Classification of Student’s E-Learning Experiences’ in Social Media via Text Mining

IOSR Journal of Computer Engineering (IOSR-JCE)
e-ISSN: 2278-0661,p-ISSN: 2278-8727, Volume 17, Issue 3, Ver. VI (May – Jun. 2015), PP 81-89
www.iosrjournals.org
DOI: 10.9790/0661-17368189 www.iosrjournals.org 81 | Page
Classification of Student’s E-Learning Experiences’ in Social
Media via Text Mining
Ms. Priyanka Patel1
, Ms. Khushali Mistry2
1
(Department of CSE, PIET, Vadodara, India, 2
(Department of CSE, PIET, Vadodara, India,
Abstract : In today’s world, social media is used every individual for expressing their feelings, opinion,
experiences’ and emotions. Applying data mining on all these emotions expressed in posts, comments and likes
called as social media data. In Existing system, only the prominent themes are identified with relatively large
number of tweets in the data. There are a variety of other issues hidden in the “others” theme. Several of these
issues may be of great interest to education researchers and practitioners. The fact that the most relevant data
which are found on engineering students’ learning experiences involve complaints, issues, and problems does
not mean there are no positive sides in students’ learning experiences’. This may entail that social media serve
as a good venue for students to utter negative emotions and seek social support. In proposed work, New Label
Good Things introduce and using the probability, the common keywords are considered for this label. Doing
this, the proposed system is identifying and classifying the e-learning problems faced by student to improve their
education quality with respect to their good and positive comments. Naive Bayes multilabel classifier is used for
classification of experiences' by finding the probability of words in tweet for each category probability of each
label contains how many users. Finally the tweets with new Label will be compared with the rest of the tweets
with existing Labels.
Keywords – Classification, social media, multilabel, learning experiences’, text mining .
I. Introduction
In today’s world we mostly use websites for Social Networking, Education, Marketing, Entertainment,
Business, Shopping, and so many other things for making life easy. Now days, Social media craze is mounting
to heights of success for every individuals. Youngsters are the most common users and most among of them are
Students. Students just comment, share, like and post their feelings on social media like Twitter, Facebook and
Youtube. Students feel free to discuss and share their experiences on social media in informal and casual
manner without considering spellings and accurate grammar. Social media data are unstructured. These social
media provides so many useful knowledge and information about the students’ emotions, feelings, experiences
and struggles in their studies outside the classroom. So following these Students track on the social media and
it’s an interesting outlook for educational researchers and practitioners to understand student learning
experiences outside the classroom. This understanding will uncover so many unknown experiences which was
not cleared or considered while the classroom discussion. This understanding about their experiences provides
so useful data for the betterment of student in decision making, enhancing students’ education quality, training
and placement, withholding and achievements.
The amount of social media data provides chances to understand students’ experiences, but their
methodological difficulties to use social media data for educational purpose. In classroom studies, to know each
and every students’ point of view surveys, reviews, group discussions, interview and concealing was done. The
innovative concept of using social media data focuses on the required information and knowledge to be
extracted for educational purpose by understanding students’ experiences.
The social media data like the students’ comments, posts and feelings have to be used for
understanding students’ learning experience with the research goals:
1) To find and classify students problem in their learning.
2) Track students’ good or bad experiences. Mining the social media data like students emotions will result to
classify the group of students’ according to their experiences and identify their problems to be solved to
improve the education quality.
Several mining techniques for text document are used for mine various textual matches through the
social area. Certain Social site applications with social websites like Facebook, Twitter, and Youtube gives so
many chances to ascertain communication via people leading to communal learning and circulation of important
knowledge such as comments, posts, emotions, along with likes.

