p–ISSN: 2723 - 6609 e-ISSN: 2745-5254
Vol. 5, No. 11 November 2024 http://jist.publikasiindonesia.id/
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5480
Application of Data Mining for Prediction of High School Student
Graduation Rates
Muhamad Kurniawan
1*
, Sani Muhamad Isa
2
Universitas Bina Nusantara, Indonesia
Email:
[email protected].id
1*
, [email protected]
2
*Correspondence
ABSTRACT
Keywords: Educational
Data Mining; Mean Squared
Error; student grades;
classification model; feature
selection
The implementation of Data Mining in the education sector aims
to develop methods that are able to discover valuable knowledge
from data generated in the educational environment. This can be
used to increase learning efficiency by paying more attention to
students who are predicted to have low grades. However, in its
application, each algorithm shows different performance
depending on the attributes and dataset used. In this study, a
dataset of semester grades and final school exam scores was used.
Some of the prediction techniques used are decision trees, support
vector machines, and neural networks. Of the four scenarios for
the science major at SMAN 2 and SMAN 3 Pangkalpinang with 3
different models, the Mean Squared Error value shows that the test
results are in accordance with the testing dataset and can be used
as predictions of students' final grades, namely the decision tree
model and support vector machine. For the Social Sciences major
at SMAN 2 and SMAN 3 Pangkalpinang with 3 different models,
the Mean Squared Error value shows that the test results are in
accordance with the testing dataset and can be used as a prediction
of students' final grades, namely the support vector machine
model.
Introduction
Education is one of the most essential parts of people's lives. It is used to enhance an
individual's academic and financial growth (Akour & Alenezi, 2022; Li, 2024). An educated
individual must contribute not only to his family but also to society and community. This is all
achieved through proper learning. To be able to produce human resources who are capable,
insightful, competitive, and creative, educational institutions are required to provide high-quality
education for their students. The implementation of this system is achieved through the
administration of school or national examinations (Akala, 2021; Wajdi et al., 2020), which serve
as a criterion for graduation at the respective educational level and provide a standard for
evaluation within the educational system. In the Implementation of Examinations Organized by
Education and National Examination Units, participants who are taught or educated can be
declared to have passed if they have completed the study period and obtained attitude or behaviour
scores. with a minimum good category score and can take part in tests held by the specified agency
or educational unit (Kementerian Pendidikan dan Kebudayaan, 2019). SMA Negeri 2
Pangkalpinang City and SMA Negeri 3 Pangkalpinang City are the high school education units
located in the Bangka Belitung Islands Province where this educational unit annually carries out
School Examinations; SMA Negeri 2 Pangkalpinang city and SMA Negeri 3 Pangkalpinang city
experienced several cases of students who got School Examination (US) scores that were below
average.
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5481
This is due to students' lack of preparation in the subjects that will be tested in the School
Examination. The method that will be used in data mining is to produce information that will be
useful so that decisions are made appropriately by the school to understand what strategies need
to be applied to students and later can place more emphasis on areas of deficiency. This is done
so that students can pursue fields that suit their respective abilities and continue to study other
subjects with as little stress as possible. The benefit in the future is that it will be easier for students
to determine their career path so that they do not fall into fields that are not suitable. This is an
anticipatory step for the school to help students who have the potential to face obstacles in learning
and lack achievement in the learning process.
The implementation of Data Mining in the education sector, known as Educational Data
Mining (EDM) (Du et al., 2020; Fernandes et al., 2019), aims to develop methods that are able to
discover valuable knowledge from data generated in the educational environment. Data mining
techniques are activities that include collecting and using historical data to find regularities,
patterns and relationships in large data sets. EDM uses a computational approach to analyze
educational data to study questions about education. One form of application of EDM is to predict
student academic performance (Altabrawee et al., 2019; Xiao et al., 2022).
This research aims to determine predictions of students' level of success in graduating based
on students' semester grades as well as knowing the working process of the Decision tree, Neural
Network, Support Vector Machine method in classifying students' level of success in graduating
and it is hoped that there will be early treatment so that it can be predicted which students need
it. Early handling displays data related to learning and predicts student grades in the form of a
dashboard. The benefits of this research make it easier to analyze graduation data so that factors
that influence student graduation rates can be identified, making it easier for schools to make
decisions to help improve the quality of student graduation.
The scope of research in this field is limited by the following: The prediction data used by
researchers was taken from class 12 of SMA Negeri 2 and SMA Negeri 3 in Pangkalpinang city.
This research only concerns the prediction of graduation rates for class 12 students of SMA Negeri
2 and SMA Negeri. 3 cities of Pangkalpinang, all calculation processes provided by the system
use data mining techniques with the Decision tree method, Neural Network, and Support Vector
Machine. The information displayed is in the form of pattern analysis reports and data mining
dashboards for student graduation rates.
Methods
Data Collection
In this stage, the researcher collected the required data, which consisted of the scores of
all subjects on the mid-term exam, final semester exam, and school exams according to each
major. The data taken comes from three different generations. The data mining process requires
training data, which will become a model for testing other data. The data required comes from
alumni data from the classes of 2021 and 2022, which will act as training data, and student data
from the class of 2023, which will act as testing data. The data required also spans the time period
from semester 1 to semester 4. Data such as knowledge and attitude scores will also be collected
so that the problems studied can be explored more deeply. The data source comes from the
school's database.
The methods used for this data collection stage were Field observations at SMA Negeri 2
and SMA Negeri 3 Pangkalpinang city to obtain data directly, make observations, and record the
necessary data; and interviews to obtain deeper information, which will be in the form of research
data attributes and knowledge of student attitudes at SMA Negeri 2 and SMA Negeri 3
Pangkalpinang city. Interviews will be conducted with the school principal or guidance and
counselling teacher from the school.
The data used in this research is data from SMA Negeri 2 Pangkalpinang City and SMA
Negeri 3 Pangkalpinang City in 2021-2023. The attributes used are based on DAPODIK data,
with the number of attributes in accordance with Table 1.
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5482
Table 1. Dataset Attributes.
Attribute
Description
Possible
Value
Final score
Students' final average grade
Numerical
Religious grades for semesters 1 - 4
Grades in religious studies
Numerical
Civics lessons grades for semesters 1 - 4
Grades in Civics lessons
Numerical
Indonesian language grades for semesters 1 - 4
Grades in Indonesian language lessons
Numerical
Mathematics grades for semesters 1 - 4
Grades in mathematics
Numerical
Chemistry grades for semesters 1 - 4
Grades in chemistry class
Numerical
Physics grades for semesters 1 - 4
Grades in Physics
Numerical
Biology grades for semesters 1 - 4
Grades in biology lessons
Numerical
Geography grades for semesters 1 - 4
Grades in geography lessons
Numerical
Economic grades for semesters 1 - 4
Grades in economics lessons
Numerical
History grades for semesters 1 - 4
Grades in history lessons
Numerical
Sociology grades for semesters 1 - 4
Grades in sociology lessons
Numerical
Local content grades for semesters 1 - 4
Grades for local content lessons
Numerical
English grades for semesters 1 - 4
Grades in English lessons
Numerical
Source: Data processed, 2024
Data Preprocessing
Clean data is needed so that the data used for the analysis process meets the researchers'
needs. The Data Preprocessing stages consist of Data Cleaning, Data Integration, Data
Selection/Reduction, and Data Transformation processes. Data Preprocessing aims to form clean
data from raw data obtained from observations and interviews with Pangkalpinang City Public
High Schools. The samples that will be taken in this research can be seen in the following table:
Table 2. Sample Data Taken.
