Data Mining Methods in Educational Process Management
ANNA HOVAKIMYAN, SIRANUSH SARGSYAN
Department of Programming and Information Technologies,
Yerevan State University,
1, Alek Manukyan st., Yerevan, 0025,
ARMENIA
Abstract: - The paper addresses the challenge of effectively managing the educational process by leveraging
intelligent data analysis of student performance during learning activities. It introduces an approach centered
around data clustering, specifically applied to the study of programming disciplines and languages. By utilizing
clustering techniques, the paper aims to identify the most challenging topics within a given academic subject,
track students' learning paths, evaluate and enhance teaching methodologies, and create personalized learning
plans tailored to individual students' needs. This approach enables educators to better understand and address
the diverse learning requirements of students, ultimately enhancing the overall educational experience.
Key-Words: - educational process, student’s educational path, programming languages teaching, knowledge
testing environments, data mining, data clustering, adaptive teaching scenarios.
Received: March 3, 2024. Revised: September 3, 2024. Accepted: October 5, 2024. Published: November 6, 2024.
1 Introduction
At present, educational institutions and training
centers gather and retain a vast amount of data about
the educational procedure. This information
includes records of student enrollment and
attendance to courses, outcomes of ongoing and
final exams, charts displaying the degree to which
students have attained the educational objectives
stated in educational programs, and so on.
Analyzing this data provides valuable insights that
aid educational institution staff in efficiently
arranging the educational process, [1], [2], [3], [4].
When dealing with large sets of data, data
mining methods can be very useful. Note that
traditional data mining algorithms may not be
suitable for solving problems interested. Therefore
the development of new algorithms that offer
specific functionality and integrity is necessary, [1].
The process of data mining aims to extract
knowledge from large amounts of raw data and has
become a crucial aspect of decision-making
systems. Data mining techniques are introduced into
different research fields, such as statistics,
databases, machine learning, artificial intelligence,
data visualization, etc. [1]. It has gained significant
attention and has become an important component
of the activities in various organizations. For
example, in the service sector, data mining helps
analyze customer behavior and improve a
company's performance. In the healthcare sector, it
serves as an additional tool for doctors to diagnose
diseases. In the marketing sector, it helps to study
the market and its needs and to make informed
decisions, [1].
Data mining is a continuous cyclic process that
does not stop after finding a solution. The results
generated from data mining lead to new business
goals, which can be used to create more focused
models.
As we know, the process of data mining
includes the following stages: problem
identification, data collection and preparation,
model building, evaluation, and model deployment,
[1].
In the first stage, the subject area as well as the
problem are analyzed, project goals and
requirements are identified, the data mining task is
formulated, and a preliminary implementation plan
is developed, [1], [5].
The second stage involves gathering and
researching data. This step helps to determine how
effectively the data gathered solves the problem.
One can remove unnecessary data or add more data
to improve the accuracy of the results. During this
stage, the data is also subjected to statistical
analysis. Proper preparation of the data can
significantly enhance the quality of information that
must be extracted using data mining techniques. In
addition, this stage involves visualizing and
interpreting the data for various stakeholders, [6].
During the model construction and evaluation
stage, different modeling methods are selected and
WSEAS TRANSACTIONS on ADVANCES in ENGINEERING EDUCATION
DOI: 10.37394/232010.2024.21.13
Anna Hovakimyan, Siranush Sargsyan
E-ISSN: 2224-3410
110
Volume 21, 2024
applied, and model parameters as well as
hyperparameters are adjusted to optimal values. At
this point, it is crucial to evaluate how well the
model satisfies the initially stated business goal, [6].
The final stage is the deployment of the model,
where data mining is used in the target environment.
In this stage, the trained model can be used to make
decisions based on real-time data, [7].
