JRSSEM 2022, Vol. 01, No. 7, 795 – 808
E-ISSN: 2807 - 6311, P-ISSN: 2807 - 6494
DOI : 10.36418/jrssem.v1i7.109
MODELING DATA MINING ALGORITHM FOR PREDICTING
TIMELY STUDENT GRADUATION AT STATE UNIVERSITY
SURABAYA
I Kadek Dwi Nuryana*
Information System Faculty of Engineering, State University of Surabaya, Indonesia
e-mail: dwinuryan[email protected]
*Correspondence: dwin[email protected].id
Submitted: 24 January 2022, Revised: 7 February 2022, Accepted: 18 February 2022
Abstract. Information technology aims to advance human activities. Even now it has penetrated
into the field of education, as a process and industrial sector, education cannot be separated from
the field of information and technology. In the implementation of the academic process, the State
University of Surabaya, which has 30,000 students, faces several obstacles in predicting the
students' graduation time on time. The information system owned has not been used optimally to
predict the time of student graduation. It is difficult to predict because this university does not
have an exact time prediction pattern to use as a basis for predicting the number of students who
will graduate on time. In helping these problems, the researcher provides a solution, namely
building a data mining algorithm model to predict the exact time of graduation for Surabaya State
University students. The data used are Strata-1 (S1) PTI, SI and TI students from 2014-2018. The
methodology in this study is FAST (Framework For The Application Of System Thinking) using the
Neural Network (NN) algorithm 761 student data with the input value of the artificial neural
network method 4, hidden layer 5 and output 2 providing an accuracy of 99.85%. Details of late
predictions are 98.06% as many as 101 students and correct predictions are 99.85% as many as
657 students. Shows that the Neural Network can be used as a prediction of the graduation of
Strata-1 (S1) PTI, SI, and IT students at the State University of Surabaya.
Keywords: data mining; neural network (NN) algorithm; prediction of student graduation.
I Kadek Dwi Nuryana | 796
DOI : 10.36418/jrssem.v1i7.109
INTRODUCTION
The development of information
technology shows very rapid progress.
These advances have made it easier for
humans to connect with one another. This
is also a fact, almost all aspects of industry
and modern society are constrained by the
breadth of information and technology,
because facts have proven that they can
provide countless values and benefits for
the development of human civilization
(Peng, 2021).
Information technology aims to
advance human activities. Even recently
penetrated into the field of education, as a
field of process and industry, education
cannot be separated from the field of
information and technology. Even UNESCO
officials and scholars believe that the
development of world information
technology has the greatest influence on
the education sector (Júnior & dos Santos,
2015). According to (Shi et al., 2017), the
peak of the application of information
technology in education will be
revolutionary, influencing the
transformation of the teaching process
from primary school to higher education or
university-level schools (Hubalovsky,
Hubalovska, & Musilek, 2019).
According to (Xiang, Magnini, &
Fesenmaier, 2015) in the world of
education, the use of information
technology is needed to support the
publication of institutions and the
implementation of the Tri Dharma
program. In today's world of education,
almost all sectors have utilized information
technology, including universities (Habib,
Jamal, Khalil, & Khan, 2021). Related to the
management of the Tri Dharma program,
universities, and university accreditation,
the role of information technology is
needed. Information technology is one of
the key factors for universities to advance
institutions, publish works, compete with
other universities, and facilitate the
teaching and learning process, all of which
are stages that cannot be avoided by
educational institution administrators
(Ngari et al., 2013).
In the implementation of the academic
process, the State University of Surabaya,
which has 30,000 students, faces several
obstacles in predicting the students'
graduation time on time. The information
system owned has not been used optimally
to predict the time of student graduation
(Handayani et al., 2017). It is difficult to
predict because this university does not
have an exact time prediction pattern to
use as a basis for predicting the number of
students who will graduate on time. The
academic information system owned by
this university is actually website-based,
but the information system has not been
used optimally to predict student
graduation times. Data mining is expected
to be useful and valuable information for
universities (Aldowah, Al-Samarraie, &
Fauzy, 2019).
