© 2020 JETIR April 2020, Volume 7, Issue 4 www.jetir.

org (ISSN-2349-5162)

Deep Learning in the field of Education

Dr. Divya Kashyap
Assistant Professor,
Indian Academy Degree College Autonomous, Bangalore.

Abstract: Educational Data Mining could be get made and multi modal data from student’s
a analysis field that focuses on the applying of learning activities in academic settings. For
machine learning, information mining and instance, these platforms record when the
statistical methods to detect patterns in students access a learning object, how many
massive collections of instructional
times they accessed it, whether the answer
information. Various machine learning
techniques area unit applied throughout this provided to an exercise is correct or not, or the
field over the years, but it’s been recently that amount of time spent reading a text or watching a
Deep Learning has gained increasing video.
attention among the academic domain. Deep
Learning may well be a machine learning
technique supported neural network
architectures with multiple layers of method All this data will be analyzed to handle
units, that has been successfully applied to a completely different educational problem, like
broad set of problems among the areas of generating recommendations, developing
image recognition and language method. This adaptive systems, and providing automatic
paper surveys the analysis dole out in Deep grading for the students’ assignments. completely
Learning techniques applied to [EDM], from Different machine learning techniques have been
its origins to the current day. The most goals
applied over time to research this information,
of this study to square measure to spot the
[EDM] tasks that have benefited from Deep however it's been in recent years that the
Learning and people that square measure utilization of Deep Learning techniques has
unfinished to be explored, to explain the main emerged within the field of EDM.
data sets used, to produce an outline of the key
ideas, main architectures, and configurations The topic of Deep Learning (DL) has gained
of Deep Learning and its applications to EDM, increasing attention within the industry and
and to debate current progressive and future research areas within the last decade,
directions on this space of analysis. revolutionizing the sector of machine learning by
getting progressive ends up in perception tasks
1. Introduction like image and speech recognition[2]. Major
companies such as Google, Facebook, Microsoft,
The analysis field of Educational Data Mining Amazon, and Apple are heavily investing in the
focuses on the applying of techniques and development of software and hardware
strategic of data mining in educational innovations in this field, trying to leverage DL
environments. EDM is concerned with potential in the production of smart products.
researching, developing, and applying machine DL is based on neural network architectures with
learning, data processing, and applied math multiple layers of processing units that apply
strategies to notice patterns in giant collections of linear and nonlinear transformations to the input
educational data that will preferable be not data. These architectures can be applied to all
possible to research [1]. type of data: image, audio, text, numerical, or
some combination of them. Many research fields
EDM leverages e-learning platforms like have benefited from applying these technologies,
Learning Management System (LMS), Intelligent and EDM is not an exception.
Tutoring Systems (ITS), and, within the last
In the previous few years there has been a
years, huge Open on-line Courses (MOOC), to
proliferation of analysis within the Educational
JETIR2004395 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 1441
© 2020 JETIR April 2020, Volume 7, Issue 4 www.jetir.org (ISSN-2349-5162)

Data Mining field using DL architectures. This complex. For example, if we want to create a
article presents a review on the literature of DL model that predicts what the house price in
methods applied to Educational Data Mining, some region next year will be, based on
from its start in 2015 to the current day. The situation on the market for the last three years,
primary contributions of this article are as we would feed the model metrics such as house
follows:
prices on the market in last three years, interest
rates and salary rates.

1.Summarize the most EDM tasks and classify The output would be the house price prediction
the present works that have applied DL on every for the next year. The process through which a
of those tasks. model learns how to make sense of input data is
2.Identify the tasks that have gained major called “model training”. Training is a key
attention and those that are still unexplored. concept in machine learning.

