WINTER SEMESTER 2022-23

CSE1901 - Technical Answers for Real World Problems (TARP)

DIGITAL ASSIGNMENT - 2

EASEL- ONLINE LEARNING PORTAL USING MACHINE LEARNING

Done by
Team NO: 17

SHIVANG KHER 20BCB0115

VARUN MURPANI 20BCE0719

KIRIT G PILLAI 20BCE2404

MANSI 20BCE2604

VANSH D JAIN 20BDS0125

SECTION I- LITERATURE REVIEW

1. The paper “EXECUTE: Exploring Eye Tracking to Support E-learning” by Ahsan Raza
Khan et al.

Problems addressed:

The COVID-19 epidemic has disrupted schooling and made remote instruction the norm. Online
teaching and learning include drawbacks, such as the absence of technology to measure student
participation and attentiveness.
This research develops an e-learning framework to capture and analyse student attention during
remote teaching sessions and profile their learning behaviour using eye-tracking data

Methodology:

● Developed a privacy-aware low-cost webcam-based eye-tracking interface to allow

remote learning.
● Developed a gaze-based attention score matrix to measure student involvement.
● Automated student involvement classification using ML and the new attention score
matrix.

Accuracy:

The results show that the logistic regression, SVM and polynomial regression have the highest
accuracy of 91.05%, 92.6% and 89.5%, respectively

Conclusion and Result:

Teaching and learning need attention monitoring. Remote learning might benefit from an
attention monitoring system that detects student engagement and redirects it. Online learning
lacks visual input for attention monitoring. Recently, pricey commercial-grade eye-tracking
technologies have been employed to measure attention in academic settings, restricting the
research to laboratories. A webcam-based attention categorization system was described in this
study.

Limitation:

Technical limitations:
Complex system requirements
2. The paper “An overview of Machine Learning Technologies and their use in E-learning”
by Ramzi FARHAT et al.

Problems addressed:

Limited understanding of the potential of machine learning technologies for enhancing the
e-learning experience

Methodology:

Various Machine Learning algorithms such as K-means, SVM, and Decision Tree, among others,
were used to improve Technology Enhanced Learning Environments (TELE).

Accuracy:

K-means, SVM, and Naive Bayes algorithms provided 97% accuracy, Backpropagation
algorithm provided 87% accuracy, Decision tree accuracy provided 85% accuracy, and K-modes
clustering provided 89% accuracy.

Conclusion and Result:

E-learning researchers have used machine learning to improve learning experience. Since
e-learning centres on the student, this makes sense. To our knowledge, no study has used
learning data to assess and enhance content quality.The major goal is to enable course creators
reengineer education using machine learning and various aspects, including prior learners'
interactions.

Limitation:

Limited Scope
3. The paper “EVALUATING THE ACCEPTABILITY AND USABILITY OF EASEL: A
MOBILE APPLICATION THAT SUPPORTS GUIDED REFLECTION FOR
EXPERIENTIAL LEARNING ACTIVITIES” by Jerry Schnepp et al.

Problems addressed:

To examine the early perceptions (acceptability) and usability of EASEL (Education through
Application-Supported Experiential Learning), a mobile platform that delivers reflection prompts
and content before, during, and after an experiential learning activity

Methodology:

Development of a mobile application that facilitates real-time guided reflection and to determine
the relevant feature set, we conducted a needs analysis with both students and faculty members.
Data collected during this stage of the evaluation helped guide the creation of a prototype. The
user experience of the prototype and interface interactions were evaluated during the usability
phase of the evaluation study.

Accuracy:

Feedback from this study guides our current software development. EASEL is currently in the
alpha phase of development. Much of the core functionality has been implemented. However, the
user interface and user experience still require refinement and testing. The initial results reported
in this work have been immensely helpful to guide our efforts

Conclusion and Result:

In the task analysis phase, initial questionnaire data revealed that participants typically did not
reflect on experiences unless a reflection activity was assigned to them. After being introduced to
the concept of EASEL by viewing an animation, most members of a second group of participants
indicated they would recommend including EASEL as a part of an experiential learning activity.
They cited the usefulness of prompts and ease of use as the application’s most valuable features.
Results collected during the usability phase indicated that participants could complete tasks using
an EASEL prototype. They found the application easy to use and most would anticipate using the
system frequently.