The research that was conducted by
(Agrusti, Bonavolontà, & Mezzini, 2019)
with the title "Comparison of Data Mining
Classification Methods for Predicting
Student Graduation". In this study, a
comparison of data mining methods,
namely the Neural Network, K-Nearest
Neighbor and Decision Tree which were
applied to student graduation data from
the three methods was carried out, it was
797 | Modeling Data Mining Algorithm for Predicting Timely Student Graduation at State
University Surabaya
found that the Neural Network method, it
was found that the neural network method
was the best way to solve the problem of
predicting student graduation. Compared
with the k-nearest neighbor method and
the Decision Tree (Adeniyi, Wei, &
Yongquan, 2016).
To overcome this problem, the
researcher provides a solution by creating
a data mining implementation model to
predict the students' graduation time on
time at the State University of Surabaya.
The data mining used in this research is the
Algorithm Neural Network for
undergraduate PTI, SI and IT students. This
model is expected to help Surabaya State
University in particular and Indonesian
universities globally in solving the problem
of predicting the exact time of graduation
for students.
METHODS
The research framework
Entitled "Modeling Data Mining
Algorithms to Determine Student
Academic Achievement and Predicting the
Right Time for Student Graduation at State
University of Surabaya" originated from the
phenomenon of academic organization.
The process at the State University of
Surabaya faces several obstacles in
predicting the exact time of student
graduation. These problems arise because
the administrative base applied is still
paper-based and not yet computerized.
Determination of student academic
achievement is still using a manual process
(Wang & Holcombe, 2010), namely by
selecting student scores from the courses
taken and sorting out the student grades
for each department and student GPA per
semester.
The procedure is carried out to predict
the exact time of graduation of students at
this university is still difficult. This is
because the university has not made an
exact pattern of graduation time as a basis
for predicting the number of students who
graduate on time based academic
information system website, but the
information system has not been
implemented optimally to predict the exact
time of student graduation. Data mining
method is an algorithm method used by
researchers to assist universities in
overcoming these problems.
In helping this problem, the researcher
provides a solution, namely building a data
mining algorithm model to predict the
exact time of graduation for students at the
State University of Surabaya.
The framework for this research is
depicted in Figure 2.
I Kadek Dwi Nuryana | 798
Figure 2. Schematic Framework for Thinking
1. System Development Methodology.
The research entitled "Data Mining
Algorithm Modeling in Predicting Timely
Graduation of State University Students
of Surabaya" was built using the FAST
(Framework For The Application Of
System Thinking) methodology
(Agarina & Sutedi, 2017). The steps
carried out in this study are based on
the FAST methodology: 1. Initial
definition range, 2. Problem analysis, 3.
Needs analysis, 4. Logical design, 5.
Decision analysis, 6. System testing.
2. Research Materials Research
materials are the elements needed
by researchers when conducting
research. The elements needed in this
research process are the Neural
Network (NN) algorithm described
below:
a. Research Object.
The object of the research of
this research is Surabaya State
University.
b. Research Data
Data used to analyze and
predict the exact time of graduation
of students at the State University
of Surabaya using the Neural
Network (NN)lectures, student
grade data, semester GPA data (per
semester), GPA data (final), and
student graduation data.
c. Data Collection Method The data
Collection method in
conducting the analysis was carried
out on the prediction of the exact
time of graduation of students at
the State University of Surabaya
using the Neural Network (NN)
algorithm. This will be explained
below:
1) Observations
Observations were carried
out by researchers to find out
academic information system
data to be analyzed for research
START
Data from
Data mining algorithm processing with Neural Network (NN)
algorithm to predict student graduation time accurately
Data mining algorithm modeling with Neural Network
(NN) algorithm to predict the exact time of student
graduation
Provide recommendations to university leaders to make decisions from data mining algorithm models to
determine student academic achievement and predict graduation mahasiswa yang tepat waktu
END
799 | Modeling Data Mining Algorithm for Predicting Timely Student Graduation at State
University Surabaya
purposes. at the State University
of Surabaya.
