identication of
inuential learners in Social Networks Communities
Radhia Toujani1 and Jalel Akaichi2
1
BESTMOD Department , Higher Institute of management, University of Tunis, Tunisia
toujaniradia@gmail.com
2
Department of Information Systems, College of Computer Science, King Khaled
University Abha, KSA
jalel.akaichi@ku.edu.sa

Abstract. The growth of social networking has gained much interest from
the research community in recent years. Social networking technology as an
e-learning tool seems promising for education instructors to combine distance
education. Several analysis researches of social media were conducted for
detection opinion leaders. While most of the existing algorithms proposed
for communities determination are destined to commercial use, in this work,
we present a new approach for detecting opinion leaders based on analyzing
online learning community interactions. In fact, we aim to identify learners
behaviors and attitudes in social network sites as productive online tools
for learning. To achieve this purpose, we describe a method of performing
detecting opinion leaders by using machine learning techniques. We focus
on the application of text mining and sentiment analysis. The output of
this work prove that education-based social network is very eective and
improvement for online communications. Experiments show the eciency of
the introduced method which can be helpful and protable for education
instructors.

Keywords: Social network, learning communities, sentiment classication
,opinion mining, opinion leader, learners behavior, inuential learners, ma-
chine learning.

1 Introduction
Researchers have become interested to analyse social network site(SNS) and com-
munities detection to explore the educational benets of using it. Using social net-
work, we construct a virtual communities enable learners to connect and collaborate
on international platforms, exceeding geographical boundaries [1], eliminate these
boundaries increases communication, collaboration, and engagement [1].
In the shared forum provided by an SNS, learners can construct a shared under-
standing and engage and collaborate in discussions, while sharing common resources,
such as readings, links, and videos [1]. In a distance education course, SNSs provide
learners with a venue for fostering and developing a community of practice through
technological aordable, such as user proles, forums, tools, and resources. Given
the inherent limitations of course management systems, the use of social network in
education settings represents a denitive shift toward social and community-based
web applications that cultivate and sustain discipline-specic social networks. For
today's technologically savvy students, social network represent improved technolo-
gies for creating a heightened sense of community resulting in the acquisition of new
student knowledge and collective intelligence [2].
The integration and use of social networking technology as an e-learning tool
seems promising for distance education. In [3], the author indicates that social soft-
ware may be the 'killer app' for distance education, given its ability to enhance
2 Radhia Toujani and Jalel Akaichi

social presence. The creation of social networking around academic topics provides
instructors and students great exibility for teaching and learning[2].
In this paper, we propose a method to detect inuential learners in the social net-
works. At the rst step, the extraction and the collection of the raw data are ex-
plained. Then, the process of sentiment analysis of online learning communities and
the detection of opinion leader are discussed.

The rest of this paper is organized as follows. Section 2 discusses the related
works; Section 3 presents the research framework; Section 4 presents experimental
results; in Section 5, concludes the paper with some suggestion for future research
directions.

2 Related works
Social networks analysis has recently gained more importance thanks to its appli-
cation in elds such as sociology. Many researchers are interested on identication
of opinion leaders. So far, some researchers were studied on opinion leaders where
rst of them is situated within the eld of sociology. Numerous studies such as
[5][10][16] have been conducted in order to understand the concept of opinion and
characteristics of leaders distinguishing them of their followers.
The studies on opinion leader identication can be divided into two parts. First,
the link-based opinion leader identication methods that consider the social inter-
active structure of the network. Second, the mixture of opinion leader identication
methods that combine the social link information with semantic-based information
embodied in documents [9][12][15].
As an example of link-based opinion leader selection, the authors in[4] have iden-
tied opinion leaders with an excellent edge rank algorithm based on excellent net-
work theory, applied it to rank excellent edges and used the ranking result to iden-
tify opinion leaders in opinion excellent network model. Alternatively, researchers in
[11][13][17] have proposed an improved combined framework identifying the opinion
leader in online learning communities, which ranked opinion leaders based on four
distinguishing features: expertise, novelty, inuence, and activity. Furthermore, the
performances of opinion leaders were further investigated in terms of longevity and
centrality.
Also, in[8] [6], authors have found that those opinion leaders are the optimum
marketing choice in terms of diusion speed and maximum cumulative number of
adopters, used a social network method and threshold model and conclude that
opinion leaders with a high sociality and centrality (users who have high degree of
connections to other people) are the best ones for fast diusion.
In case of social network, many of the trust of customers to the social network
communities are based on opinion leaders' recommendations for products and ser-
vices [7][18]. So, how to identify the opinion leaders eectively is the key to raise
sales and brand awareness. In sales and marketing eld, some studies focus on de-
veloping various indexes such as in-degree, out-degree, betweenness and closeness in
social network analysis to identify opinion leaders. These studies are interested in
identifying opinion leaders who will forward be marketing messages to other users
via their trust and distrust networks.