Limitation:

E-Learning, still tends to struggle with student feedback.

4. The paper “Improvement of an Online Education Model with the Integration of Machine
Learning and Data Analysis in an LMS” by William Villegas-Ch et al.

Problems addressed:

the learning management systems (LMSs), which, in recent years, have lost prominence because
they are considered by certain institutions as simple repositories. This vision took a drastic turn
due to current circumstances, in which LMS is the medium that allows students to maintain
interaction with their institutions

Methodology:

This objective is outlined in a new normality that seeks robust educational models, where certain
activities are carried out in an online mode, surrounded by technologies that allow students to
have virtual assistants to guide them in their learning.The system after the analysis adjusts the
activities that are presented as an evaluation mechanism and aligns them with the characteristics
and needs of the student.

Accuracy:

The algorithm performed the analysis of 51 instances to identify the reason why the scores in the
evaluations present a performance below the expected. Of the 51 instances, 48 were classified as
correct, with 94.1176%. This value was considered as true to assume the decision of the analysis.

Conclusion and Result:

This work allows technology to become the ideal assistant for both students and teachers.
Further, it allows the management of calendars to the students, as well as the generation of
events, reminders, and notifications that indicate to the student what activities must be fulfilled,
and based on its results, it can carry out a continuous accompaniment that allows the student to
improve their performance.

Limitation:

Data Breach
5 .The paper "Analyzing and Predicting Students’ Performance by Means of Machine
Learning: A Review (MDPI)" by Juan et al. (2022)

Problem addressed:

The article aims to analyze the existing literature on the topic, classify the techniques and
objectives into categories, and present the results of the literature review through tables and
graphs. The study considers 64 recent articles (90% published in the last 6 years) from various
online databases and journal sources, with a focus on machine learning techniques. The results
show the most common techniques, objectives, and student levels used in the studies and the
algorithms and methods applied.

Methodology:

A qualitative research study was conducted on 64 recent articles related to techniques for
predicting student behavior.
The papers were filtered based on quality, contribution, peer-review, and impact factor, and 35%
of the analyzed papers were journal articles.
The literature review was mainly classified based on techniques and objectives, and the
techniques were implemented using various algorithmic methods.
The results of the study were presented through graphs and tables that summarize the techniques,
objectives, student level, and algorithms used in the reviewed articles.

Accuracy:

The supervised learning method of Rule Induction produced a 94% result for retention
prediction, while SVM with SMOTE generated a 90.24% outcome. A number of algorithms
were suggested to identify struggling students with 97% accuracy, including Sequential Forward
Selection (SFS), C4.5, Random Forest (RF), K-Nearest Neighbors (KNN), and Naive Bayes
(NB). Out of all the algorithms, Artificial Neural Networks (ANN) and Support Vector Machines
(SVM) showed the highest accuracy.

Conclusions and Results:

The conclusion derived from the text is that a qualitative research study was conducted to
analyze the different techniques used for predicting students' behavior. The study was based on
64 recent articles related to the subject and was conducted by reviewing literature from various
online databases. The articles were screened to ensure that only high-quality, peer-reviewed
sources were included. The literature review was then classified into techniques and objectives.
The techniques were analyzed to reach the objectives of each reference and were implemented
through several algorithmic methods. The results showed that the academic level of the students
can play a significant role in predicting their performance and that different datasets built from
students of different levels can have different latent factors.

Limitations :

The study only considered articles from a limited number of online databases and excluded
papers without impact factor or not peer-reviewed.
The study was qualitative in nature and relied on a literature review, rather than original data
collection and analysis.
The study was limited to a specific academic level and may not be applicable to other levels or
contexts.

6 .The paper " e-Learning: Challenges and Research Opportunities Using Machine
Learning & Data Analytics (IEEE) by Abdallah Moubayed et al

Problem addressed:

The problems addressed are the challenges facing the field of e-learning, which include issues
with content transmission and delivery, and the limitations of enabling technologies. The paper
proposes the use of machine learning (ML) and data analytics (DA) techniques as potential
solutions to overcome these challenges, and highlights the importance of these techniques in
improving the learning experience. The text presents several research opportunities for using ML
and DA techniques in e-learning, and provides an overview of the different challenges in the
field, as well as the different machine learning and data analytics techniques.