2) Library Research
Is a data collection method
that is carried out by collecting
data from various sources that
support research with the
Neural Network (NN) algorithm
by searching for information
through the internet or
scientific journals, international
journals, papers, books,
proceedings, or other articles
that support research. The
results of library research are in
the form of models and the
latest developments on
concepts, data mining, data
mining classification, Neural
Network (NN) algorithms,
association rules, and other
supporting theories.
RESULTS AND DISCUSSION
In this study the research method used
is the Framework for the Applications of
System Technology (FAST) method with the
following stages:
1. Initial definition range,
The data used comes from the
Directorate of Information Technology,
State University of Surabaya in the form
of undergraduate student data (S1) PTI,
TI, SI batch 2014-2018. The data
obtained is the output of several
existing systems, so that several stages
of processing must be carried out for
the accuracy of graduation to be
determined based on the length of the
study. The grades range from 3.5 to 7
years. Based on the available programs,
it is determined that on-time
graduation occurs when a student has
successfully completed his studies in
3.5 and 4 years. If it exceeds then it is
considered not on time.
2. Problem analysis,
According to (Hla & Teru, 2015) the
definition of scope, problem analysis
through PIECES (Performance,
Information, Economics, Control,
Efficiency, and Service) can be
determined in the following table:
Table 1. Analysis of System Problems using PIECES.
Parameter
Solution
Performance
Optimization of information
systems to predict the exact time
of graduation for students.
Information
Processing data mining
algorithms with Neural Network
(NN) to accurately predict
student graduation times
I Kadek Dwi Nuryana | 800
Parameter
Solution
Economy
of information costscosts
because the system can be
accessed by related units so that
information can be viewed
digitally.
Control
With the information system , the
data and information conveyed
is more controlled and accurate.
Efficiency
Through the information system
the required resources are less
because the data and
information are obtained with an
integrated system.
Service
The data and information
needed by the unit can be
obtained more quickly and there
is no need to wait for data and
information.
3. Needs analysis,
In the needs analysis stage, the
information system to be built is
determined based on the definition of
the scope and problem analysis. Based
on the scope of analysis of the
questions above, a system of
requirements is needed in the form of
student data from academic Siakad.
The functional requirements of the
system are as follows (Kobayashi,
Morisaki, Atsumi, & Yamamoto, 2016):
a. Data taken from the academic
information system (SIAKAD)
includes:
1. Student grade
2. Data, semester GPA data (per
semester),
3. GPA data (final).
b. Student graduation data.
The data used in this study is
student data for semester GPA (per
semester) and GPA data (end), Odd
Semester Period for the 2018-2019
academic year and Even Semester
2019-2020. The data used is data from
PTI, SI, and IT students from the 2014-
2018 class in the Department of
Informatics Engineering (JTIF). The
number of data obtained is 762 data.
4. Logical
a. Design, Process Design of Neural
Network Algorithms.
The design of the neural
network process consisting of
801 | Modeling Data Mining Algorithm for Predicting Timely Student Graduation at State
University Surabaya
processes, training, and testing can
be seen in the following figure:
Figure 3. Design of Neural Network Algorithms
5. Decision analysis
a. Decision Analysis in predicting
student graduation.
Algorithm Back propagation
neural network, the parameters
displayed are training cycle (epoch),
learning rate, momentum and
hidden layer. Learning Rate is 0.1.
The experimental results can be
seen in Table 2 below.
Table 2. Experiment Value of Training Cycle (Epoch)
Training Cycle (Epoch)
Learning Rate
Momentum
Accuracy
200
400
600
800
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
%
99.48
81.73%
64.7 %
Based on the experimental
results above, the value of the
Training Cycle (Epoch) was chosen,
which was 400, and the accuracy
was 84.59%. This value is then used
in experiments to determine the
Learning Rate. Get the Learning
Rate value by entering a value
between 0.2 and 0.9. The
experimental results can be seen in
Table 3 below.