School Name
The class
year 2021
The class
year 2023
Total
SMAN 2 PANGKALPINANG
306
274
875
SMAN 3 PANGKALPINANG
0
151
314
Source: Data processed, 2024
In Table 2, the totals from SMA Negeri 2 Pangkalpinang City and SMA Negeri 3
Pangkalpinang City, the researchers took a total of data samples with a total of 1,189 students.
The data that has been processed is then divided into two data sets, namely training data and
testing data, with a ratio of 80% for training data and 20% for testing data. The purpose of the
training data is to help the algorithmic model to learn patterns from the data that has been
collected. , while the training data is used to evaluate and measure the accuracy of the prediction
results from the previously created model.
Results and Discussion
Preprocessing Data
At this stage, data preprocessing is carried out before the data can be used to avoid data
that makes the model inaccurate. There are several stages of this preprocessing, such as Removing
Missing Values, Removing Duplicate Values, and Sampling.
1. Remove Missing Values
At this stage, the first stage in pre-processing, data cleaning is carried out not only by
selecting attributes appropriate to the subjects of each class of students who graduate with subjects
that will be tested in the final school exam but also by cleaning up any missing power that exists
in training data and testing data.
In the SMA Negeri 2 Pangkalpinang dataset, data is missing in both available datasets.
Missing data occurs due to many possibilities. The possibility of what happens is as follows:
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5483
Table 3. Possible Causes of Missing Data.
Missing Data
Possible Occurrence
There is no data on student grades in semesters 3
and 4
Students change schools when they are in the 11th
grade of high school
There is no data on student grades in semesters 1,
2, 3 and 4
Students change schools when they enter the 10th
grade of high school
There is no data on student grades in semesters 2,
3 and 4
Students move schools when entering the 2nd
semester of grade 10 high school
There is no data on student grades in semesters 1,2
New students enter the school in grade 11 of high
school
Source: Data processed, 2024
Table 3 shows that four possibilities caused missing data in these semesters; the
possibility is that the student in question has gone to another school, or the possibility that the
missing is also a student who just entered the school when the student was in grade 11 of high
school.
Table 4. Number of Missing Data for Each Department.
Missing Data
Data Training
Data Testing
Possible
Occurrence
IPS
IPA
IPS
IPA
Semester 1
11 Record
9 Record
0 Record
0 Record
Students continue at
the school
Semester 1 and
2
13 Record
14 Record
0 Record
0 Record
Students change
schools when they
enter the 11th grade
of high school
Semester 1, 2,
and 3
8 Record
12 Record
0 Record
0 Record
New transfer student
Semesters 1, 2,
3, and 4
10 Record
15 Record
0 Record
0 Record
New transfer student
entering 12th grade
of high school
Source: Data processed, 2024
Table. 4 provides the amount of missing data obtained in the training data and data set in
each scenario, and also for each department and the possible causes of the missing data. In the
science department training data for the first scenario for semester 1, there are 11 missing data;
for the second scenario for semesters 1 and 2, there are 13 missing data; for the third scenario for
semesters 1, 2 and 3, there are eight records missing data, and for the fourth-semester scenario 1,
2, 3, and 4 contain missing data of 10 records.
Meanwhile, in the IPS department training data for the first scenario for semester 1, there
were nine records missing data; for the second scenario for semesters 1 and 2, there were 14
missing data; for the third scenario for semesters 1, 2 and 3 there were 12 records missing data,
and for the fourth scenario Semesters 1, 2, 3, and 4 contained 15 missing data. There were no
missing data in testing data in any scenario for either science or social studies at SMAN Negeri 2
Pangkalpinang; for testing data and training data at SMA Negeri 3 Pangkalpinang from scenarios
1 to scenario 4, there were no missing data in the science and social studies majors. Missing
records can disrupt the data mining process later, so the names of students who have missing
records must be cleaned or removed from the dataset table.
Data mining
At this stage, data mining processing is carried out, which aims to find and build a model
from a dataset and then use the model in another dataset where you want to predict a result. In
this research, data from students from the class of 2021, 2022 from semester 1 to semester 4 of
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5484
SMAN 2 and data from students from the class of 2022 from semester 1 to semester 4 from SMAN
3 Pangkal Pinang will act as training data, which will produce a classification model (Courtiol et
al., 2019). Later, this model will be used in grade data from the class of 2023 at SMAN 2 and
SMAN 3 Pangkalpinang as testing data. In this research, 4 test scenarios were carried out in order
to determine scenarios that accurately determine predictions of students' final grades using three
models, namely Decision Tree (DT), Support Vector Machine (SVM), and Neural Network (NN).
1. Decision Models tree
In this research, the Decision model tree is used to predict student grades from semester
1 to semester 4 by using two datasets, namely dataset training and dataset testing, to find out
whether the Decision model This tree has accurate predictions to predict the final grades of
students at SMAN 2 and SMAN 3 Pangkalpinang. This model is used to determine the scenario
that has the most accurate final grade predictions so that conclusions can be drawn to find out
which students need to do extra learning in order to graduate. Final exam (Mienye et al., 2019).
a. MSE Model Decision Value tree Scenario 1 Science Department of SMAN 2 and SMAN
3 Pangkalpinang
The first scenario where the value to be predicted is the average value from semester 1
with testing data, namely the class of 2023. The MSE evaluation value obtained in scenario 1 is
4.959515007849291, with calculated differences and percentage differences, the difference
between the actual and predicted values. The MSE value is obtained from training data and testing
data; the actual value is obtained from testing data; the predicted value is obtained from the results
of MSE management; the calculated value The difference is obtained from the actual value results
with predicted value, calculate percentage The difference is obtained from the actual value with
the Difference value. Actual, predicted, difference and percentage values differences can be seen
in Table .
Table 5. Scenario 1 Science Value actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
85.93
85.71
0.22
0.25
82.86
85.93
-3.07
-3.70
84.14
83.71
0.43
0.51
79.36
83.50
-4.14
-5.21
81.00
84.57
-3.57
-4.40
80.57
80.00
0.57
0.70
83.00
81.79
1.21
1.45
79.42
79.57
-0.14
-0.17
82.00
82.14
-0.14
-0.17
80.93
84.71
-3.78
-4.67
Source: Data processed, 2024
b. MSE Model Decision Value tree Scenario 2 Science Department of SMAN 2 and SMAN
3 Pangkalpinang
The second scenario where the value to be predicted is the average value from the
combination of semester 1 and semester 2 with testing data, namely the class of 2023. The MSE
evaluation value obtained in scenario 2 is 5.905944629014399, with calculated differences and
percentage differences, the difference between the actual and predicted values. The MSE value is
obtained from training data and testing data; the actual value is obtained from testing data; the
predicted value is obtained from the results of MSE management; the calculated value The
difference is obtained from the actual value results with predicted value, calculate percentage The
difference is obtained from the actual value with the Difference value. Actual, predicted,
difference and percentage value differences can be seen in Table 6.
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5485
Table 6 Scenario 2 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
81.36
83.86
-2.50
-3.07
82.71
79.86
2.85
3.44
82.57
88.71
-6.14
-7.43
80.21
76.86
3.35
4.18
85.79
84.86
0.93
1.08
82.50
80.79
1.71
2,072
82.43
82.50
-0.07
-0.08
83.00
84.50
-1.50
-1.80
82.29
84.14
-1.85
-2.24
81.43
82.29
-0.86
-1.05
Source: Data processed, 2024
c. MSE Model Decision Value tree Scenario 3 Science Department of SMAN 2 and SMAN
3 Pangkalpinang
The third scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, and semester 3 with testing data, namely the class of 2023.