3.Describe and categorise the most public and The best way to describe the potential of machine
personal datasets employed to train and test DL learning is to explore how people and companies
models in Educational Data Mining tasks. are currently taking advantage of it. Some
examples could be:
4.Introduce key DL ideas and technologies,
describing the techniques and configurations
most generally utilized used in Educational Data
Mining and its specific tasks.  Natural language processing: - Google
Translate is created from a set of machine
5.Discuss future directions for analysis in DL
applied to Educational Data Mining based on the learning algorithms that updates the service
data gathered during this study. over time based on input from users, like
new words and syntax. Siri, Alexa, and,
2. Machine learning and its application in last,Cortana, and, Google Assistant all rely
education area on natural language processing to recognize
speech and synthesis, permiting them to
Machine learning can be considered a part of know or pronounce words they have never
Artificial Intelligence [AI]. Machine learning is the encountered before.
method of granting a machine or model access to
information and property it learn for itself. In 1959,  Recommendation systems: - On Amazon,
Arthur Samuel came up with the good concept we Netflix, Google, etc., everything that's
must alaways not need to teach computers, but suggested to you depends on your search
rather, we have a tendency to might allow them to activity. These websites deliver
their own. He coined the term “machine learning” recommendations across platforms, devices,
to describe his theory, which is now a standard and apps. Machines match buyers with
definition for the ability of computers to learn sellers and their products, digital content
autonomously [3]. with viewers who want to see them - all of
which improves our online experiences
Machine learning is programming computers to
significantly. The company even owns a
optimize a performance criterion using example
patent for Anticipatory Shipping, a system
data or experience [4] . Implementing a machine
that ships a product to the closest warehouse
learning algorithm means implementing a model
thus you'll be able to order and receive your
that outputs correct information given that we
item on the constant day although it's unclear
have provided input data. You can think of a
whether or not they have enforces it yet.
model as a black box: data go in at the
beginning, and some other data go out at the
end — but the processes in between are
JETIR2004395 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 1442
© 2020 JETIR April 2020, Volume 7, Issue 4 www.jetir.org (ISSN-2349-5162)

 Algorithmic trading: - Algorithmic trading is 3. Research

a process that involves random behaviour,
ever-changing data, and a variety of factors.
While financiers cannot predict all of that The above-mentioned research questions were
behaviour, machine learning algorithms will extracted when we studied the aims of this
— and that they reply to changes within the research. To answer these research questions,
market quicker than somebody's. we performed a literature study. We decided to
use the SLR method to collect the relevant
primary studies and followed the guidelines
There are many alternative business given by[5].
implementations of machine learning, heap of
them area unit in education area . a number of
fascinationg areas are: In the next phase, we created four categories
and classified the papers into those. The reason
 Predict Student Performance (A nice and
for this classification was that most of the
great application of machine learning is
papers published were relatively distinctive in
predicting student performance. By learning
terms of the research objective, methodology,
concerning every student, the machine
and application. To be simplest as possible, we
learning model will conclude weaknesses
created four categories of studies without
and suggests ways that to enhance like extra
ignoring the variations of themes. This way, we
lectures or study extra literature ).
examined the research papers that fell under the
 Test Students & Grade Students Fairly same. The final number of papers collected was
(Machine learning can help creating 67. Some studies was included in more than just
computerized adaptive assessments . The one category.
machine learning based assessment provides
constant feedback to faculty members and
students regarding how the student learns, The research categories and the total number of
the support they have and also the progress selected studies are as follows:
towards their learning goals ).
(A) Grading students (12 studies)
 Improve Retention (Machine learning, like
learning analytic, will also help improve (B) Improving student retention (17 studies)
retention rates. By identifying “at risk”
(C) Predicting student performance (42 studies)
students, facilities will reach intend on those
students and obtain them the assistance they (D) Testing students (6 studies)
have to achieve successful ).
Letters A, B, C and D are just used as tags and
 Support academics and institution stuff serve as a mark for further referencing in the rest
(Machine learning based mostly algorithms of this study.
will facilities with classification of students
written assessment papers) 4. Result
This section describes the main characteristics
of each category. We present all four main
categories in subsections, supported with
relevant papers.