Table 3. Experiment Determination of Learning Rate
Training Cycle (Epoch)
Learning Rate
Momentum
Accuracy
400
400
400
400
0.2
0.5
0.7
0.9
0.1
0.1
0.1
0.1
85.12%
91.89%
89.91%
94, 84%
I Kadek Dwi Nuryana | 802
The two experiments above use
a parameter value of 400 in the
Training Cycle (epoch), and learning
rate 0.9. The maximum accuracy of
input node number 5 is 94.84%. This
study uses 5 input nodes. The
experimental results are as follows
table 4 shows :
Table 4. Experiment Number of Input Nodes
Number of Input
Nodes
Training Cycles
(epoch)
Learning
Rate
Momentum
accuracy
(%)
2
400
0.9
0.1
94
3
400
0.9
0.1
92.6
4
400
0.9
0.1
93.35
5
400
0.9
0.1
94.84
6. System Testing,
The structure of the Neural Network
(NN) network consists of 3 layers,
namely the input layer, the hidden
layer, and the output layer. In the input
layer there are 4 nodes as input, namely
(GPA 1, GPA 2, GPA 3 and GPA 4), the
hidden layer is composed of 5 nodes,
and the output layer is 2 nodes. If the
final GPA is above 2.75 it means on
time, if it is below 2.75 it means the
opposite). Its structure is shown in
Figure 4.
Figure 4. Structure of the Neural Network (NN)
Result set cross validation is shown in the graph as shown in Figure 5 below:
Figure 5. Cross Validation result set.
803 | Modeling Data Mining Algorithm for Predicting Timely Student Graduation at State
University Surabaya
System testing using the Neural
Network , the AUC value illustrated in
the ROC graph can be seen in the
following figure:
Figure 6. The AUC graph ROC
Determine the level of prediction
accuracy from the error matrix obtained
by processing 761 data using 75%
training data, namely 571 and 25%
testing data 190 to test the neural
network algorithm method, as shown in
Table 5 below:
Table 5. Neural Network Algorithm Method
Conducted using Neural
Network showed final accuracy of
99.61%, late detail prediction 98.06% as
many as 101 students, and accurate
prediction accuracy 99.85% as many as
657 students.
7. Implementation
Implementation Prediction of
student graduation for each batch from
the predictions of the 2014 – 2018
batch as follows:
a. Graduation Prediction for the 2014
batch.
Determining the level of
accuracy of the 2014 graduation
prediction from the error matrix
obtained in 81 data with the neural
network algorithm, as shown in
Table 6 below:
Table 6. Accuracy Value of Graduation Prediction Class of 2014
I Kadek Dwi Nuryana | 804
DOI : 10.36418/jrssem.v1i7.109
Tests conducted using the
Neural Network showed final
accuracy of 97.50%, late detail
prediction of 90% of 9 students, and
accurate prediction of 98% accuracy
of 70 students. The AUC value
illustrated in the ROC graph can be
seen in the following figure:
Figure 7. The AUC Graph ROC predicts graduation class 2014
b. Prediction Graduation batch 2015.
Determines the level of
prediction accuracy for class 2015
graduation from the error matrix
obtained in 165 data with neural
network algorithms, such as It can
be seen in Table 7 below:
Table 7. Accuracy Value of Graduation Prediction Class of 2015
Tests conducted using the
Neural Network showed final
accuracy of 83.79%, late detail
prediction 71.01% as many as 49
students, and accurate prediction
accuracy 92.71% as many as 89
students. The AUC value illustrated
in the ROC graph can be seen in the
following figure:
Figure 8. The AUC Graph ROC for 2015 graduation
c. Predictions 2016 Graduation
Predictions
Determines the accuracy level of
2016 graduation predictions from
the error matrix obtained in 128
805 | Modeling Data Mining Algorithm for Predicting Timely Student Graduation at State
University Surabaya
data with neural network
algorithms, as shown in Table 8
below:
Table 8. Accuracy Value of Graduation Prediction for Class of 2016
Tests conducted using the
Neural Network showed final
accuracy of 100%, prediction of
100% late details for 2 students, and
ccurate prediction accuracy of
100% with 126 students. The AUC
value illustrated in the ROC graph
can be seen in the following figure:
Figure 9. The AUC Graph ROC predicts graduation class 2016.