The MSE evaluation value obtained in scenario 3 is 4.689620471281302 with calculated
differences And calculated percentage difference, the difference between the actual and predicted
values. The MSE value is obtained from training data and testing data; the actual value is obtained
from testing data; the predicted value is obtained from the results of MSE management; the
calculated value The difference is obtained from the actual value results with predicted value,
calculate percentage The difference is obtained from the actual value with the Difference value.
Actual, predicted, difference and percentage values differences can be seen in Table 7
Table 7. Scenario 3 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
82.71
79.71
2.99
3.62
82.21
82.07
0.14
0.17
83.36
80.71
2.65
3.17
81.43
83.71
-2.28
-2.79
85.86
85.29
0.57
0.66
82.00
84.35
-2.35
-2.87
84.64
82.43
2,210
2.61
82.00
81.36
0.64
0.78
73.64
76.93
-3.29
-4.46
83.71
80.57
3.14
3.75
Source: Data processed, 2024
d. MSE Model Decision Value tree Scenario 4 Science Department of SMAN 2 and
SMAN 3 Pangkalpinang
The fourth scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, semester 3, and semester 4 with testing data, namely the
class of 2023. The MSE evaluation value obtained in scenario 4 is 5.013858037676613 with
calculate differences And calculate percentage difference, the difference between the actual and
predicted values. The MSE value is obtained from training data and testing data; the actual value
is obtained from testing data; the predicted value is obtained from the results of MSE
management; the calculated value The difference is obtained from the actual value results with
predicted value, calculate percentage The difference is obtained from the actual value with the
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5486
Difference value. Actual, predicted, difference and percentage values differences can be seen in
Table 8
Table 8. Scenario 4 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
81.21
80.93
0.28
0.34
82.21
81.21
1.00
1.21
80.29
81.50
-1.21
-1.50
81.64
82.21
-0.57
-0.69
81.00
78.92
2.07
2.55
82.50
85.79
-3.29
-3.98
81.14
82.43
-1.29
-1.58
82.64
83.71
-1.07
-1.29
86.36
88.71
-2.35
-2.72
85.79
85.93
-0.14
-0.16
Source: Data processed, 2024
e. MSE Model Decision Value tree Scenario 1 Social Sciences Department of SMAN 2 and
SMAN 3 Pangkalpinang
The first scenario where the value to be predicted is the average value from semester 1
with testing data, namely the class of 2023. The MSE evaluation value obtained in scenario 1 is
8.531357587650442 with calculate differences And calculate percentage difference, the
difference between the actual and predicted values. The MSE value is obtained from training data
and testing data; the actual value is obtained from testing data; the predicted value is obtained
from the results of MSE management; the calculated value The difference is obtained from the
actual value results with predicted value, calculate percentage The difference is obtained from the
actual value with the Difference value. Actual, predicted, difference and percentage values
differences can be seen in Table 9
Table 9. Scenario 1 Social Studies Score actual, predicted, difference, and
percentage differences
Actual
Predicted
Difference
Percentage
Difference
75.86
78.86
-3.00
-3.95
81.64
83.00
-1.36
-1.66
81.21
84.64
-3.43
-4.22
80.93
79.00
1.93
2.38
83.07
77.50
5.57
6.70
82.07
81.29
0.78
0.95
80.07
79.86
0.21
0.26
80.0
78.07
1.93
2.41
79.50
74.71
4.79
6.02
82.07
82.79
-0.72
-0.87
Source: Data processed, 2024
f. MSE Model Decision Value tree Scenario 2 Social Sciences Department of SMAN 2 and
SMAN 3 Pangkalpinang
The second scenario where the value to be predicted is the average value from the
combination of semester 1 and semester 2 with testing data, namely the class of 2023. The MSE
evaluation value obtained in scenario 2 is 7.767002838827839 with calculate differences And
calculate percentage difference, the difference between the actual and predicted values. The MSE
value is obtained from training data and testing data; the actual value is obtained from testing
data; the predicted value is obtained from the results of MSE management; the calculated value
The difference is obtained from the actual value results with predicted value, calculate percentage
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5487
The difference is obtained from the actual value with the Difference value. Actual, predicted,
difference and percentage values differences can be seen in Table 10
Table 10. Scenario 2 IPS scores actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
76.36
82.64
-6.28
-8.22
78.21
77.00
1.21
1.54
75.86
77.71
-1.85
-2.43
81.79
78.79
3.00
3.66
78.07
77.85
0.21
0.27
75.57
75.07
0.50
0.66
79.71
78.57
1.14
1.43
81.93
79.00
2.93
3.57
76.57
82.00
-5.43
-7.09
83.50
79.21
4.29
5.13
Source: Data processed, 2024
g. MSE Model Decision Value tree Scenario 3 Social Sciences Department of SMAN 2
and SMAN 3 Pangkalpinang
The third scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, and semester 3 with testing data, namely the class of 2023.
The MSE evaluation value obtained in scenario 3 is 6.318856035234607 with calculate
differences And calculate percentage difference, the difference between the actual and predicted
values. The MSE value is obtained from training data and testing data; the actual value is obtained
from testing data; the predicted value is obtained from the results of MSE management; the
calculated value The difference is obtained from the actual value results with predicted value,
calculate percentage The difference is obtained from the actual value with the Difference value.
Actual, predicted, difference and percentage value differences can be seen in Table 11.
Table 11. Scenario 3 IPS Score actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
78.43
79.21
-0.78
-0.99
79.79
74.71
5.08
6.36
74.71
78.07
-3.36
-4.49
77.79
77.57
0.22
0.28
82.43
81.86
0.57
0.69
78.29
80.00
-1.71
-2.18
84.14
85.21
-1.07
-1.27
78.79
79.21
-0.42
-0.53
79.50
77.00
2.50
3.14
83.07
85.79
-2.72
-3.27
Source: Data processed, 2024
h. MSE Model Decision Value tree Scenario 4 Social Sciences Department of SMAN 2
and SMAN 3 Pangkalpinang
The fourth scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, semester 3, and semester 4 with testing data, namely the
class of 2023. The MSE evaluation value obtained in scenario 4 is 5.337855004238116 with
calculate differences And calculate percentage difference, the difference between the actual and
predicted values. The MSE value is obtained from training data and testing data; the actual value
is obtained from testing data; the predicted value is obtained from the results of MSE
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5488
management; the calculated value The difference is obtained from the actual value results with
predicted value, calculate percentage The difference is obtained from the actual value with the
Difference value. Actual, predicted, difference and percentage values differences can be seen in
Table 12.