In data mining, various approaches and
algorithms are used to find connections between
different data characteristics. These methods include
classification, regression, cluster analysis, factor
analysis, social network analysis, association rules
searching, sequential pattern analysis, etc. The
results are used to predict future events and the
forecast results are presented via various visual
representations such as pie charts, histograms, Gantt
charts, tables, etc. This helps people to better
understand the patterns and forecast results, which
can then be used for decision-making, [6].
2 Data Mining in Education
Educational Data Mining (EDM) is a specific
research field in Data Mining that focuses on using
various studies, techniques, and tools to extract
useful information from vast amounts of data related
to educational processes, [8], [9]. This data can
include various details such as the performance of
applicants in entrance exams, the results of students'
session exams, how the independent and research
work is carried out, the academic subjects that
students frequently access during online learning, as
well as the students’ preferred format of educational
materials (text, multimedia, etc.), [10].
Data mining is widely used in education to
analyze a large amount of data produced by
information systems in schools, universities, and
educational centers. Experts from various fields
such as computer and communication technologies,
pedagogy, psychology, and statistics work together
to improve the educational process by combining
traditional and innovative teaching and learning
methods. The ultimate goal is to enhance the quality
of education through advanced data analysis
techniques (Figure 1), [11], [12].
Automated data analysis through data mining in
the field of education provides detailed results that
are difficult for a person to obtain through manual
analysis. These results can be used by the learning
process management system to establish a
correlation between a student's educational path,
final grades, achievements, and educational goals
[9], [10], [13].
Fig. 1: Data mining infrastructure in education
Data mining can be a valuable tool for educators
and researchers in the field of education. It can
assist in analyzing curriculum to ensure high-quality
learning for students, altering the list and content of
academic disciplines and workshops according to
student and employer requirements, studying the
educational paths of students and their connection
with educational goals, identifying patterns and
anomalies in student work, suggesting the most
effective criteria for selecting courses in
asynchronous learning, and much more, [3], [4], [9],
[12].
To comprehensively evaluate the entire
educational process and identify patterns between
students' academic achievements and different
educational approaches, it is necessary to analyze
and visualize information. Research has shown that
pre-processing algorithms should be applied to
learning process data before any specific data
mining methods can be applied, [1], [3], [7].
Data mining methods are utilized in various
forms of learning, such as offline, online, and hybrid
formats. They are applied to analyze the learning
outcomes of individual subjects and the entire
educational program, [3], [11], [13]. In a single
subject, it is possible to examine the results of
mastery of individual topics and patterns that reflect
the pace of mastery of a specific topic, depending on
the degree of mastery of other topics. Ultimately,
this assessment can provide an evaluation of the
students' acquisition of knowledge, competencies,
and skills in the entire academic discipline, [9], [11],
[13].
It is possible to establish a correlation between
accessibility and the degree of mastery of specific
topics within an academic discipline. The same
method can be used to evaluate the entire
educational program and determine the extent to
which the educational goals have been met, [14].
WSEAS TRANSACTIONS on ADVANCES in ENGINEERING EDUCATION
DOI: 10.37394/232010.2024.21.13
Anna Hovakimyan, Siranush Sargsyan
E-ISSN: 2224-3410
111
Volume 21, 2024
Automated testing environments can generate
data that can be used to build machine learning
models. With a target variable in mind, regression
models can be constructed to predict expected
output values. These predictive models can be used
to develop warning and recommendation systems, to
prevent undesirable academic outcomes, [11], [15].
3 Evaluating the Effectiveness of
Implementing the Educational
Program
One of the key factors in evaluating the
effectiveness of an educational program is to
determine the extent to which students have
achieved the goals outlined in the program. This is
done by assessing the educational outcomes
produced by various components of the curriculum
including final and current exams, tests, independent
and research work, practice, projects, final papers,
and so on. The points obtained by the students for
these components are then used for certification
purposes. Every unit of the syllabus is linked with
indicators that show whether it covers educational
objectives in the form of knowledge, abilities,
competencies, and skills that students acquire.
Gathering, processing, and analyzing the data will
enable students to be grouped based on the extent to
which they have achieved the stated educational
objectives.