Classification of Student’s E-Learning Experiences’ in Social Media via Text Mining
Mining text oriented files is the acquaintance breakthrough technique that provides computational
brainpower and this technique comprises of multidisciplinary fields, such as information rescue, text
investigation, natural language dispensation, and in sequence classification based on logical and non-petty
matches from colossal data sets. Users do not care about spelling mistakes and grammar rules. For text mining
classification and clustering which has been extensively used for the examination of the shapeless text available
on the large scale systems like Social area there are two approaches.
1.1 Problem Definition
Educational researches have been using old ways such as surveys, interviews, focus groups, classroom
activities to collect data related to students’ learning experiences. That will reducing scalability problem. As
optimistic about their experiences, students need to reflect on what they were thinking and doing sometime in
the past, which may have become obscured over time. There is no research found to directly mine and analyze
student-posted content with considering the students’ problem from uninhibited spaces on the social web with
the clear goal of understanding students’ learning experiences. The existing work has not measured student
academic performance to identify the students’ problem and classify them accurately for enhancing E-learning
experiences.
1.2 Motivation
Many times, students’ gets shy or afraid of clearing their problem during the classroom and this social
media help them to just post whatever they feel at that time about their emotions. The schools and departments
have been struggling with student recruitment and withholding issues. Graduates play an important role in
nation’s future workforce and which directly effects the nation’s economic growth and global competency. The
concept of combining Students’ learning experiences for enhancing E-learning experiences is innovative for
improving the training style or teaching style by which student to correct them at some required time without
continuous concealing or surveying. Based on understanding of issues and problems in students’ life,
policymakers and educators can make more informed decisions on proper interventions and services that can
help students overcome barriers in learning[1].Student can be trained or improve education quality as they have
been classified. Students’ learning experiences from social media will save the time to collect the data manually.
1.3 Objective
 To separate out meaningful information from the label “other”.
 To classify student based on content shared in the social media.
 To understand issues and problems students encounter in their learning experiences.
 To classify new label “Good things” for students’ combining with their emotional aspects and performance
grade.
 Improved the “Probability of the labels and keywords” for new introduced labels.
 It can be used to inform educational administrators, practitioners and other relevant decision makers to gain
further understanding of students’ college experiences.
II. Literature Survey
2.1 Social Media
With the intensification of the Internet communiqué techniques, the World Wide Web have became a
extremely imperative display place for users for a interaction with each other. In the course of this display place,
users could share with a ease and broaden in sequence and dreams to any one and from anywhere.
Next to with the people’s contacts with each other via these all social networking websites,
progressively data and un-like in sequence information have been aggregated. Surrounded by all different
information on these social networking websites, a quantity of has been considered as negative ones because
they could be utilize to attack and to malign. Twitter is an exceptionally trendy micro blogging site, where users
rummage around for appropriate and social information referred as breaking news, posts about celebrities, and
trending topics. Users send short transcript messages referred as tweets, have limitation of 140 characters by
length and could be view by user’s followers. A follower is any person called when he is choosing to have
other’s tweets posted on one’s timeline. For real-time, Twitters have been worn as a channel information giving
out and it has been used in various brand elections, campaigns, and likely to a news media.
2.2Naive Bayes Classification
The Naive Bayes Classifier technique has been based using Bayesian theorem and the best well-
matched to the dimensionality of the inputs has sky-scraping. Over and over again outperforms more
sophisticated classification methods, yet it is simple to operate. For the models, Maximum Likelihood estimates