Table 12. Scenario 4 IPS Score actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
78.21
79.21
-0.99
-1.27
82.07
81.86
0.21
0.25
78.50
76.93
1.57
2.00
76.57
77.64
-1.07
-1.39
83.07
81.86
1.21
1.45
76.35
74.50
1,857
2.43
78.29
81.43
-3.14
-4.01
78.50
77.71
0.78
1.00
77.07
77.93
-0.85
-1.11
76.79
79.71
-2.92
-3.80
Source: Data processed, 2024
2. Support Models Vector Machine
In this research, the Support model was also used Vector Machine was used to predict
student grades from semester 1 to semester 4 by having two datasets, namely dataset training and
dataset testing, to find out whether the model supports Vector. This machine has accurate
predictions to predict the final grades of students at SMAN 2 and SMAN 3 Pangkalpinang; this
model is used to determine the scenario that has the most accurate final grade predictions so that
conclusions can be drawn to find out which students need to do extra learning in order to graduate
in Final exam (Dou et al., 2020).
a. MSE Model Support Value Vector Machine Scenario 1 Science Department of SMAN
2 and SMAN 3 Pangkalpinang
The first scenario where the value to be predicted is the average value from semester 1
with testing data, namely the class of 2023. The MSE evaluation value obtained in scenario 1 is
41.21334272134476 with calculate differences And calculate percentage difference, the
difference between the actual and predicted values. The MSE value is obtained from training data
and testing data; the actual value is obtained from testing data; the predicted value is obtained
from the results of MSE management; the calculated value The difference is obtained from the
actual value results with predicted value, calculate percentage The difference is obtained from the
actual value with the Difference value. Actual, predicted, difference and percentage value
differences can be seen in Table 13.
Table 13. Scenario 1 Science Value actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
85.93
85.85
0.07
0.08
82.86
82.34
0.511
0.61
84.14
82.60
1.53
1.81
79.36
82.13
-2.77
-3.50
81.00
81.14
-0.14
-0.17
80.57
80.3
0.25
0.32
83.00
81.18
1.81
2.19
79.42
80.10
-0.67
-0.84
82.00
81.15
0.84
1.02
80.93
81.73
-0.80
-0.99
Source: Data processed, 2024
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5489
b. MSE Model Support Value Vector Machine Scenario 2 Science Department of SMAN
2 and SMAN 3 Pangkalpinang
The second scenario where the value to be predicted is the average value from the
combination of semester 1 and semester 2 with testing data, namely the class of 2023. The MSE
evaluation value obtained in scenario 2 is 3.494705329823478 with calculate differences And
calculate percentage difference, the difference between the actual and predicted values. The MSE
value is obtained from training data and testing data; the actual value is obtained from testing
data; the predicted value is obtained from the results of MSE management; the calculated value
The difference is obtained from the actual value results with predicted value, calculate percentage
The difference is obtained from the actual value with the Difference value. Actual, predicted,
difference and percentage value differences can be seen in Table 14.
Table 14. Scenario 2 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
81.36
82.06
-0.70
-0.86
82.71
81.27
1.43
1.73
82.57
83.80
-1.23
-1.49
80.21
78.69
1.52
1.89
85.79
84.7
1.01
1.18
82.50
82.21
0.28
0.33
82.43
81.25
1.17
1.42
83.00
82.89
0.10
0.12
82.29
81.8 4
0.44
0.53
81.43
80.97
0.45
0.56
Source: Data processed, 2024
c. MSE Model Support Value Vector Machine Scenario 3 Science Department of SMAN
2 and SMAN 3 Pangkalpinang
The third scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, and semester 3 with testing data, namely the class of 2023.
The MSE evaluation value obtained in scenario 3 is 14.182750048882673 with calculate
differences And calculate percentage difference, the difference between the actual and predicted
values. The MSE value is obtained from training data and testing data; the actual value is obtained
from testing data; the predicted value is obtained from the results of MSE management; the
calculated value The difference is obtained from the actual value results with predicted value,
calculate percentage The difference is obtained from the actual value with the Difference value.
Actual, predicted, difference and percentage values differences can be seen in Table 15
Table 15. Scenario 3 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
82.71
81.02
1.68
2.03
82.21
80.88
1.32
1.61
83.36
82.12
1.23
1.48
81.43
81.36
0.06
0.08
85.86
81.20
4.65
5.42
82.00
78.96
3.03
3.70
84.64
83.70
0.93
1.09
82.00
80.77
1.22
1.49
73.64
80.57
-6.93
-9.41
83.71
81.26
2.44
2.92
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5490
Source: Data processed, 2024
d. MSE Model Support Value Vector Machine Scenario 4 Science Department of SMAN
2 and SMAN 3 Pangkalpinang
The fourth scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, semester 3, and semester 4 with testing data, namely the
class of 2023. The MSE evaluation value obtained in scenario 4 is 3.029739424194293 with
calculate differences And calculate percentage difference, the difference between the actual and
predicted values. The MSE value is obtained from training data and testing data; the actual value
is obtained from testing data; the predicted value is obtained from the results of MSE
management; the calculated value The difference is obtained from the actual value results with
predicted value, calculate percentage The difference is obtained from the actual value with the
Difference value. Actual, predicted, difference and percentage values differences can be seen in
Table 16
Table 16. Scenario 4 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
81.21
83.33
-2.12
-2.61
82.21
80.68
1.52
1.85
80.29
80.05
0.23
0.28
81.64
82.74
-1.10
-1.35
81.00
79.85
1.14
1.41
82.50
81.70
0.79
0.96
81.14
81.57
-0.43
-0.53
82.64
81.70
0.93
1.13
86.36
86.81
-0.45
-0.52
85.79
86.45
-0.66
-0.77
Source: Data processed, 2024
e. MSE Model Support Value Vector Machine Scenario 1 Social Sciences Department,
SMAN 2 and SMAN 3 Pangkalpinang
The first scenario where the value to be predicted is the average value from semester 1
with testing data, namely the class of 2023. The MSE evaluation value obtained in scenario 1 is
5.130493760927023 with calculate differences And calculate percentage difference, the
difference between the actual and predicted values. The MSE value is obtained from training data
and testing data; the actual value is obtained from testing data; the predicted value is obtained
from the results of MSE management; the calculated value The difference is obtained from the
actual value results with predicted value, calculate percentage The difference is obtained from the
actual value with the Difference value. Actual, predicted, difference and percentage value
differences can be seen in Table 17.
Table 17. Scenario 1 Social Studies Score actual, predicted, difference, and
percentage differences
Actual
Predicted
Difference
Percentage
Difference
75.86
78.19
-2.33
-3.07
81.64
79.19
2.44
2.99
81.21
79.11
2.09
2.58
80.93
79.59
1.33
1.64
83.07
77.48
5.58
6.72
82.07
80.20
1.86
2.27
80.07
80.32
-0.25
-0.32
80.00
79.87
0.12
0.15
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5491
79.50
80.54
-1.04
-1.31
82.07
81.63
0.43
0.53
Source: Data processed, 2024
f. MSE Model Support Value Vector Machine Scenario 2 Social Sciences Department,
SMAN 2 and SMAN 3 Pangkalpinang
The second scenario where the value to be predicted is the average value from the
combination of semester 1 and semester 2 with testing data, namely the class of 2023. The MSE
evaluation value obtained in scenario 2 is 5.1600334873366345 with calculate differences And
calculate percentage difference, the difference between the actual and predicted values. The MSE
value is obtained from training data and testing data; the actual value is obtained from testing
data; the predicted value is obtained from the results of MSE management; the calculated value
The difference is obtained from the actual value results with predicted value, calculate percentage
The difference is obtained from the actual value with the Difference value. Actual, predicted,
difference and percentage values differences can be seen in Table 18
Table 18. Scenario 2 IPS scores actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
76.36
79.68
-3.32
-4.35
78.21
77.76
0.44
0.57
75.86
78.94
-3.08
-4.0 6
81.79
78.54
3.24
3.96
78.07
79.44
-1,379
-1.76
75.57
78.30
-2,733
-3.61
79.71
77.82
1.88
2.36
81.93
79.9 9
1.93
2.35
76.57
82.28
-5.71
-7.46
83.50
81.40
2,098
2.51
Source: Data processed, 2024
g. MSE Model Support Value Vector Machine Scenario 3 Social Sciences Department of
SMAN 2 and SMAN 3 Pangkalpinang
The third scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, and semester 3 with testing data, namely the class of 2023.