We represent the knowledge, skills,
competencies, and abilities acquired by a student
with an ID as a vector
j-th knowledge/ability/
competence/skill acquired by the student identified
by this ID.
Cluster analysis of data should be conducted on
both individual components and groups of
components while excluding others, [16]. Typically,
the number of clusters is set to three, corresponding
to the categories of "bad," "satisfactory," and
"good." Visualizing the analysis results can aid in
performing a SWOT analysis on the implementation
of an educational program by highlighting the
program's strengths, weaknesses, and possible risks.
The outcomes of the analysis can be used to take
corrective measures to eliminate disadvantages and
enhance educational processes.
As a part of this research, some methods have
been created to evaluate the outcomes of students'
education in programming subjects, including
programming fundamentals and programming
languages. To accomplish this, the e-judge
interactive environment was used, which enables
automated verification of code written in various
programming languages, [17].
As is known programming is a discipline that
involves creating algorithms to solve problems
effectively. A good algorithm should be optimal in
terms of complexity. For the student, it is essential
to have the ability to create correct algorithms and
evaluate their complexity. Programming training
involves mastering the syntax of a programming
language, as well as acquiring the skills to build
effective and accurate programs in that language.
This means that the program should work correctly
for all proposed test cases.
Automated systems designed to support
programming disciplines such as e-judge, Code
Signal, etc. are equipped with tools that help check
the syntax of programs in the desired programming
language. They support the launch of the programs
on specific input data and provide results on test
cases. These results are available in a convenient
format for both manual and automated analysis,
[17].
This paper examines the use of data intellectual
analysis methods to improve education by analyzing
data from knowledge-testing environments. The
purpose of this study is to create a toolset for
exploring students' educational paths, monitoring
their progress, identifying challenging topics, and
implementing effective teaching methods. To solve
the problem, data clustering is proposed, [16].
Our research is conducted to evaluate students’
achievements in programming disciplines. The
students are given tasks to programming in different
programming languages such as C, C++, and
Assembler. The students must implement the tasks
in the e-judge environment, [17]. The tasks are
provided with test cases that the e-judge system uses
to test the programs.
The e-judge system presents program execution
results on test cases in different forms. If there are
compilation errors in the program that point gaps in
learning by students the syntax of the programming
language, the system returns a "Compilation Error"
message. If not all tests have been passed, an
"Execution Error" message is returned. If the
program is ineffective and exceeds the maximum
operating time, the system returns a "Maximum
Operating Time has been Exceeded" message.
Finally, if the program executes successfully on all
the test cases, the system returns an "Awaits for
Confirmation" message, [17].
Data clustering can be used to identify how
students learn the syntax of programming
languages, whether they create complete programs
with all the necessary functionality, how effective
WSEAS TRANSACTIONS on ADVANCES in ENGINEERING EDUCATION
DOI: 10.37394/232010.2024.21.13
Anna Hovakimyan, Siranush Sargsyan
E-ISSN: 2224-3410
112
Volume 21, 2024
their programs are, and which tasks are the most
challenging for them.
The data that needs to be analyzed consists of
vectors that represent students. Each vector reflects
the progress made by the respective student in a
training module section and provides a summary of
his task completion. The student vector contains
various components such as the student's ID,
number of tasks completed, task numbers, number
of completed tasks that ended in a particular status,
and the dates of the first and last submission of
completed tasks. These vectors will be used to
cluster the data.
Clustering is a process of dividing a set of
objects into groups called clusters. The main goal is
to place similar objects in one cluster while keeping
significantly different objects in separate clusters,
[16]. A cluster can be defined as a group of objects
that share common properties. Data clustering helps
us to understand the key issues related to learning
and to identify weaknesses in the educational
process. The results of data analysis can be used to
create an individualized learning approach for each
student by developing an adaptive learning scenario,
[11], [18].
The process of cluster data analysis is carried
out in several stages. First, features are selected
based on which clustering will be performed. Next,
a measure of distance between objects is chosen.