all the parameters. To estimate the parameters, it has been require of small number of training. It operates
glowing and powerfully in supervised learning.
In Bayesian analysis,
 Prior Probability: It is a belief and based on previous experience. It is a ratio of number of single objects
and number of total objects.
 Likelihood: To classify a new object that this object belongs to which case.
 Posterior Probability: The final classification is made by combining both sources of information i.e. Prior
and Likelihood to form a Posterior Probability using Bayes rule.
 Posterior Probability of X being a object α Prior Probability of total objects xLikelihood of X given objects.
Above theorem guide us to solve the conditional probability of contrary and independent events. The
estimation of the probability for an opinion could be containing positive, negative or neutral. It would produce
good results .Conclusion is that Naive Bayes perform glowing in sky-scraping dependant features and
outperforms regularly compared to Neural Networks, Decision Trees etc.,. Nevertheless criterion maximum
likelihood parameter learning for Naïve Bayes classifier tends to be suboptimal.
2.3Text Mining
The data mining done on social media data covers many uncover features of the social media or the
social web i.e. Twitter, Facebook and Youtube. Mining file containing text has been known as Intelligent Text
breakdown or acquaintance discovery in Text or it can said that Text Data Mining which can be used to mine
the social media data. Mostly the social media data are unstructured format and to retrieve information from that
is complex due the massive information. So, it requires specific processing methods and algorithms to extract
useful information from that social web data.
2.4 Comparisons between Multiclass and Multilabel Classification
Table 2.1 Comparison between Multiclass and Multilabel classification
MULTICLASS CLASSIFICATION MULTILABEL CLASSIFICATION
PURPOSE This means a classification task with more
than two classes but not at the same time.
This means a classification task with more than two
classes at same time.
PROPERTIES It makes the assumption that each sample is
assigned to one and only one label.
It is a predicting property of a data point.
NATURE They are mutually exclusive. They are not mutually exclusive.
EXAMPLE A fruit can be either an apple or a mango, but
not can be apple or mango at the same time.
A text can be at any document at same time.
NOTATION Examples : D={x1,….xn}
Labels : L={ L1,…, Lm}
Each example is associated with one label (x,
l L)
Examples : D={x1,….xn}
Labels : L={ L1,…, Lm}
Each example is associated with a subset of labels (x,
S L)
2.5 Study of Papers
SR
NO
TITLE AUTHOR DESCRIPTION PUBLI-CATION
&YEAR
1 Understanding
Customers Using
Facebook Pages: Data
Mining Users
Feedback Using Text
Analysis
Hsin-Ying
Wu, Kuan-
Liang Liu
and Charles
Trappey
For start-ups companies, Following steps are
done in this paper:
 Key phrases extraction and summarization.
 Clustering algorithm: K-means.
 Clustering quality evaluation: cohesion and
separation.
IEEE,
2014
2 Mining Social Media
Data for
Understanding
Students’ Learning
Experiences
Xin Chen,
Mihaela
Vorvoreanu,
and Krishna
Madhavan
 Developed a workflow to integrate both
qualitative analysis and large-scale data
mining techniques.
 Used a multi-label classification algorithm
to classify tweets reflecting students’
problems.
IEEE,
2013

3 Content Matters: A
study of hate groups
detection based on
social networks
analysis and web
mining
I-Hsien Ting,
Shyue-Liang
Wang, Hsing-
Miao Chi and
Jyun-Sing
Wu
Two groups hate and like group.
 Facebook Data Extraction
 Data Pre-processing
 Social Network Structure Features
Extraction
 Content Features Extraction by TF-IDF.
 Group vectors
IEEE,
2013
4 Recognizing User
Identity in Twitter
Social Networks via
Text Mining
Sara Keretna,
Ahmad
Hossny and
DougCreight
on
They followed the steps:
 Input: unlabeled tweets.
 Feature extractor: feature vector is extracted
 Classifier: to predict identity of user.
IEEE,
2013
5 Twitter Trending
Topic Classification
Kathy Lee,
Diana
Palsetia,
Ramanathan
Narayanan,
Md. Mostofa
Ali Patwary,
Ankit
Agrawal, and
Alok
Choudhary
 They classify Twitter Trending Topics into
18general categories such as sports, politics,
technology ,etc.
 They experiment with 2approaches for topic
classification;
(i) the well-known Bag-of-Words approach
for text classification
(ii) network-based classification.
IEEE,
2011
III. Proposed Work
3.1 Overview of Proposed Work
Proposed work focus on the students’ emotions express on the social site by which they can be classify
into good and bad experiences. The proposed work is only the first step towards enlightening actionable insights
from student-generated content on social media in order to pick up education quality. The proposed system aims
at which are related to engineering students’ Twitter posts that are to understood arguments and problems for
their educational experiences. Initially to be conducted was qualitative investigation on such a gathered sample
taken from tweets which are related to engineering students’ belongs to college life. The proposed system finds
engineering students bump into problems are as follows: lack of social engagement, heavystudy load, and sleep
dispossession. Considering the outcomes to be brought the proposed system will implement of multi labelled
classifiers algorithm to identify and organize tweets which will be reflecting students’ problems.
The proposed system has been used for the supervised learning technique for the organization of classified the
twitter tweets and such a Foremost it has to be employed a healthy known text arrangement technique called
Naive Bayes (N.B.) [10]. A document for NB would be modelled for the presence and absence of particular
words. A variation of NB is Naive Bayes multinomial (NBM) and this will considers the rate of recurrence of
words and could be expressed as follows:
In which it represents as : P (c | d): the probability of a document as d individual in class as c, P ( c ):
the prior probability of a document taking place in class as c, and P (t k | c ): the provisional probability of term
as tk happening in a document of class as c. A document as d in such a case has been trending defined or tweets
associated with tags such as: #hashtag.
The proposed system has been sequences of steps which will recognize the social media data for
educational purposes which would overcomes all major limitations of both manual qualitative investigation and
large scale computational investigation of user-generated textual at ease There can be no pre-defined categories
of the data, subsequently the proposed system needed to explore what all the students are saying in the tweets
and what they want to say.