3 METHODOLOGY
This section presents our methodology. Firstly, the general architecture is presented.
Secondly, the extraction and the collection of the raw data are explained. Then, the
process of sentiment analysis of online learning communities and the detection of
opinion leader are discussed.
Title Suppressed Due to Excessive Length 3

Fig. 1. System Targeted Architecture

3.1 Raw data collection
With the widespread adoption of social media, online learning communities are
perceived as a network of knowledge comprised of interconnected individuals with
varying roles. Opinion leaders are important in social networks because of their
ability to inuence the attitudes and behaviours of others. The introduced work is
based on the analysis of learning communities in social network namely Facebook.
In this study, we chose the type group 'We Read for you'( On a lu pour vous
(OLPV))3 . We used the programming interface available on Facebook site, called
Facebook API, providing tools and applications that allowed us to access and get the
necessary for information our study. Unlike closed and secret groups, the extraction
of database from learning community 'OLPV' group is permitted. This later group
has 13,537 members and allows its members to discuss and exchange views on issues
related to learning especially reading. Posts the online learning community OLPV
are grouped by interest and has many forms namely "status", "photo", "video",
"URL" and "event". These posts have multiple messages types:
An interrogative message where the user requests the opinions of others,
Quote, poem or proverb message,
a message where the user expresses his opinion or a message without an opinion.
In fact, these types of posts are shown in the gure 2.

3.2 Machine Learning and Opinions extraction
We introduce machine learning method namely Fuzzy SVM for mining opinions of
social network learners. Fuzzy SVM is explored for multi-class scenarios (positive,
negative, fuzzy). In fact, we propose a renement which aims to consider the spring's
variables which refer to the membership functions of fuzzy classes. This later is
dened as follows:
i = γw,b − (Yi fw,b (xi ))+ (1)
where (x)+ represents the positive part of x ∈ R. This quantity is related to the
concept of margin.It measures the location of a pair (Xi , Yi ) in relation to the mar-
gin of the classier fwb . Therefore, three cases are introduced:

3
https ://www.facebook.com/groups/oalpv.tn/members/
4 Radhia Toujani and Jalel Akaichi

Fig. 2. (a) interrogative message , (b) Quote Message , (c)Message expressing opinion ,
(d) Message without opinion

if i > γw,b so the pair (Xi , Yi ) is misclassied by fw,b
if 0< i < γw,b so so the couple is well ranked but close to the hyperplane (the
distance from the hyperplane is less than the margin)
i = 0 so the couple is in its majority class (the distance to the hyperplane is
greater than the margin).
Indeed, the introduction of these springs variables allow to relax the mathematical
formulation of our fuzzy SVM. While the main classes treated on sentiment analysis
in social networks are the positive and negative opinion, in our research we focus
on fuzzy aspects. To solve this problem, we integrate fuzzy class to labeled input
terms. In this context, the training samples are :

S = {(Xi , yi , si ), i = 1, ...N ∈ RN } (2)
Where each Xi ∈ R denotes a training sample and yi ∈ {-1,1} represents its class
N

label. si = {1,2,...N} present our fuzzy class with σ ≤ si ≤ 1 and a constant σ ≥
0.Denote a set as:

Q = {Xi |(Xi , yi , si ) ∈ S} (3)
Clearly, it contains two classes. One class contains positive opinion such sample
point Xi with yi =1, denoting this class by C + , then,

C + = {(Xi |(Xi ∈ S)}andyi = 1 (4)
The negative class contains such sample point Xi with yi =-1, denoting this class
by C − , then,

C − = {(Xi |(Xi ∈ S)}andyi = −1 (5)
Indeed,

Q = C+ ∪ C− (6)
Thus, the mathematical formulation for our sentiment classication problem and
the linear Fuzzy Support Vector Machine classier can be obtained by solving the
following optimization problem:
n
(7)
X
min 1/2kW k2 + C si i
i=1

yi (wT Φ(Xi + b)) ≥ 1 − i (8)
i ≥ 0, i = {1, ...N } (9)
Title Suppressed Due to Excessive Length 5

where C denotes a constant, the fuzzy class si is the fuzzy labeled opinion terms
Xi toward one class and the parameter i the percentage of opinion class , the term
si i can be looked as a measure of degree of fuzziness.
Once opinions are detected, a score opinion ScoreOp is assigned . This score is
weighted or equal to the number of words contained in the document.

3.3 Detection of inuential learners
The detection of online inuential learners, we opted for an assessment score cor-
responding to the frequency of opinion terms contained in each learner message. In
addition, to express score of inuential learners, we opted a threshold below which
the inuence of the post is not signicant. This threshold is equal to the average of
the ScoreOp obtained for all posts.
In other words, any post having a "ScoreOp" higher than or equal to the dened
threshold, will be considered as inuential leaders. Figure 3 depicts extracted posts
from online learning community and presented opinions leaders learners.