Then, a clustering method is selected. Finally, the
reliability of the clusters is interpreted and assessed,
[16].
In this work, two cluster analysis methods were
used: the K-means method and the DBSCAN
method, [16].
The K-means method involves selecting K
(K>=2) centroids randomly among the data. The
distance from each data point to each centroid is
then calculated, and each data point is assigned to
the cluster with the nearest centroid. After this, the
centroids are recalculated by taking the average of
all points in a given cluster. The algorithm ends
when the centroids of the newly formed clusters do
not change, or the data remains in one cluster
without moving from one cluster to another, or if the
maximum number of assigned iterations has been
completed, [7], [15], [16].
The DBSCAN algorithm, short for Density-
based spatial clustering of applications with noise, is
a clustering technique that takes into account the
distribution density of a random variable. It works
by grouping points that are located close to each
other and labeling the points that are situated in
areas of low density as noise. The algorithm
identifies neighboring points within a specific
neighborhood of a chosen point. If the number of
such points exceeds a given threshold, a new cluster
is created. Otherwise, the point is marked as noise.
Then the algorithm assigns all the points from the
given neighborhood to the same cluster as the main
point. These steps are repeated for unvisited points
as well as for those that are marked as noise, [16].
Clustering provides an easily interpretable
pattern of results by using the centroid values of
each cluster. The centroid represents the most
typical data or prototype in a cluster. However, it
does not necessarily describe any specific instance
in that cluster.
This work addresses the challenges of analyzing
task complexity, tracking individual student
progress, and evaluating educational paths in a
course topic.
To categorize the tasks based on their level of
difficulty and identify topics that may require
further study, k-means cluster analysis is conducted
[16]. The problems are categorized into three
different levels of difficulty: easy, medium, and
advanced. This is accomplished by computing the
percentage of correct answers for each problem, and
then conducting cluster analysis to group the tasks
based on their level of difficulty for students (Figure
2). The DBSCAN method is utilized to assess
students' individual work on a specific task. The
data processed by the method pertains to each
student and the particular task, reflecting the amount
of effort the student put into solving the given task.
The data vector includes information about the
student's ID number, task number, number of
attempts, and the date when results such as
"compilation error," "working time exceeded," or
"incorrect answer" were recorded.
Fig. 2: Cluster analysis of tasks by complexity
The DBSCAN method clusters a set of these
vectors, and the results are visualized (Figure 3).
WSEAS TRANSACTIONS on ADVANCES in ENGINEERING EDUCATION
DOI: 10.37394/232010.2024.21.13
Anna Hovakimyan, Siranush Sargsyan
E-ISSN: 2224-3410
113
Volume 21, 2024
Fig. 3: DBSCAN analysis of one specific task
By using statistical analysis of data provided by
the e-judge system, one can generate summary
information that includes both solved and unsolved
tasks. This information can be used to develop an
effective learning strategy (Figure 4), [11], [18].
Fig. 4: Information about solutions to problems
The teacher can obtain information about the
student՛s progress regarding the specific task.
Filtering is performed by the student name and a
task, and the student’s progress in solving the
problem is visualized (Figure 5).
Fig. 5: Filtering by the student’s name and a specific
task
4 Conclusion
Most of the studies on the use of data mining
techniques in the field of education (EDM) focus on
the use of e-learning technologies such as Moodle,
WebCT, Blackboard, Classroom, etc., and student
knowledge automated testing using tools like
Google Forms, Moodle, e-judge, Code Signal, etc.,
[17], [19].
Currently, the main challenge is to evaluate the
level of attainment of the final educational outcomes
(learning outcomes), where student performance
plays a crucial role. Predicting student performance
based on their learning path can help both students
and teachers enhance their learning and teaching
methods. Frequently, the current research is limited
by the statistical techniques utilized for processing
data. The application of data mining methods to the
educational process is an important issue that
requires attention. This includes data collection,
problem formulation, clarification of the methods
used, determination of forecasting goals, and
practical application of the results obtained.