JETIR2004395 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 1443
© 2020 JETIR April 2020, Volume 7, Issue 4 www.jetir.org (ISSN-2349-5162)

1. Grading students recent years in this area, as we said before. For

example, study[6] employs the machine learning
approach called the Recursive Clustering method
Machine learning can grade students by
to group the students of the programming course
removing human biases. Some recent examples
into groups based on their performance within the
are use of the supervised Machine Learning for
necessity courses, co-requisite and current course
text classification to predict students’ final course
work result. Students present in the lower groups
grades in some course and exhibited the potential
will be taken into consideration since they are
of using ML classified messages to identify
highly prone to fail.In another interesting study in
students at risk of course failure [9]. In addition,
this category, authors have proposed a new model
there are aims to improve the assessment of
to categorize students into three categories to
problem solving in education by employing
determine their learning capabilities and to help
language technologies and computational-
them to improve their studying techniques. They
statistical machine learning methods to grade
have chosen the state of the art of machine
students' natural language responses automatically
learning approach to classify student's nature of
[10]. Great example of use machine learning for
study by choosing distinguished options of their
grading students is by comparing their actions to
activity in their academic field. They have chosen
a model of expert behaviour [11].
a data driven approach where key factors that
determines the base of student and classify them
into high, medium and low ranks . Yet another
2. Improving student retention study in this category, but from other perspective
brings .
As we said before, by identifying at risk students
early, faculties will find and call those students
and facilitate them to be more successful. Student 4. Testing students
retention is an essential part of many enrollment
The machine learning based assessment provides
systems. It affects most all segments of university
constant feedback to lecturers, students and
school or college metrics: name, financials,
parents regarding however the student learns, the
reputation, ranking. Specially, student retention
support they need and the progress they are
has become one of the most important things for
making towards their learning goals. Authors in
managers in higher education institutions. There
study [10] introduced a system for training of
are few studies, which developed models to
students' ability to construct correct proofs in
predict and to introduce the reasons behind
propositional or predicate logic. In addition to
student’s number decreasing.
common techniques,as well as shows supported
by slides and exercises they used animations,
which were based on carefully selected
3. Predicting student performance demonstrative examples and their step-by step
solutions. as to check students' knowledge, they
Probably a serious good thing anout machine prepared a questionnaire that captured the whole
learning (regarding range of studies in scientific method of a logic proof construction. A student
databases) is its ability to predict student constructed a proof and then answered queries
performance. By “learning” regarding every from the questionnaire. They described the
student, the technology can identify weaknesses design of the questionnaire and discussed its
and suggests ways in which to enhance, like dis/advantages. At the end, they then applied
additional practice tests. This appears to be very frequent sub graph mining along with supervised
popular analysis trend; there are lot of studies in
JETIR2004395 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 1444
© 2020 JETIR April 2020, Volume 7, Issue 4 www.jetir.org (ISSN-2349-5162)

machine learning algorithms to perform an early as possible. It affects university rankings,