d. Prediction Graduation class 2017
Determines the level of
prediction accuracy for class 2017
graduation from the error matrix
obtained in processing 128 data
with neural network algorithms, as
shown in Table 9 below:
Table 9. Accuracy Value of Graduation Prediction for Class of 2017
Tests conducted using the
Neural Network showed final
accuracy of 98.95%, late detail
prediction 0%, and accurate
prediction accuracy of 98.94% as
many as 187 students. The AUC
value illustrated in the ROC graph
can be seen in the following figure:
I Kadek Dwi Nuryana | 806
DOI : 10.36418/jrssem.v1i7.109
Figure 10. The AUC Graph ROC predicts graduation class 2017
e. Prediction Graduation class 2018
Determines the level of
prediction accuracy for class 2018
graduation from the error matrix
obtained in processing 198 data
with a neural network algorithm, as
shown in Table 10 below:
Table 10. Accuracy Value of Graduation Prediction for class of 2018
Tests conducted using the
Neural Network showed final
accuracy of 98%, late detail
prediction of 71.43% for 10
students and accurate prediction
accuracy of 100% for 184
students.AUC illustrated in the ROC
graph can be seen in the following
figure:
Figure 11. The AUC Graph ROC predicts graduation for class 2018.
CONCLUSIONS
In predicting the graduation of
undergraduate students (S1) PTI, SI, and IT,
State University of Surabaya from the 2014
to 2018 batch, an algorithm is used neural
networks. Algorithm Neural Network \
used is a Neural Network, the input value
for the artificial neural network method is 4,
hidden layer 5 and output is 2, the accuracy
is 99.85%. Detailed information on
predictions of 98.06% late for 101 students,
and accurate predictions for 99.85% for 761
807 | Modeling Data Mining Algorithm for Predicting Timely Student Graduation at State
University Surabaya
students. Predictions for graduation for
each batch are as follows:
1. Prediction of PTI Graduation 2014
batch of final accuracy 97.50%,
Prediction of late details 90% as many
as 9 students, and 98% correct
prediction as many as 70 students.
2. Prediction of Graduation PTI, SI, TI 2015
final accuracy 83.79%, late detail
prediction 71.01% as many as 49
students, and accurate predictions
92.71% as many as 89 students.
3. Prediction of Graduation PTI, SI, TI 2016
final accuracy 100%, 100% late detail
predictions as many as 2 students, and
100% accurate predictions as many as
126 students.
4. Prediction of Graduation PTI, SI, TI class
2017 final accuracy 98.95%, late detail
prediction 0%, and predictions right
98.94% as many as 187 students.
5. Predictions for Graduation PTI, SI, TI
2018 final accuracy 98%, late detail
prediction 71.43% as many as 10
students, and exactly 100% as many as
184 students.
6. Based on these results, it can be
concluded that the Neural Network can
be used to predict the graduation of
undergraduate (S1) PTI, TI, and SI at the
State University of Surabaya from 2014
to 2018.
REFERENCES
Adeniyi, David Adedayo, Wei, Zhaoqiang, &
Yongquan, Y. (2016). Automated web
usage data mining and
recommendation system using K-
Nearest Neighbor (KNN) classification
method. Applied Computing and
Informatics, 2(12), 90–108.
https://doi.org/10.1016/j.aci.2014.10.0
01
Agarina, Melda, & Sutedi, Sutedi. (2017).
Information System Design Of Goods
Stock Using Framework For The
Application Of System Thinking (FAST)
Method (Case Study CV. Aneka Mandiri
Lestari). International Conference on
Information Technology and Business
(ICITB), 125–128.
Agrusti, Francesco, Bonavolontà,
Gianmarco, & Mezzini, Mauro. (2019).