One practical way to implement research results
is to develop recommendation systems for
personalized learning. These systems will provide
students with a learning path that is tailored to their
needs. Recent research on personalized learning
systems examines some simple features such as
learner preferences, interests, and learning and
testing behavior.
This paper demonstrates the outcomes of a
cluster analysis that was applied to data related to
teaching programming and programming languages.
The purpose of this analysis is to identify the
shortcomings in the educational process and to
develop new teaching and learning strategies that
are tailored to the needs of the students.
If more data becomes available in the future, it
will be possible to apply additional data analysis
methods that will yield more effective results. It
would also be desirable to have the ability to
automatically connect to databases of automated
testing systems to process real-time data, [17], [19].
The analysis via data mining methods provides
insights into the importance of required courses in
the syllabus. A valuable tool for teachers is an
analysis tool that predicts the degree to which a
student will achieve her/his educational goals.
We consider also an opportunity to use machine
learning technologies to predict the expected
learning outcomes for a student, as well as to create
warning and recommendation systems to avoid
undesirable academic outcomes, [6], [15].
WSEAS TRANSACTIONS on ADVANCES in ENGINEERING EDUCATION
DOI: 10.37394/232010.2024.21.13
Anna Hovakimyan, Siranush Sargsyan
E-ISSN: 2224-3410
114
Volume 21, 2024
References:
[1] A. A. Barseghyan, M. S. Kupriyanov, I. I.
Kholod, M. D. Tess, S. I. Elizarov. Analysis of
data and processes: textbook. 3rd ed., BHV-
Petersburg, 2009 (in Russian).
[2] Коvalev E.E. A System Model and Tools for
Modernization of Federal and Regional
Digital Services of Statistics and Data
Analytics in Education, Lecture Notes in
Networks and Systems (Volume 1). Germany.
Springer Nature, 2021.
[3] Fiofanova O. A. Data Analysis Competencies
in Professional Standards: From Data-Experts
to Evidence-Based Education / Advances in
Natural, Human-Made, and Coupled
HumanNatural Systems Research, Lecture
Notes in Networks and Systems (Volume 1).
Germany, Springer Nature, 2021.
[4] Ch. Fischer, Z. A. Pardos, R. Sh. Baker,
.J.Williams, P. Smyth, R. Yu,
S.Slater,R.Baker, M. Warshauer, Mining Big
Data in Education: Affordances and
Challenges. Review of Research in Education.
Vol. 44, issue 1, 2020. pp. 130–160. doi:
https://doi.org/10.3102/0091732X20903304.
[5] Nong Ye. The Handbook of Data Mining.
CRC Press, 2004.
[6] Mohammed J.Zaki and Wagner Meira. Data
Mining and Machine Learning: Fundamental
Concepts and Algorithms. 2nd ed., Cambridge
University Press, 2020.
[7] Siranush Sargsyan, Anna Hovakimyan,
Varditer Kerobyan. An Approach to
Developing and Implementing a
Recommendation System. International
Journal of Economics and Management
Systems. ISSN: 2367-8925, Vol.7, 2022, pp.
270-273, [Online].
https://www.iaras.org/home/caijems/an-
approach-to-developing-and-implementing-a-
recommendation-system (Accessed Date:
October 20, 2024).
[8] A. H.Cairns, B.Gueni, М. Fhima, A. Cairns,
S.David, N. Khelifa. Process Мining in the
Education Doмain. International juornal on
Advances in Intellegent Systems, Vol.8,
no.1&2, 2015, pp.219-232, [Online].
https://www.thinkmind.org/library/IntSys/IntS
ys_v8_n12_2015/intsys_v8_n12_2015_18.ht
ml (Accessed Date: October 20, 2024).