automatic evaluation of correctness of the proofs. faculty name, school reputation, and financial
wellbeing. Student retention has become one in
Conclusion akk the foremost vital priorities for decision
This study has reviewed the emergence of DL makers in higher education institutions, so there
applications to EDM, a trend that started in are lot of studies in that category. Reviewing
2015 with three papers published, increasing its studies under category marked as C, showed us
presence once a year so far with 19 papers how major good thing of machine learning
published in 2019. After a systematic search, 41 regarding number of studies in scientific
works were retrieved in this area. It is worth databases is its ability to predict student
mentioning the presence of these approaches in performance. By “learning” about each student,
relevant EDM forums such as the annual the technology can identify weaknesses and
International Conference in Educational Data suggests ways to improve. According to our
Mining, with 7 papers published in the last survey, this is most interesting area of machine
edition (for a total of 16 in the last three learning application to researchers. There are lot
years).Based on the taxonomy of EDM of studies in recent years in that category, and
applications defined by[11] , only 4 of the 13 lot of machine learning models were provided
tasks proposed in that study have been to predict student performance on totally
addressed by DL techniques. This reveals that different parameters. We might say that this
there are several open opportunities for the use category is definitely the trend.Reviewing
of DL in unexplored EDM tasks, moreover studies under category marked as D, showed
taking into account the promising results some models however machine learning will
obtained by these models in the works reviewed favilitate move far from standardized testing.
(67% of them reported that DL outperformed Machine learning based assessment provides
the “traditional” machine learning baselines in constant feedback to lecturers, students and
all their experiments).The aim of this study was parents regarding how the student learns, the
to judge the present state of the art within the support they need and the progress they're
application of machine learning in education creating towards their learning goals. As we
space. The amount of studies (papers and have found earlier, research has been made over
articles) was large, so only some of studies, several relevant databases, but of course, not all
which we found as good representatives, were were involved,so this can be considered as
mentioned in results this study. This study limitation of study. In addition, there is a
shows that there a square measure important possibility that some of the relevant studies may
other ways to profit from machine learning be skipped by chance.
application in education area.As we tend to In the future, we have a tendency to arrange to
expressed in introduction section, one of our implement own machine learning model for
goals was attempt to classify studies within the suggesting potential student to enroll or not to
field of machine learning application in enroll on University College algebra, Study of
education space. Based on our survey, the Software Engineering, based on different
papers reviewed under category marked as A parameters. As we have rich database with lot
research a ways how machine learning can of information of students on previous years,
grade students by removing human biases we believe that study would be of help to
(fairly grading).Reviewing studies under support our admission office as help in student
category marked as B, showed how machine- enrolment process.
learning algorithms will facilitate faculties or
schools or faculties to reach out to students and
get them the help they need to be successful as
JETIR2004395 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 1445
© 2020 JETIR April 2020, Volume 7, Issue 4 www.jetir.org (ISSN-2349-5162)

References
1. C. Romero, S. Ventura, M. Pechenizkiy, and R.
Baker, Handbook of educational data mining,
CRC Press, 2010.
2. I. Goodfellow, Y. Bengio, and A. Courville, Deep
Learning, MIT Press, Cambridge, Mass, USA,
2016.
3. Samuel, A.L. (1959) Some Studies in Machine
Learning Using the Game of Checkers, IBM
Journal of Research and Development, vol. 3,
no. 3, pp. 210-229, doi: 10.1147/rd.33.0210
4. https://www.thetechedvocate.org/8-ways-
machine-learning-will-improve-education,
(2018), Accessed on: 2018- 09-25
5. Kitchenham. B.A., Charters,S. (2007)
Guidelines for performing Systematic
Literature Reviews in Software Engineering.
Version 2.3, EBSE Technical Report EBSE-
2007-01, Keele University, Keele, Staffordshire,
United Kingdom
6. Wu, J. Y., Hsiao, Y. C., & Nian, M. W. (2018).
Using supervised machine learning on large-
scale online forums to classify course-related
Facebook messages in predicting learning
achievement within the personal learning
environment. Interactive Learning
Environments, 1-16.
7. Wang, H. C., Chang, C. Y., & Li, T. Y. (2008).
Assessing creative problem-solving with
automated text grading. Computers &
Education, 51(4), 1450-1466.
8. Abbott, R. G. (2006, June). Automated expert
modeling for automated student evaluation. In
International Conference on Intelligent
Tutoring Systems (pp. 1-10). Springer, Berlin,
Heidelberg.
9. Anand, V. K., Rahiman, S. A., George, E. B., &
Huda, A. S. (2018, March). Recursive clustering
technique for students' performance evaluation
in programming courses. In Majan
International Conference (MIC), 2018 (pp. 1- 5).
IEEE.
10. Vaculík, K., Popelínský, L., Mráková, E., &
Jurco, J. (2013, October). Tutoring and
automatic evaluation of logic proofs. In
European Conference on e-Learning (p. 495).
Academic Conferences International Limited.
11. B. Bakhshinategh, O. R. Zaiane, S. ElAtia,
and D. Ipperciel, “Educational data mining
applications and tasks: A survey of the last
10 years,” Education and Information
Technologies, vol. 23, no. 1, pp. 537–553,
2018.

JETIR2004395 Journal of Emerging Technologies and Innovative Research (JETIR) www.jetir.org 1446