Methodology:

The methodology of utilizing big data in e-learning is presented in the text. With more students
accessing online education, big data is being generated and machine learning and data analytics
are proposed as solutions. Different types of machine learning algorithms are described,
including supervised, unsupervised, semi-supervised, and reinforcement learning, each having a
unique approach to training..

Accuracy:

According to the study referenced (Hogo,2010) the Fuzzy C-Means (FCM) algorithm was
outperformed by the Kernel Fuzzy C-Means (KFCM) algorithm. The results showed that the
KFCM algorithm had an accuracy rate of approximately 78%, while the FCM algorithm was less
accurate.
.Conclusions and Results:

The growth of data from various devices requires analysis to extract useful information, which
machine learning and data analytics can provide. These algorithms are proving to be beneficial in
fields like healthcare, business, and energy. The field of e-learning, which has gained interest due
to technology proliferation, faces challenges like content delivery and personalization. This
survey explores machine learning and data analytics algorithms, their uses in different fields, and
the e-learning field specifically, including its definitions, challenges, and examples of
implementation. Research opportunities have also been proposed.

Limitations :

The focus of previous works has only been on comparing the delivery mechanisms in terms of
resources rather than their impact on student learning.
The field of e-Learning still requires more research opportunities to personalize the learning
experience for students based on learning preferences.
There is a need for more research to determine the relationships between certain activities and
their impact on students' performance and engagement.

7 .The paper " Eye gaze tracking-based adaptive e-learning for enhancing teaching and
learning in virtual classrooms.” by Rastrollo-Guerrero et al

Problem addressed:

Students today prefer to conceal their questions and avoid approaching their teachers.Especially
in the case of virtual classrooms, the staff has no way of knowing whether or not the material is
tailored to the aptitude of the students.

Methodology:

The EyeTribe Tracker is an eye-tracking system that can determine where a person is looking by
extracting information from their face and eyes. The eye gaze coordinates are calculated relative
to the screen a person is looking at and are represented by a pair of (x, y) screen coordinates.

Accuracy:

The accuracy is not mentioned. However, the marks scored correspond well to the attention paid
by the student.

Conclusions and Results:

The proposed eye gaze tracking-based adaptive e-learning system can facilitate learning in
virtual classrooms by making the work of teachers and students easier when combined with
conventional instruction. The effectiveness of learning and accessibility of course materials to
the intended audience can be enhanced. Our system can also be utilised for business purposes,
assisting companies in gauging the allure of advertisements to potential customers. Eye tracking
as a field of study is still in its infancy, and EyeTribe is the first affordable tracker of its kind;
however, it lacks the precision of other high-end options.

8 .The paper "DEVELOPMENT OF AN E-LEARNING WEB PORTAL: The Foss

Approach " by Azeta A. A. Oyelami M. O. Ayo C. K.

Problem addressed:

Electronic learning (e-Learning), facilitated by communication networks, has made it possible

for individuals to receive education from anywhere at any time using the Internet, wide area
networks, or local area networks. Despite the central role that e-Learning applications play in the
learning process, they can be challenging to develop using proprietary programming tools. This
is due to the complex nature of acquiring the necessary tools and licenses, as well as the high
cost involved in purchasing and using them to develop large software applications. This has
limited the widespread usage of such systems.To address these challenges, the use of open source
applications has become a viable solution. Open source applications offer high flexibility and
provide software engineers and institutions with the ability to reuse, study, distribute, and
customize the code to meet user requirements.