[9] Ginica Mahajan, Bhavna Sahini,
EducationalData Minig: a state-of-the-art
survey on tools and techniques used in EDM,
International Journal of Computer
Applications & Information Tecnology,
Vol.12, No.1, 2020, pp. 310-316, [Online].
https://www.researchgate.net/publication/340
983783 (Accessed Date: October 20, 2024).
[10] Boyarinov D. A. Knowledge integration maps
in the context of the issue of automated
pedagogical information processing Problemy
sovremennogo obrazovaniya. 2019, No. 6, pp.
232–239, [Online].
http://www.pmedu.ru/images/2019-6/22.pdf
(in Russian). (Accessed Date: October 20,
2024).
[11] Khan M. A., Khojah M., Vivek V. Artificial
Intelligence and Big Data: The Advent of
NewPedagogy in the Adaptive E-Learning
System in the Higher Educational Institutions
of Saudi Arabia. Education Research
International. Vol.2, 2022, Article ID
1263555, 10p, doi:
https://doi.org/10.1155/2022/1263555.
[12] C.Romero, S.Ventura, М. Pechenizkiy and
R.Baker. Handbook of Educational Data
Мining., Taylor&Francis, 2010.
[13] V.Sothavilay, K. Yacef and R.A.Calvo,
Process mining to support student’s
collaborative writing, 3rd International
Conference on Educational Data Мining
Proceedings, Pittsburgh,PA, June 11-13,
2010, pp.257-266, [Online].
https://educationaldatamining.org/EDM2010/
uploads/proc/2010%20Proceedings%20Prefac
e,%20TOC.pdf (Accessed Date: October 20,
2024).
[14] М. Pechenizkiy, N.Treka, E.Vasilyeva and P.
De Bra. Process mining online assesment
data, 2nd International Conference on
Educational Data Мining (EDМ09)
Proceedings, Cordoba, Spain.July 1-3, 2009,
pp.279-288, [Online].
https://www.educationaldatamining.org/EDM
2009/uploads/proceedings/edm-proceedings-
2009.pdf (Accessed Date: October 20, 2024).
[15] Andreas C. Müller, Sarah Guido, Introduction
to Machine Learning with Python: A Guide
for Data Scientists, O’Reilly, 2017.
[16] L. Kaufman and P. Rousseeuw, Finding
groups in data: an introduction to cluster
analysis, Wiley&Sons, Inc. 1990.
[17] eJudge, [Online]. http://ejudge-y.ispras.ru/c/,
http://ejudge-y.ispras.ru/asm/ (Accessed Date:
June 26, 2024).
[18] Boyarinov D. A. Pedagogical Model for
Creating Individual Learning Paths Based
onEducational Maps. VI International Forum
on Teacher Education,ARPHA Proceedings 3,
WSEAS TRANSACTIONS on ADVANCES in ENGINEERING EDUCATION
DOI: 10.37394/232010.2024.21.13
Anna Hovakimyan, Siranush Sargsyan
E-ISSN: 2224-3410
115
Volume 21, 2024
Kazan, 2020, pp. 277–289.
https://doi.org/10.3897/ap.2.e0277.
[19] C.Romero, S.Ventura and E. Garcia. Data
mining in course management systems:
moodle case study and tutorial. Computers&
Education, Vol.51, No.1, 2008, pp.368-384,
https://doi.org/10.1016/j.compedu.2007.05.01
6.
Contribution of Individual Authors to the
Creation of a Scientific Article (Ghostwriting
Policy)
- Anna Hovakimyan carried out the gathering of
information from the e-judge system.
- Siranush Sargsyan was responsible for the
implementation of clustering algorithms.
Sources of Funding for Research Presented in a
Scientific Article or Scientific Article Itself
No funding was received for conducting this study.
Conflict of Interest
The authors have no conflicts of interest to declare.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en
_US
WSEAS TRANSACTIONS on ADVANCES in ENGINEERING EDUCATION
DOI: 10.37394/232010.2024.21.13
Anna Hovakimyan, Siranush Sargsyan
E-ISSN: 2224-3410
116
Volume 21, 2024