The MSE evaluation value obtained in scenario 3 is 4.220445766997599 with calculate
differences And calculate percentage difference, the difference between the actual and predicted
values. The MSE value is obtained from training data and testing data; the actual value is obtained
from testing data; the predicted value is obtained from the results of MSE management; the
calculated value The difference is obtained from the actual value results with predicted value,
calculate percentage The difference is obtained from the actual value with the Difference value.
Actual, predicted , difference and percentage values differences can be seen in Table 19
Table 19. Scenario 3 IPS Score actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
78.43
80.16
-1.73
-2.21
79.79
79.11
0.677
0.84
74.71
77.69
-2.98
-4.00
77.79
77.53
0.255
0.32
82.43
82.11
0.31
0.38
78.29
80.10
-1.81
-2.31
84.14
82.93
1.20
1.42
78.79
79.96
-1.17
-1.48
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5492
79.50
79.67
-0.17
-0.21
83.07
83.36
-0.29
-0.35
Source: Data processed, 2024
h. MSE Model Support Value Vector Machine Scenario 4 Social Sciences Department of
SMAN 2 and SMAN 3 Pangkalpinang
The fourth scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, semester 3, and semester 4 with testing data, namely the
class of 2023. The MSE evaluation value obtained in scenario 4 is 4.665472066573534 with
calculate differences And calculate percentage difference, the difference between the actual and
predicted values. The MSE value is obtained from training data and testing data; the actual value
is obtained from testing data, the predicted value is obtained from the results of MSE
management, the calculated value The difference is obtained from the actual value results with
predicted value, calculate percentage The difference is obtained from the actual value with the
Difference value. Actual, predicted, difference and percentage values differences can be seen in
Table 20
Table 20. Scenario 4 IPS Score actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
78.21
78.70
-0.48
-0.62
82.07
81.64
0.42
0.51
78.50
79.81
-1.31
-1.68
76.57
79.09
-2.52
-3.29
83,070
84.36
-1.29
-1.55
76.35
77.70
-1.34
-1.75
78.29
79.90
-1.61
-2.06
78.50
79.49
-0.99
-1.27
77.07
78.50
-1.42
-1.85
76.79
78.93
-2.14
-2.79
Source: Data processed, 2024
3. Neural Network Models
In this research, a Neural Network model was also used to predict student grades from
semester 1 to semester 4 by having two datasets, namely the dataset training and dataset testing,
to find out whether this Neural Network model has accurate predictions to predict the final grades
of students at SMAN 2 and SMAN 3 Pangkalpinang, this model is used to determine the scenario
that has the most accurate final grade predictions so that conclusions can be drawn to find out
what students should do extra learning in order to pass the Final Exam (Sharifzadeh et al., 2019).
a. MSE Value of Neural Network Model Scenario 1 Science Department at SMAN 2 and
SMAN 3 Pangkalpinang
The first scenario where the value to be predicted is the average value from semester 1
with testing data, namely the class of 2023. The MSE evaluation value obtained in scenario 1 is
24660.441532902347 with calculate differences And calculate percentage difference, the
difference between the actual and predicted values. The MSE value is obtained from training data
and testing data; the actual value is obtained from testing data; the predicted value is obtained
from the results of MSE management; the calculated value The difference is obtained from the
actual value results with predicted value, calculate percentage The difference is obtained from the
actual value with the Difference value. Actual, predicted , difference and percentage values
differences can be seen in Table 21.
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5493
Table 21. Scenario 1 Science Value actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
85.93
106.71
-20.78
-24.19
82.86
93.09
-10.23
-12.35
84.14
61.40
22.73
27.01
79.36
64.17
15.18
19.13
81.00
62.72
18.27
22.55
80.57
71.39
9.17
11.38
83.00
84.78
-1.78
-2.15
79.42
98.70
-19.27
-24.26
82.0
74.01
7.98
9.73
80.93
80.99
-0.06
-0.076
Source: Data processed, 2024
b. MSE Value of Neural Network Model Scenario 2 Science Department at SMAN 2 and SMAN
3 Pangkalpinang
The second scenario where the value to be predicted is the average value from the
combination of semester 1 and semester 2 with testing data, namely the class of 2023. The MSE
evaluation value obtained in scenario 2 is 55.90357441058514 with calculate differences And
calculate percentage difference, the difference between the actual and predicted values. The MSE
value is obtained from training data and testing data; the actual value is obtained from testing
data; the predicted value is obtained from the results of MSE management; the calculated value
The difference is obtained from the actual value results with predicted value, calculate percentage
The difference is obtained from the actual value with the Difference value. Actual, predicted,
difference and percentage values differences can be seen in Table 4.22
Table 22. Scenario 2 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
81.36
75.85
5.50
6.76
82.71
78.36
4.34
5.24
82.57
80.41
2.15
2.60
80.21
74.18
6.03
7.51
85.79
81.13
4.65
5.42
82.50
77.38
5.11
6.19
82.43
72.27
10.15
12.31
83.00
89.92
-6.92
-8.33
82.29
82.84
-0.55
-0.67
81.43
86.50
-5.07
-6.22
Source: Data processed, 2024
c. MSE Value of Neural Network Model Scenario 3 Science Department at SMAN 2 and
SMAN 3 Pangkalpinang
The third scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, and semester 3 with testing data, namely the class of 2023.
The MSE evaluation value obtained in scenario 3 is 824.1271515136524 with calculate
differences And calculate percentage difference, the difference between the actual and predicted
values. The MSE value is obtained from training data and testing data; the actual value is obtained
from testing data; the predicted value is obtained from the results of MSE management; the
calculated value The difference is obtained from the actual value results with predicted value,
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5494
calculate percentage The difference is obtained from the actual value with the Difference value.