Methodology:

The approach taken in creating the e-Learning system involves two steps: Firstly, obtaining the
open source e-Learning software (PHPBB) from the Internet, and secondly, modifying and
customizing the software to fit the local environment. The development process utilized tools
such as PHP for the front-end, Apache for the middleware, and a MySQL database for the
back-end.In conclusion, the authors have presented a novel approach to e-Learning by
developing an e-Learning class bulletin board that enables virtual interaction between students
and lecturers in an academic setting. The system is unique in its utilization of Free and Open
Source Software (FOSS), which allowed for an efficient and cost-effective solution compared to
using proprietary programming tools. The paper highlights the benefits of using open source for
e-Learning, including its cost-effectiveness, flexibility, and speed of development. The
 Conclusions and Results:

authors' aim was to create a system that fosters better interaction among members of the
academic community and provides a low-cost method for disseminating educational resources.
With the growth of Information and Communication Technology (ICT), open source is becoming
a popular choice for e-Learning solutions in developing economies, and the authors hope that
their work will introduce academic communities to e-Learning and online learning.

Limitations:

Other Content Management Systems (CMS) could be able to accomplish some of the
functionalities attained in the built e-Learning application.
The authors' main goal is to construct an e-learning system that is flexible, affordable, and quick
by adapting already-existing open source code, especially when contrasted to trying to
accomplish the same thing using proprietary programming.

SECTION II- LITERATURE SURVEY TABULAR FORMAT

S.no Problem Methodology Dataset Used Accuracy Limitations
Addressed
Adopted

1 The COVID-19 Developed a The dataset The results Technical

epidemic has privacy-aware includes the show that the limitations:
[1] Khan, A. disrupted low-cost following: logistic Complex
R. et al schooling and webcam-based Eye regression, system
made remote eye-tracking movements, SVM and requirements
instruction the interface to Gaze polynomial
norm. Online allow remote patterns , regression
teaching and learning. User have the
learning •Developed a interactions, highest
includes gaze-based User accuracy of
drawbacks, such attention score demographic 91.05%,
as the absence of matrix to s Learning 92.6% and
technology to measure student outcomes 89.5%,
measure student involvement. respectively
participation and • Automated
attentiveness. student
This research involvement
develops an classification
 e-learning using ML and
framework to the new
capture and attention score
analyse student matrix.
attention during
remote teaching
sessions and
profile their
learning
behaviour.
using
eye-tracking
data