Fig. 3. Example of inuential learners post

4 SIMULATION RESULTS
Here we present the experimental results of the introduced opinion leaders method
on both learning communities OLP V 1 and OLP V 2. We use Graph API 4 to extract
real time data from these two later communities. In fact, the use of data from the
same learner group (OLPV) in two dierent periods allows us to show the position
of opinion leaders over time.
The inuential learners belonging to community OLPV1 are shown, in ascending
order, in Figure 4
4
https ://developers.facebook.com/docs/graph-api
6 Radhia Toujani and Jalel Akaichi

Fig. 4. Diagram of inuential learners in OLPV1

Using graph API the ID of the inuential learners, we can extract information
about the opinion leader's as name and published message. The following gure
shows the opinion leader having the highest ScorInf l from the OLPV1 learning
community.

Fig. 5. Illustration of name and message of the opinion leader in OLPV1

Figure 6 shows the diagram of inuential learner sorted by ScoreInf l of the
second learning community OLPV2.
Title Suppressed Due to Excessive Length 7

Fig. 6. Diagram of inuential learners in OLPV2

Figure 8 depicts information related to inuential learner in OLPV2 community.

Fig. 7. Illustration of name and message of the opinion leader in OLPV2

Moreover, we use precision and recall to validate the utility of the introduced
method. The precision measures the exactness of a classier and recall represents
the number of correct classications penalized by the number of missed items. These
two measures are calculated as follows:

I ∩ NI
P recision = (10)
NI

NI ∩ I
Recall = (11)
I

where I denotes the number of outputs inuential learners, N I represents the num-
ber of outputs non-inuential learners. Figure 8 depicts information related to in-
uential learner in OLPV2 community.
8 Radhia Toujani and Jalel Akaichi

Fig. 8. Illustration of name and message of the opinion leader in OLPV2

According to the curve above, which present the recall and the precision our
method gets a higher precision but a lower recall. We can conclude that the perfor-
mance of the proposed methodology is good.

5 Conclusion
In this work, we have investigated the utility of opinion mining on a novel collection
of dataset which is online learning communities. Using the most well-known machine
learning algorithm namely SVM, we identify inuential learners. First of all, we
analyse popularity and reputation of posts to determine which are more important
than others referring to the frequency of opinion words in learning posts. Then, we
identify inuential posts leading to detect opinion leaders learners.
The overall performance of the proposed methodology is satisfactory, however, we
would like to further improve our research needs to explore how SNSs can be used
most eectively in distance education courses as a technological tool for improved
online communications among students in higher distance education courses.

References
1. K. H. McCann., Virtual communities for educators: An overview of supports and best
practices., Proceedings from Technology, Colleges, and Community Conference, pages
137142,2009.
2. I. E. Allen and J. Seaman., Online nation: Five years of growth in online learning.,
MA: Sloan Consortium 2007.
3. T Anderson., Distance learningsocial softwares killer ap.,Adelaide, South Australia:
University of South Australia,2005.
4. B. Mullen.et al. ,Shared leadership in teams: an investigation of antecedent conditions
and performance.Acad. Manag. J, 50:12171234,2007.
5. J.B. Carson and al., Salience, motivation, and artifact as contributions to the relation
between participation rate and leadership,J. Exp. Soc. Psychol, 25:545,559,1989.
6. Y. Cho and al. Identication of eective opinion leaders in the diusion of technological
innovation: a social network approach.,Technol. Forecast. Soc, 79:79-106,2012.
7. S. Misra K.K. Chan. Centrality in social networks conceptual clarication.,TSoc. Netw,
3:215-239,1979.
8. L.C. Freeman. Characteristics of the opinion leader: a new dimension.,J. Advert,19:53-
60,2012.
9. J.M. Kleinberg. Authoritative sources in a hyperlinked environment.,J. ACM (JACM),
46:604-632,1999.
10. Y. Li and al.An improved mix framework for opinion leader identication in online
learning communities.KnowledgeBased Syst, 43:34-51,2013.
Title Suppressed Due to Excessive Length 9

11. K. H. McCann. Virtual communities for educators: An overview of supports and best
practices.Proceedings from Technology, Colleges, and Community Conference, pages
137-142,2009.
12. S. Neti. Social media and its role in marketing.Int. J. Enterp. Comput. Bus. Syst,
14:1-15,2011.
13. E.C. Nisbet.The engagement model of opinion leadership: testing validity within a
european context Int. J. Public Opin. Res, 18:3-30, 2006.
14. L. Page and al.The pagerank citation ranking: Bringing order to the web.Int. J. Public
Opin. Res, 1999.
15. B. Batinic T. Gnambs. A personality-competence model of opinion leadership.Psychol.
Mark, 29:606-621, 2012.
16. M.P. Venkatraman. Opinion leaders, adopters, and communicative adopters: a role
analysis.Psychol. Mark, 6:51-68, 1989.
17. G. Weimann and al.Looking for opinion leaders: traditional vs. modern measures in
traditional societies.Int. J. Public Opin. Res, 19:137-190, 2007.
18. Li.et al. Y.An improved mix framework for opinion leader identication in online learn-
ing communities KnowledgeBased Syst, 43:43-51, 2007.