Actual, predicted, difference and percentage values differences can be seen in Table 23
Table 23. Scenario 3 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
82.71
90.32
-7.61
-9.21
82.21
83.09
-0.88
-1.07
83.36
82.87
0.48
0.58
81.43
79.09
2.33
2.87
85.86
78.97
6.88
8.01
82.00
85.53
-3.53
-4.31
84.64
83.17
1.46
1.73
82.00
80.92
1,073
1.30
73.64
86.27
-12.63
-17.15
83.71
78.86
4.84
5.79
Source: Data processed, 2024
d. MSE Value of Neural Network Model Scenario 4 Science Department at SMAN 2 and
SMAN 3 Pangkalpinang
The fourth scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, semester 3, and semester 4 with testing data, namely the
class of 2023. The MSE evaluation value obtained in scenario 4 is 27.31309375459551 with
calculate differences And calculate percentage difference, the difference between the actual and
predicted values. The MSE value is obtained from training data and testing data; the actual value
is obtained from testing data; the predicted value is obtained from the results of MSE
management; the calculated value The difference is obtained from the actual value results with
predicted value, calculate percentage The difference is obtained from the actual value with the
Difference value. Actual, predicted, difference and percentage values differences can be seen in
Table 24
Table 24. Scenario 4 Science Values actual, predicted, difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
81.21
83.33
-2.12
-2.61
82.21
80.68
1.52
1.85
80.29
80.05
0.23
0.28
81.64
82.74
-1.10
-1.35
81.00
79.85
1.14
1.41
82.50
81.70
0.79
0.96
81.14
81.57
-0.43
-0.53
82.64
81.70
0.93
1.13
86.36
86.81
-0.45
-0.523
85.79
86.45
-0.66
-0.77
Source: Data processed, 2024
e. MSE Value of Neural Network Model Scenario 1 Social Sciences Department at SMAN
2 and SMAN 3 Pangkalpinang
The first scenario where the value to be predicted is the average value from semester 1
with testing data, namely the class of 2023. The MSE evaluation value obtained in scenario 1 is
277.72135373958025 with calculate differences And calculate percentage difference, the
difference between the actual and predicted values. The MSE value is obtained from training data
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5495
and testing data; the actual value is obtained from testing data; the predicted value is obtained
from the results of MSE management; the calculated value The difference is obtained from the
actual value results with predicted value, calculate percentage The difference is obtained from the
actual value with the Difference value, Actual, predicted , difference and percentage values
differences can be seen in Table 25
Table 25. Scenario 1 Social Studies Score actual, predicted , difference, and percentage
differences
Actual
Predicted
Difference
Percentage
Difference
75.86
58.44
17.41
22.96
81.64
80.15
1.48
1.82
81.21
71.77
9.43
11.61
80.93
66.02
14.90
18.41
83.07
75.11
7.95
9.57
82.07
68.16
13.9 0
16.94
80.07
71.10
8.96
11.19
80.00
80.22
-0.22
-0.27
79.50
90.93
-11.43
-14.3
82.07
70.88
11.18
13.62
Source: Data processed, 2024
f. MSE Value of Neural Network Model Scenario 2 Social Sciences Department of SMAN
2 and SMAN 3 Pangkalpinang
The second scenario where the value to be predicted is the average value from the
combination of semester 1 and semester 2 with testing data, namely the class of 2023. The MSE
evaluation value obtained in scenario 2 is 76.35980830234871 with calculate differences And
calculate percentage difference, the difference between the actual and predicted values . The MSE
value is obtained from training data and testing data, the actual value is obtained from testing data
, the predicted value is obtained from the results of MSE management, the calculated value The
difference is obtained from the actual value results with predicted value , calculate percentage
The difference is obtained from the actual value with the Difference value . Actual , predicted ,
difference and percentage values differences can be seen in Table 26
Table 26. Scenario 2 IPS scores actual, predicted, difference, and percentage differences
Actual
Predicted
Difference
Percentage
Difference
76.36
73.20
3.15
4.12
78.21
77.19
1.01
1.29
75.86
82.54
-6.68
-8.81
81.79
72.97
8.81
10.77
78.07
77.01
1.05
1.34
75.57
78.49
-2.92
-3.87
79.71
78.62
1,089
1.36
81.93
79.83
2.09
2.55
76.57
78.39
-1.82
-2.38
83.50
78.10
5.39
6.45
Source: Data processed, 2024
g. MSE Value of Neural Network Model Scenario 3 Social Sciences Department of SMAN
2 and SMAN 3 Pangkalpinang
The third scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, and semester 3 with testing data, namely the class of 2023.
The MSE evaluation value obtained in scenario 3 is 122.88905701875385 with calculate
differences And calculate percentage difference, the difference between the actual and predicted
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5496
values. The MSE value is obtained from training data and testing data; the actual value is obtained
from testing data, the predicted value is obtained from the results of MSE management, the
calculated value The difference is obtained from the actual value results with predicted value,
calculate percentage The difference is obtained from the actual value with the Difference value.
Actual, predicted, difference and percentage values differences can be seen in Table 27.
Table 27. Scenario 3 IPS Score actual, predicted , difference, and percentage differences
Actual
Predicted
Difference
Percentage
Difference
78.43
75.01
3.41
4.35
79.79
77.95
1.83
2.30
74.71
81.64
-6.93
-9.27
77.79
76.13
1.65
2.12
82.43
80.75
1.67
2.02
78.29
79.10
-0.81
-1.04
84.14
84.87
-0.73
-0.86
78.79
78.98
-0.19
-0.25
79.50
80.69
-1.19
-1.50
83.07
92.76
-9.69
-11.66
Source: Data processed, 2024
h. MSE Value of Neural Network Model Scenario 4 Social Sciences Department of SMAN
2 and SMAN 3 Pangkalpinang
The fourth scenario where the value to be predicted is the average value from the
combination of semester 1, semester 2, semester 3, and semester 4 with testing data, namely the
class of 2023. The MSE evaluation value obtained in scenario 4 is 15.625493356140588 with
calculate differences And calculate percentage difference, the difference between the actual and
predicted values. The MSE value is obtained from training data and testing data; the actual value
is obtained from testing data, the predicted value is obtained from the results of MSE
management, the calculated value The difference is obtained from the actual value results with
predicted value, calculate percentage The difference is obtained from the actual value with the
Difference value. Actual, predicted, difference and percentage values differences can be seen in
Table 28.
Table 28. Scenario 4 IPS Score actual, predicted, difference, and percentage differences
Actual
Predicted
Difference
Percentage
Difference
78.21
78.70
-0.48
-0.62
82.07
81.64
0.42
0.51
78.50
79.81
-1.31
-1.68
76.57
79.09
-2.52
-3.29
83.07
84.36
-1.29
-1.55
76.35
77.70
-1.34
-1.75
78.29
79.90
-1.61
-2.06
78.50
79.49
-0.99
-1.27
77.07
78.50
-1.42
-1.85
76.79
78.93
-2.14
-2.79
Source: Data processed, 2024
Model Evaluation
The evaluation model used is Mean Squared Error (MSE) with objectives For each
prediction made by the model (Chicco et al., 2021), MSE measures the squared difference
between the predicted value and the actual value; MSE also provides information about how
accurate the model is in predicting the data, but can also be used as a tool to improve model
performance and understand the characteristics of the data being processed. Once calculated for
each prediction, MSE then takes the average of these squared differences to provide an overall
picture of the model's performance. The scenarios tested were as follows: Scenario 1, namely
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5497
predicting the first semester grades of students at SMAN 2 and SMAN 3 Pangkalpinang; Scenario
2 is predicting the combined 1st and 2nd-semester grades of students at SMAN 2 and SMAN 3
Pangkalpinang; Scenario 3 is predicting the combined 1st, 2nd and 3rd-semester grades of
students at SMAN 2 and SMAN 3 Pangkalpinang; Scenario 4 is predicting the combined grades
of semesters 1, 2, 3, and 4 of students at SMAN 2 and SMAN 3 Pangkalpinang
The MSE value in each scenario is different because each scenario has different data. In
determining the MSE value of the data set, extracting features and targets from the variables is
carried out so that the training dataset is also tested using regression with the model being tested,
namely the decision model. tree (DT), Support Vector Machine (SVM), and Neural Network
(NN). The following are the results of the MSE values from scenarios 1 to 4 for the science and
social studies majors in the decision model tree (DT), Support Vector Machine (SVM), and
Neural Network (NN)
In scenario 1, the science major for the lowest MSE value is the decision model tree MSE
results show that the test results are in accordance with actual data and can be used for test
calculations in the future period. The results of scenario 1 of the science majors at SMAN 2 and
SMAN 3 Pangkalpinang can be seen in Table 29 below:
Table 29. MSE scores for Scenario 1 Science at SMAN 2 and SMAN 3 Pangkalpinang
Model
MSE value
Decision tree
4.959515007849291
Support Vector
Machine
41.21334272134476
Neural Networks
24660.441532902347
Source: Data processed, 2024
In scenario 2, the science major for the lowest MSE value is the Support model Vector
Machine, and the MSE results show that the test results are in accordance with the actual data and
can be used for test calculations in the future period. The results of scenario 2 science majors at
SMAN 2 and SMAN 3 Pangkalpinang can be seen in Table 30 below:
Table 30. MSE scores for Scenario 2 Science at SMAN 2 and SMAN 3 Pangkalpinang
Model
MSE value
Decision tree
5.905944629014399
Support Vector
Machine
3.494705329823478
Neural Networks
55.90357441058514
Source: Data processed, 2024
In scenario 3, the science major for the lowest MSE value is the decision model tree, and
the MSE results show that the test results are in accordance with the actual data and can be used
for test calculations in the future period. The results of scenarios for 3 science majors at SMAN
2 and SMAN 3 Pangkalpinang can be seen in Table 31 below:
Table 31. MSE scores for Scenario 3 Science at SMAN 2 and SMAN 3 Pangkalpinang
Model
MSE value
Decision tree
4.689620471281302
Support Vector
Machine
14.182750048882673
Neural Networks
824.1271515136524
Source: Data processed, 2024
In scenario 4, the science major for the lowest MSE value is the Support model Vector
Machine, and the MSE results show that the test results are in accordance with the actual data and
can be used for test calculations in the future period. The results of scenarios for 4 science majors
at SMAN 2 and SMAN 3 Pangkalpinang can be seen in Table 32 below:
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5498
Table 32. MSE scores for Scenario 4 Science at SMAN 2 and SMAN 3 Pangkalpinang
Model
MSE value
Decision tree
5.013858037676613
Support Vector
Machine
3.029739424194293
Neural Networks
27.31309375459551
Source: Data processed, 2024
Of the four scenarios for the science major at SMAN 2 and SMAN 3 Pangkalpinang with
three different models, the MSE value shows that the test results are in accordance with the testing
dataset and can be used to predict the student's final score, namely the decision model, tree, and
support vector machine.