2 Limited Various K-means, Limited Scope

understanding of Machine A collection SVM, and
[2] Farhat, the potential of Learning of datasets Naive Bayes
R.et al machine algorithms such was used. algorithms
learning as K-means, provided
(Not
technologies for SVM, and 97%
specified)
enhancing the Decision Tree, accuracy,
e-learning among others, Backpropagat
experience were used to ion algorithm
improve provided
Technology 87%
Enhanced accuracy,
Learning Decision tree
Environments accuracy
(TELE). provided
85%
accuracy, and
K-modes
clustering
provided
89%
accuracy.
3 To examine the Development of E-Learning,
early a mobile Understandi Feedback still tends to
[3] Jerry perceptions application that ng student's from this struggle with
Schnepp et (acceptability) facilitates study guides student
online
al. and usability of real-time our current feedback.
EASEL guided behavior and
software
(Education reflection and to their
through determine the development.
relations EASEL is
Application-Sup relevant feature
ported set, we inside real currently in
Experiential conducted a class the alpha
Learning), a needs analysis phase of
gradings
mobile platform with both development.
that delivers students and
reflection faculty This Dataset Much of the
prompts and members. Data was core
content before, collected during compiled functionality
during, and after this stage of the after 4 has been
an experiential evaluation months of an implemented.
learning activity helped guide Algorithm
However, the
the creation of a Introductory
Class at a user interface
prototype. The
user experience Brazilian and user
of the prototype University. experience
and interface still require
interactions refinement
were evaluated and testing.
during the The initial
usability phase
results
of the
evaluation reported in
study. this work
have been
immensely
helpful to
guide our
efforts
4 the learning This objective The dataset The Data Breach
management is outlined in a of e-learning algorithm
[4] systems (LMSs), new normality research and performed
Villegas-Ch that seeks publication
which, in recent the analysis
et al. robust during the
years, have lost educational COVID-19 of 51
prominence models, where pandemic, instances to
because they are certain which was identify the
considered by activities are indexed by reason why
certain carried out in an Scopus in the scores in
institutions as online mode, 2020. The the
surrounded by dataset
simple evaluations
technologies contains data
repositories. that allow authors, present a
This vision took students to have authors ID performance
a drastic turn virtual Scopus, title, below the
due to current assistants to year, source expected. Of
circumstances, guide them in title, volume, the 51
in which LMS is their issue, article instances, 48
learning.The number in
the medium that were
system after the Scopus,
allows students analysis adjusts DOI, link, classified as
to maintain the activities affiliation, correct, with
interaction with that are abstract, 94.1176%.
their institutions presented as an index This value
evaluation keywords, was
mechanism and references,
considered as
aligns them corresponde
with the nce Address, true to
characteristics editors, assume the
and needs of the publisher, decision of
student. conference the analysis.
name,
conference
date,
conference
code, ISSN,
language,
document
type, access
type, and
EID.
5.Analyzing The article aims A qualitative A group of The study only
and to analyze the research study data sets was The considered
Predicting existing was conducted utilized in supervised articles from a
learning limited number
Students’ literature on the on 64 recent the study.
method of of online
Performance topic, classify articles related The specific databases and
by Means of the techniques to techniques datasets Rule
excluded papers
Machine and objectives for predicting were not Induction without impact
Learning: A into categories, student named. produced a factor or not
Review and present the behavior. 94% result peer-reviewed.
(MDPI) results of the The papers for retention The study was
prediction, qualitative in
literature review were filtered
while SVM nature and
through tables based on relied on a
and graphs. The quality, with SMOTE
literature
study considers contribution, generated a review, rather
64 recent peer-review, 90.24% than original
articles (90% and impact outcome. data collection
published in the factor, and 35% Sequential and analysis.
Forward The study was
last 6 years) of the analyzed
Selection limited to a
from various papers were specific
online databases journal articles. (SFS),Rando academic level
and journal The literature m Forest and may not be
sources, with a review was (RF), KNN, applicable to
focus on mainly and Naive other levels or
Bayes contexts.
machine classified based
learning on techniques (NB),etc
techniques. The and objectives, were
results show the and the suggested.
most common techniques were Among them,
techniques, implemented Artificial
objectives, and using various Neural
student levels algorithmic Networks
used in the methods. (ANN) and
studies and the The results of Support
algorithms and the study were Vector
methods presented Machines
applied. through graphs (SVM)
and tables that showed the
summarize the highest
techniques, accuracy
objectives,
 student level, with 97%
and algorithms accuracy.
used in the
reviewed
articles.

The problems The A group of According to The focus of

6. addressed are methodology of data sets was the study previous works
the challenges utilizing big utilized in referenced has only been
e-Learning: on comparing
facing the field data in the study. (Hogo,2010)
Challenges the delivery
of e-learning, e-learning is The specific the Fuzzy mechanisms in
and
which include presented in the datasets C-Means terms of
Research
issues with text. With more were not (FCM) resources rather
Opportunitie
content students named algorithm than their
s impact on
transmission and accessing was
Using student
delivery, and the online outperformed
Machine learning.
Learning & limitations of education, big by the Kernel
The field of
Data enabling data is being Fuzzy e-Learning still
Analytics technologies. generated and C-Means requires more
(IEEE) The paper machine (KFCM) research
proposes the use learning and algorithm. opportunities to
of machine data analytics The results personalize the
learning (ML) are proposed as showed that learning
experience for
and data solutions. the KFCM
students based
analytics (DA) Different types algorithm on learning
techniques as of machine had an preferences.
potential learning accuracy rate There is a need
solutions to algorithms are of for more
overcome these described, approximatel research to
determine the
challenges, and including y 78%, while
relationships
highlights the supervised, the FCM between certain
importance of unsupervised, algorithm activities and
these techniques semi-supervised was less their impact on
in improving the , and accurate. students'
learning reinforcement performance
experience. The learning, each and
engagement.
text presents having a unique
several research approach to
opportunities for training..
 using ML and
DA techniques
in e-learning,
and provides an
overview of the
different
challenges in the
field, as well as
the different
machine
learning and
data analytics
techniques.