Fig.3.1 Architecture Diagram
As a consequence proposed system first conducted an inductive content analysis on the engineering
Problems dataset. Inductive content analysis is one popular qualitative research method for manually analyzing
text content. The proposed system extends the understanding of students’ experiences to the social and
emotional aspects based on their informal online conversations. These are important components of the learning
experiences that are much less emphasized and understood compared with academic performance.Existing
system have five prominent themes as under by feature extraction:
Table 3.1 list of category with their keywords[1]
CATEGORY MOST PROBABLE KEYWORDS THAT OCCURS IN TWEETS
Heavy Study Load
hour, homework, exam, day, class, work, negtoken, problem, study, week, toomuch, all, lab, still, out, time,
page, library, spend, today, long, school, due, engineer, already, disgusting
Lack of Social
Engagement
negtoken, Friday, homework, out, study, work, weekend, life, class, engineer, exam ,drink, break, Saturday,
people, social, lab, spend, tonight, watch, game, miss, party, sunny, beautiful
Negative Emotion hate, f***, shit, exam, negtoken, week, class, hell, engineer, suck, study, hour, homework, time, equate, FML,
lab, sad, bad, day, feel, tire, damn, death, hard
Sleep Problems sleep, hour, night, negtoken, bed, allnight, exam, homework, nap, coffee, time, study,more, work, class,
dream, ladyengineer, late,week, day, long, morning, wake, awake, nosleep
Diversity Issues girl, class, only, negtoken, guy, engineer, Asia,professor, speak, English, female, hot, kid, more, toomuch,
walk, people, teach, understand, chick, China, foreign, out, white, black
In Proposed system, we used six existing prominent themes or labels with new theme or label: “GOODTHING”.
The most probable features extracted under new label from analysing the tweets are as follows:

Table 3.2 Details of new Label with keywords
CATEGORY MOST PROBABLE KEYWORDS THAT OCCURS IN TWEETS
Others
All rest words are considered to be in this category
NOTE: Now separating out some useful words which will be coming under new label “GOODTHING” by analysing the extracted
tweets and then separated out the keywords for this new label.
GoodThing Advantage, proud, best, free, Congrats, creativity, skill, good, challenge, free, happy, early, like, love, relax,
easy, fun, excite, progress, thank, god, lucky
3.2 Flow Chart of Proposed Flow.
Figure 3.2 Flow chart of proposed work
Stop word
removal
Stemming
Pre-processed
data
Probability of
tweet category
Each word
probability in
tweet
Calculate probability
and non probability
for new label
Multilabel
classification
Probability of tweet
under each label
Collect new label :
“GOODTHING” to
classify with new
keywords
Twitter data
Extraction: Select
user
Collect Particular
user data
Detection of
learning experience
shared by tweet
tweet with highest
probability label

IV. Experimental Results
This section describes several experimental evolutions on inputted Student tweets on twitter extracted
from import.io specific dataset.
Existing system result
1st
train data: No. Of tweets: 90, No. Of label: 6 and keywords: 215
Fig 4.1 Probability of keywords falls in which label existing system
Fig 4.2 Result Category Probability of Keywords Labels existing system

Proposed system result
Fig 4.3 Improved Probability of “Goodthing” from “Others”
Fig 4.4 1st train data: Improved Category Probability of Keywords Labels
After Successfully Evaluate the result of Proposed Approach with compare to Existing Approach, we will
see a comparative study of both system.