University dropout prediction through
educational data mining techniques: A
systematic review. Journal of E-Learning
and Knowledge Society, 15(3), 161–182.
https://doi.org/10.20368/1971-
8829/1135017
Aldowah, Hanan, Al-Samarraie, Hosam, &
Fauzy, Wan Mohamad. (2019).
Educational data mining and learning
analytics for 21st century higher
education: A review and synthesis.
Telematics and Informatics, 3(7), 13–49.
https://doi.org/10.1016/j.tele.2019.01.0
07
Habib, Muhammad Nauman, Jamal,
Waseef, Khalil, Uzma, & Khan,
Zunnoorain. (2021). Transforming
universities in interactive digital
platform: case of city university of
science and information technology.
Education and Information
Technologies, 6(1), 517–541.
I Kadek Dwi Nuryana | 808
Handayani, Putu Wuri, Hidayanto, Achmad
Nizar, Pinem, Ave Adriana, Hapsari, Ika
Chandra, Sandhyaduhita, Puspa
Indahati, & Budi, Indra. (2017).
Acceptance model of a hospital
information system. International
Journal of Medical Informatics, 9(9), 11–
28.
https://doi.org/10.1016/j.ijmedinf.2016.
12.004
Hla, Daw, & Teru, Susan Peter. (2015).
Efficiency of accounting information
system and performance measures.
International Journal of
Multidisciplinary and Current Research,
3(2), 976–984.
Hubalovsky, S., Hubalovska, Marie, &
Musilek, Michal. (2019). Assessment of
the influence of adaptive E-learning on
learning effectiveness of primary school
pupils. Computers in Human Behavior,
9(2), 691–705.
https://doi.org/10.1016/j.chb.2018.05.0
33
Júnior, José Policarpo, & dos Santos,
Claudemir Inacio. (2015). “ Learning to
live together in peace and harmony”:
um olhar comprometido com a
formação humana. Conjectura: Filosofia
e Educação, 20(2), 94–111.
Kobayashi, Nobuhide, Morisaki, Shuji,
Atsumi, Noritoshi, & Yamamoto,
Shuichiro. (2016). Quantitative Non
Functional Requirements evaluation
using softgoal weight. J. Internet Serv.
Inf. Secur., 6(1), 37–46.
Ngari, S. M., Ndungu, A., Mwonya, R.,
Ngumi, O., Mumiukha, C.,
Chepchieng, M., & Kariuki, M.
(2013). Levels of stress among
secondary school administrators
and its implication in education
management in Kenya. Educational
Research and Reviews, 8(11), 677–
681.
https://doi.org/10.5897/ERR10.021
Peng, Bingzhuan. (2021). Intercultural
Communicative Competence Teaching
and Assessment Based on Modern
Information Technology. International
Journal of Emerging Technologies in
Learning, 16(7).
Shi, Victor Guang, Baines, Tim, Baldwin,
James, Ridgway, Keith, Petridis,
Panagiotis, Bigdeli, Ali Ziaee, Uren,
Victoria, & Andrews, Daniel. (2017).
Using gamification to transform the
adoption of servitization. Industrial
Marketing Management, 6(3), 82–91.
https://doi.org/10.1016/j.indmarman.2
016.12.005
Wang, Ming Te, & Holcombe, Rebecca.
(2010). Adolescents’ perceptions of
school environment, engagement, and
academic achievement in middle
school. American Educational Research
Journal, 4(3), 633–662.
https://doi.org/10.3102/000283120936
1209
Xiang, Zheng, Magnini, Vincent P., &
Fesenmaier, Daniel R. (2015).
Information technology and consumer
behavior in travel and tourism: Insights
from travel planning using the internet.
Journal of Retailing and Consumer
Services, 2(2), 244–249.
https://doi.org/10.1016/j.jretconser.20
14.08.005
© 2022 by the authors. Submitted
for possible open access publication
under the terms and conditions of the Creative
Commons Attribution (CC BY SA) license
(https://creativecommons.org/licenses/by-sa/4.0/).