In scenario 1, the Social Sciences major for the lowest MSE value is the Support model
Vector Machine. MSE results show that the test results are in accordance with actual data and can
be used for test calculations in future periods. The results of scenario 1 of the Social Sciences
Department at SMAN 2 and SMAN 3 Pangkalpinang can be seen in Table 33 below:
Table 33. MSE scores for Scenario 1 IPS SMAN 2 and SMAN 3 Pangkalpinang
Model
MSE value
Decision tree
8.531357587650442
Support Vector
Machine
5.130493760927023
Neural Networks
277.72135373958025
Source: Data processed, 2024
In scenario 2 of the Social Sciences major, the lowest MSE value is the Support model
Vector Machine, and the MSE results show that the test results are in accordance with the actual
data and can be used for test calculations in the future period. The results of scenario 2 of the
Social Sciences majors at SMAN 2 and SMAN 3 Pangkalpinang can be seen in Table 34 below:
Table 34. MSE scores for Scenario 2 IPS SMAN 2 and SMAN 3 Pangkalpinang
Model
MSE value
Decision tree
7.767002838827839
Support Vector
Machine
5.1600334873366345
Neural Networks
76.35980830234871
Source: Data processed, 2024
In scenario 3 of the Social Sciences major, the lowest MSE value is the Support model
Vector Machine , that the MSE results show that the test results are in accordance with the actual
data and can be used for test calculations in the future period. The results of scenarios for 3 social
studies majors at SMAN 2 and SMAN 3 Pangkalpinang can be seen in Table 35 below:
Table 35. MSE scores for Scenario 3 IPS SMAN 2 and SMAN 3 Pangkalpinang
Model
MSE value
Decision tree
6.318856035234607
Support Vector
Machine
4.220445766997599
Neural Networks
122.88905701875385
Source: Data processed, 2024
In scenario 4 of the Social Sciences major, the lowest MSE value is the Support model
Vector Machine, that the MSE results show that the test results are in accordance with the actual
data and can be used for test calculations in the future period. The results of scenarios for 4 social
studies majors at SMAN 2 and SMAN 3 Pangkalpinang can be seen in Table 36 below:
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5499
Table 36. MSE scores for Scenario 4 IPS SMAN 2 and SMAN 3 Pangkalpinang
Model
MSE value
Decision tree
5.337855004238116
Support Vector
Machine
4.665472066573534
Neural Networks
15.625493356140588
Source: Data processed, 2024
Of the four scenarios for the Social Sciences major at SMAN 2 and SMAN 3
Pangkalpinang with three different models, the MSE value shows that the test results are in
accordance with the testing dataset and can be used to predict students' final grades, namely the
support model vector machine.
1. Prediction results for the science majors at SMAN 2 and SMAN 3 Pangkalpinang
After getting the MSE value, the data will be trained to get the final prediction. In this
process, datasets training Final exam score data and semester 1 to semester 4 scores are combined
to determine the final results of the decision model predictions tree (DT), Support Vector Machine
(SVM), and Neural Network (NN) in scenarios 1 to scenario 4, science majors at SMAN 2 and
SMAN 3 Pangkalpinang.
Figure 2. Prediction Results for the Science Department of SMAN 2 and SMAN 3 Pangkalpinan
Source: Data processed, 2024
Figure 2 shows the final result of a prediction using the three decision models tree (DT),
Support Vector Machine (SVM), and Neural Network (NN). For students with the initials AAS
who attend SMAN 2 with an average UAS score compared to student subject scores in each
semester, the predicted score from the model is obtained. decision tree is 85.64, Support model
Vector Machine is 86.40, and the Neural Network model is 86.22. Likewise, with students with
the initials ARF who attend SMAN 3 Pangkalpinang with an average UAS score compared to
student scores in each semester, the predicted score from the model is obtained. decision tree is
73.64, the Support model Vector Machine is 79.88, and the Neural Network model is 87.50.
2. Prediction results for the Social Sciences majors at SMAN 2 and SMAN 3 Pangkalpinang
After getting the MSE value, the data will be trained to get the final prediction. In this
process, datasets training Final exam score data and semester 1 to semester 4 scores are combined
to determine the final results of the decision model predictions tree (DT), Support Vector Machine
(SVM), and Neural Network (NN) in scenarios 1 to scenario 4 of the Social Sciences Department
at SMAN 2 and SMAN 3 Pangkalpinang.
Figure 3. Prediction Results for the Social Sciences Department of SMAN 2 and SMAN 3
Pangkalpinang
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5500
Source: Data processed, 2024
Figure 3 shows the final result of a prediction using the three decision models tree (DT),
Support Vector Machine (SVM), and Neural Network (NN). For students with the initials AYP
who attend SMAN 2 with an average UAS score compared to student subject scores in each
semester, the predicted score from the model is obtained. decision tree is 77.79, the model Support
Vector Machine is 78.99, and the Neural Network model is 80.37. Likewise, students with the
initials NLS who study at SMAN 3 Pangkalpinang with an average UAS score compared to
student subject scores in each semester will get a predicted score from the model. decision tree is
80.64, the Support model Vector Machine is 78.58, and the Neural Network model is 75.36.
Data Visualization & Dashboard Evaluation
In this section, a dashboard will be created that displays the student's initials, school of
origin, subject grades, semester, scenarios 1 to 4, and prediction range for each scenario. The
application used is Power BI. The results of the predictions that have been made will be submitted
to the programmer to create a dashboard. The aim of creating this dashboard is so that the school
can easily see the values and become the basis for making decisions regarding the learning
strategies that must be implemented.