Table 5.6 Comparison Table of Proposed and Existing system
Proposed System considering new label “Goodthing” and
“Others”
Existing System
Dataset :
(implemented)
 Real time Twitter data
 1st
train data: 90 tweets, 215 keywords-Real dataset
 2nd
train data:200 of tweets, 215 keywords-Real
dataset
Only Prudue University.
Parameter:  Probability of keywords.
 Probability of labels. (Category and non-category)
Probability of tweets under five labels.
Labels : Seven Labels:
HeavystudyLoad
Lack of Social-Engagement
Negative Emotion
SleepProblems
Diversity Issues
Others
Goodthing
Only consideration of five labels and
“Others” label as extra case to fetch
unwanted tweets.
Extractor: Use import.io extractor.
(collect large tweets monthly)
Prudue University
V. Conclusion
The conclusion to the research done till now is those researchers in learning analytics, educational data
mining, and learning technologies can be very advantageous. It provides a workflow for analyzing social media
data for educational purposes that overcomes the major limitations of both manual qualitative analysis and large
scale computational analysis of user-generated textual content.
From study of various labels for classifying student learning experiences’ many possibility of falling
students under that different category classified as per the analysed result, it is cleared that after implementing
proposed work many of the useful tweet which were going under the others and not considered. That all real
tweets were collected, analysed and found the probability of that tweets falling under “Goodthing”. By using
Naive Bayes probability rules, we classified the tweets for finding the Probability of the keywords and
probability tweets under labels.
From studies of different papers, only knowing the sentiment of student shared tweets does not provide
much actionable knowledge on relevant interventions and services for students. In sentiment analysis, tweets
can be classified positive, negative or neutral but cannot be judged that it belongs to which category. So, in the
proposed system multi label classification model was used which will make fall tweet into multi-categories at
same time.
The proposed system successfully improved the probability and category probability of the two labels:
“Others” and “Goodthing” labels. And concluded that student just don’t only post their bad experiences’ but
also good experiences’ on social media. And even used a new extractor import.io which helps in making
database and dataset and can be used as web crawler.
References
[1]. Xin Chen, Mihaela Vorvoreanu, and Krishna Madhavan “Mining Social Media Data for Understanding Students’ Learning
Experiences,” IEEE transactions on learning technologies, manuscript id 1, pp. 1-14, 2013.
[2]. Ajay Kumar Pal, Saurabh Pal, “Classification model of Predication for Placement of Students,” in I.J. modern Education and
Computer Science, pp. 49-56, November 2013.
[3]. Hsin-Ying Wu, Kuan-Liang Liu and Charles Trappey, “Understanding Customers Using Facebook Pages: Data Mining Users
Feedback Using Text Analysis”, IEEE 18th International Conference on Computer Supported Cooperative Work in Design, 2014,
pp. 346-350.
[4]. I-Hsien Ting, Shyue-Liang Wang, Hsing-Miao Chi and Jyun-Sing Wu, “Content Matters: A study of hate groups detection based on
social networks analysis and web mining ”, IEEE/ACM International Conference on Advances in Social Networks Analysis and
Mining, 2013,pp. 1196-1201.
[5]. Sara Keretna, Ahmad Hossny and DougCreighton,” Recognising User Identity in Twitter Social Networks via Text Mining”, IEEE
International Conference on Systems, Man, and Cybernetics, 2013 , pp. 3079-3082.
[6]. PAPER 5: Kathy Lee, Diana Palsetia, Ramanathan Narayanan, Md. Mostofa Ali Patwary, Ankit Agrawal, and Alok Choudhary,
“Twitter Trending Topic Classification”, 11th IEEE International Conference on Data Mining Workshops, 2011 , pp. 251-258.
[7]. N. Anitha, B. Anitha, S. Pradeepa, “Sentiment Classification Approaches – A Review,” in International Journal of Innovations in
Engineering and Technology, vol. 3, Issue 1, pp. 22-31, October 2013.
[8]. Nitin Pawar, Prof. S. K. Sonkar, “An Approach Towards E-Learning Using SVM Classification Technique and Ranking Technique
in Microblog Supported Classroom: A Survey,” in International Journal of Emerging Technology and Advance Engineering,
Volume 3, Issue 7, July 2013.
[9]. A. K. Santra, S. Jayasudha, “Classification of Web Log Data to Identify Interested Users Using Naïve Bayesian Classification,” in
International Journal of Computer Science Issues, Vol. 9, Issue 1, pp. 381-387, January 2012 .
[10]. C. D. Manning, P. Raghavan, and H. Schtze, “Introduction toInformation Retrieval.” New York, NY, USA: Cambridge University
Press, 2008.

Classification of Student’s E-Learning Experiences’ in Social Media via Text Mining

More Related Content

What's hot (20)

Similar to Classification of Student’s E-Learning Experiences’ in Social Media via Text Mining (20)

More from iosrjce (20)

Recently uploaded (20)

Classification of Student’s E-Learning Experiences’ in Social Media via Text Mining