Students today
7. [7] prefer to conceal The EyeTribe The author The The main
Rastrollo-G their questions Tracker is an collected effectiveness problem faced
uerrero et al and avoid eye-tracking data locally of learning was ensuring
approaching system that can using their and the accuracy of
their teachers. determine own setups. accessibility the EyeTribe
Especially in the where a person Data of course readings, which
case of virtual is looking by columns materials to could be
classrooms, the extracting include: : the intended affected by the
staff has no way information fixations, audience can students' head
of knowing from their face fixations be enhanced. movements.
whether or not and eyes. The duration Our system Further analysis
the material is eye gaze mean, can also be was also
tailored to the coordinates are complete utilised for required to
aptitude of the calculated fixation business determine the
students. relative to the times at AOI purposes, effectiveness of
screen a person assisting the content.
is looking at companies in
and are gauging the
represented by a allure of
pair of (x, y) advertisemen
screen ts to potential
coordinates. customers.
 Electronic Other Content
8 learning The e-Learning Not Included Not Management
(e-Learning), system was Mentioned Systems (CMS)
[8] Azeta, A developed using
which facilitates could be able to
et al. two different
education using accomplish
communications methodologies: some of the
networks, has Start by functionalities
made learning downloading attained in the
possible from PHPBB, a free built e-Learning
anywhere at open source application.
anytime by e-learning The authors'
using the programme, main goal is to
Internet, wide from the construct an
area networks or Internet. e-learning
local area Second, system that is
networks. broaden and flexible,
Notably,Learnin modify FOSS to affordable, and
g applications fit local quick by
which have conditions. adapting
become central Tools like PHP already-existing
to the learning as the front-end, open source
process may be Apache as the code, especially
developed using middleware, when
proprietary and MySQL as contrasted to
programming the back-end trying to
tools. were used to accomplish the
create the same thing
system. using
proprietary
programming.
REFERENCES:

1) Khan, A. R., Khosravi, S., Hussain, S., Ghannam, R., Zoha, A., & Imran, M. A. (2022,
March). Execute: Exploring eye tracking to support e-learning. In 2022 IEEE Global
Engineering Education Conference (EDUCON) (pp. 670-676). IEEE.
2) Farhat, R., Mourali, Y., Jemni, M., & Ezzedine, H. (2020, February). An overview of
Machine Learning Technologies and their use in E-learning. In 2020 International
Multi-Conference on:“Organization of Knowledge and Advanced Technologies”(OCTA)
(pp. 1-4). IEEE.
3) William Villegas-Ch 1,* , Milton Román-Cañizares 1 and Xavier Palacios-Pacheco 2 (4
August 2020 ) Improvement of an Online Education Model with the Integration of
Machine Learning and Data Analysis in an LMS. MDPI .
4) Jerry Schnepp , Christian Rogers ,EVALUATING THE ACCEPTABILITY AND
USABILITY OF EASEL: A MOBILE APPLICATION THAT SUPPORTS GUIDED
REFLECTION FOR EXPERIENTIAL LEARNING ACTIVITIES. JITE.IIP.ORG.
Volume 16, 2022
5) Rastrollo-Guerrero, J. L., Gómez-Pulido, J. A., & Durán-Domínguez, A.
(2020,February). Analyzing and predicting students’ performance by means of machine
learning: A review. Applied sciences, 10(3), 1042.
6) Moubayed, A., Injadat, M., Nassif, A. B., Lutfiyya, H., & Shami, A. (2018,July).
E-learning: Challenges and research opportunities using machine learning & data
analytics. IEEE Access, 6, 39117-39138.
7) Joe Louis Paul, I., Sasirekha, S., Uma Maheswari, S., Ajith, K. A. M., Arjun, S. M., &
Athesh Kumar, S. (2019). Eye gaze tracking-based adaptive e-learning for enhancing
teaching and learning in virtual classrooms. In Information and Communication
Technology for Competitive Strategies: Proceedings of Third International Conference on
ICTCS 2017 (pp. 165-176). Springer Singapore.
8) Azeta, A. A., Oyelami, M. O., & Ayo, C. K. (2008). Development of an e-learning web
portal: The FOSS approach. Turkish Online Journal of Distance Education, 9(2),
186-199.