Figure 4 SMAN 2 Pangkalpinang Science department dashboard
Figure 5 Social Sciences department dashboard at SMAN 3 Pangkalpinang
Figure 4 and Figure 5 are displays of the dashboard that has been created. The dashboard
is created with a simple display by using the slicer function to select the student's initials and
choose a scenario, a bar chart to display the range of predicted UAS scores for semester 1 to
semester 4 from 3 different models, namely Decision tree (DT), Support Vector Machine (SVM),
and Neural Network (NN), meanwhile the table is used to view students' subject scores based on
their subjects, UAS scores and school origin will automatically appear if you have selected the
student's initials and scenario.
Initially, schools can select the desired student's initials in the dropdown in the top left
corner of the dashboard, and the school must select the desired scenario. Then, automatically, the
Application of Data Mining for Prediction of High School Student Graduation Rates
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5501
UAS and School Origin scores will appear in the dashboard table, the purpose of using the
student's initials for privacy reasons. Predicted student final grades will appear on the bar chart at
the bottom of the dashboard along with the numbers in each Decision model tree (DT), Support
Vector Machine (SVM), and Neural Network (NN), which will move dynamically following
student data and the selected scenario (Aydoğdu, 2020). It is hoped that this dashboard will make
it easier for schools to see students' learning results each semester with the aim of students
receiving further treatment so they can pass their final school exams.
Conclusion
The results of students' exams and their graduation can be shown from the quality and
weaknesses of the students in terms of learning, so efforts are always made to support students
who have difficulty learning in educational units/programs. Student achievement, student
progress and student potential are very important for measuring learning outcomes and selecting
learning materials and learning activities. There are various influencing factors in the success rate
of graduating students. One of the main factors is the daily score which has an impact on test
scores, but that is not the only thing in predicting student performance in increasing graduation.
Other factors also play a role, including behaviour and discipline, which also influence it.
Increasing the number of students graduating in an educational unit/program can elevate and
increase the popularity of that educational unit/program. If an educational unit/program is
declining in its rating, this could have an impact on the attractiveness of prospective new students
who will enter that educational unit/program. So, of course, it can cause concerns for educational
units/programs. The aim is to find and build a model from a dataset and then use the model in
another dataset where you want to predict a result.
The MSE value in each scenario is different because each scenario has different data. In
determining the MSE value of the data set, extracting features and targets from the variables is
carried out so that the training dataset is also tested using regression with the model being tested,
namely the decision model. tree (DT), Support Vector Machine (SVM), and Neural Network
(NN). From these results, it can be concluded that: Of the four scenarios for the science major at
SMAN 2 and SMAN 3 Pangkalpinang with three different models, the MSE value shows that the
test results are in accordance with the testing dataset and can be used to predict the student's final
score, namely the decision model, tree, and support vector machine. Of the four scenarios for the
Social Sciences major at SMAN 2 and SMAN 3 Pangkalpinang with 3 different models, the MSE
value shows that the test results are in accordance with the testing dataset and can be used as a
prediction of students' final grades, namely the support model vector machine. The dashboard
created using Power BI shows that each scenario has a different predicted range for UAS scores
for each model. It is hoped that schools will take action so that students whose scores are still
below receive special and intensive treatment. The dashboard also shows the scores for each
subject, which students are expected to study more actively in subjects with scores that are still
quite low.
.
Bibliography
Akala, D. B. M. (2021). Revisiting education reform in Kenya: A case of Competency Based
Curriculum (CBC). Social Sciences & Humanities Open, 3(1), 100107.
https://doi.org/10.1016/j.ssaho.2021.100107
Akour, M., & Alenezi, M. (2022). Higher Education Future in the Era of Digital Transformation.
Education Sciences, 12(11), 784. https://doi.org/10.3390/educsci12110784
Altabrawee, H., Ali, O. A. J., & Ajmi, S. Q. (2019). Predicting Students’ Performance Using
Machine Learning Techniques. Journal of University of Babylon, Pure and Applied
Sciences, 27(1), 194–205.
Muhamad Kurniawan, Sani Muhamad Isa
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 11, November 2024 5502
Aydoğdu, Ş. (2020). Predicting student final performance using artificial neural networks in
online learning environments. Education and Information Technologies, 25(3), 1913–1927.
https://doi.org/10.1007/s10639-019-10053-x
Chicco, D., Warrens, M. J., & Jurman, G. (2021). The coefficient of determination R-squared is
more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis
evaluation. PeerJ Computer Science, 7, e623. https://doi.org/10.7717/peerj-cs.623
Courtiol, P., Maussion, C., Moarii, M., Pronier, E., Pilcer, S., Sefta, M., Manceron, P., Toldo, S.,
Zaslavskiy, M., Le Stang, N., Girard, N., Elemento, O., Nicholson, A. G., Blay, J.-Y.,
Galateau-Sallé, F., Wainrib, G., & Clozel, T. (2019). Deep learning-based classification of
mesothelioma improves prediction of patient outcome. Nature Medicine, 25(10), 1519–
1525. https://doi.org/10.1038/s41591-019-0583-3
Dou, J., Yunus, A. P., Bui, D. T., Merghadi, A., Sahana, M., Zhu, Z., Chen, C.-W., Han, Z., &
Pham, B. T. (2020). Improved landslide assessment using support vector machine with
bagging, boosting, and stacking ensemble machine learning framework in a mountainous
watershed, Japan. Landslides, 17(3), 641–658. https://doi.org/10.1007/s10346-019-01286-
5
Du, X., Yang, J., Hung, J.-L., & Shelton, B. (2020). Educational data mining: a systematic review
of research and emerging trends. Information Discovery and Delivery, 48(4), 225–236.
https://doi.org/10.1108/IDD-09-2019-0070
Fernandes, E., Holanda, M., Victorino, M., Borges, V., Carvalho, R., & Erven, G. Van. (2019).
Educational data mining: Predictive analysis of academic performance of public school
students in the capital of Brazil. Journal of Business Research, 94, 335–343.
https://doi.org/10.1016/j.jbusres.2018.02.012
Kementerian Pendidikan dan Kebudayaan. (2019). Peraturan Menteri Pendidikan dan
Kebudayaan Nomor 43 Tahun 2019 Tentang Penyelenggaraan Ujian yang
Diselenggarakan Satuan Pendidikan dan Ujian Nasional.
Li, L. (2024). Reskilling and Upskilling the Future-ready Workforce for Industry 4.0 and Beyond.
Information Systems Frontiers, 26(5), 1697–1712. https://doi.org/10.1007/s10796-022-
10308-y
Mienye, I. D., Sun, Y., & Wang, Z. (2019). Prediction performance of improved decision tree-
based algorithms: a review. Procedia Manufacturing, 35, 698–703.
https://doi.org/10.1016/j.promfg.2019.06.011
Sharifzadeh, M., Sikinioti-Lock, A., & Shah, N. (2019). Machine-learning methods for integrated
renewable power generation: A comparative study of artificial neural networks, support
vector regression, and Gaussian Process Regression. Renewable and Sustainable Energy
Reviews, 108, 513–538. https://doi.org/10.1016/j.rser.2019.03.040
Wajdi, M. B. N., Iwan Kuswandi, Umar Al Faruq, Zulhijra, Z., Khairudin, K., & Khoiriyah, K.
(2020). Education Policy Overcome Coronavirus, A Study of Indonesians. EDUTEC :
Journal of Education And Technology, 3(2), 96–106. https://doi.org/10.29062/edu.v3i2.42
Xiao, W., Ji, P., & Hu, J. (2022). A survey on educational data mining methods used for predicting
students’ performance. Engineering Reports, 4(5). https://doi.org/10.1002/eng2.12482