P-ISSN 3025-6615

Volume 1 Number 1
September 2023

Literature Review: Data Mining For Student Data Classification

Billi Mahardika1*
1
Laboratory Staff, Universitas Muhammadiyah Palembang
*billymahardika123@gmail.com

Ahmad Yani street No. 13, Palembang, South Sumatra 30263, Indonesia

Received September 25th, 2023; Revised September 29th, 2023; Accepted September 30th, 2023

Abstract
The abundance of student data and student graduation number data, hidden information can be found by
processing student data to be useful to the university. The processing of student data needs to be done to uncover
important information in the form of new knowledge (knowledge discovery) such as information on student data
classification based on profile and academic data. Therefore, in this research, the researcher plans to conduct a
literature review related to data mining for student data classification with the aim of finding out about data
mining data processing classification and collecting all designs used in identifying data starting from problems,
methodology, equations and results. For this research, researchers used historical data from students from 2007
to 2011 who had graduated. There were 9 research journals that researchers managed to find, each of which used
different algorithms or classification techniques. To conduct a literature review, researchers conducted a journal
review using PICOT. The results of this research are the success of researchers in classifying student data using
data mining techniques.

Keywords: Classification, Students, Data Processing.

1. INTRODUCTION
Higher education today must gain a competitive advantage using every available resource. In
addition to infrastructure and human resources, information systems are one of the resources that can be
used to collect, process and disseminate information to support day-to-day operations as well as
outsourcing activities. strategic decisions. Education has a very important role to improve and prepare
superior and highly competitive human resources (HR). This is where the role of higher education
institutions becomes very important in creating experts who are able to develop knowledge and
contribute to development. Higher education as one of the educational institutions, is required to be able
to provide quality and quality education to its stakeholders[1]. The abundance of student data and
student graduation count data, hidden information can be found by processing student data to be useful
to higher education institutions. The processing of student data should be performed to uncover
important information in the form of new knowledge (knowledge discovery), for example information
on classification of student data by profile and academic data. This new knowledge could help
universities rank student graduation rates to determine strategies to increase graduation rates in later
years. Data mining is a series of processes that aim to manually extract added value in the form of
unknown information from a database by extracting patterns from data with the aim of manipulating
the data to obtain More valuable insights by extracting and recognizing important patterns. or dig into
the data contained in the database [2].
In relation of Data Mining, there are several previous research that have become references for
researchers. The first is research from Cornelia Selvi Diana, Latifah Hanum and Saut Parsaoran Tamba,
who in this research implemented data mining using the K-means algorithm to determine the title of the
thesis and research journal by making FTIK UNPRI the research object. The results of this application
itself certainly provide convenience or solutions for students and their scope to find out ideas for thesis
titles and research journals. The second is research conducted by Ni Luh Putu Purnama Dewi, I Nyoman

1|Page
 Volume 1 Number 1
September 2023

Purnama and Nengah Widya Utami. In this research, data mining was applied to cluster lecturer
performance assessments using the K-means algorithm at STMIK Primakara. The results obtained from
this research are research on lecturer performance based on student satisfaction, namely very good
cluster 312 (31.74%) student data, good cluster 401 (40.79% student data, quite good cluster 189
(19.23%) student data and the less good cluster was 81 (8.24%). The DBI accuracy level was 0.270 or
27%. The last research was conducted by Hozairi, Anwari and Syarifu Alim, in this research
implementing orange data mininh to classify student graduation using a model. K-Nearest Neighbor,
Decision Treee and Naive Bayes in the Informatics Engineering Study Program, Madura Islamic
University, class of 2016. In this study, a comparison of classification models was carried out with the
results of K-Nearest Neighbor having an accuracy level of 77%, Decision Tree with an accuracy level
of 74%. % and Naive Bayes is 89%, thus the most recommended model is Naive Bayes.
From these three research, it can certainly be seen that there are many classification models or
classification algorithms for data mining. So in this research, researchers will conduct a literature review
related to data mining for student data classification with the aim of finding out about data mining
classification for data processing and collecting all designs used in identifying data starting from
problems, methodology, equations and results. For this research, researchers used historical data from
students from 2007 to 2011 who had graduated. Regarding the classification algorithm consists of 6
parts consisting of: Decision Tree Analysis, is a technique that belongs to the machine-learning family,
arguably the most popular classification technique in the data mining area. Statistical Analysis [3]
Statistical engineering was the source of popular classification algorithms for many years until the
emergence of machine learning techniques. Statistical classification techniques include logistic
regression and discriminant analysis, both of which assume that the relationship between input and
output variables is essentially linear, that the data are normally distributed, and that the data are normally
distributed. The variables are neither interdependent nor independent of each other. The nature of these
questionable assumptions eventually led to the shift towards machine learning techniques. Neural
Networks is one of the most popular techniques in Machine-Learning that can be used for classification
problems. Case-Based Reasoning, this approach uses historical cases to recognize similarities to
determine a new case into the most probable category. Bayesian Classifiers, this approach uses
probability theory to create classification models based on past events that can place a new instance into
the most probable class. Genetic Algorithms, using analogies to natural evolution to create a purposeful
search-based mechanism for classifying data samples.

2. RESEARCH METHODOLOGY
In this research, the researcher conducted a literature review, in which case the researcher
reviewed journals that matched the PICOT and search terms for journals through MESH (Medical
Subject Heading), limitations taken by journals and other things. The journal is used in literature reviews
obtained through databases that provide Indonesian scientific journals through Google Scholar and
websites such as Garuda Kemdikbud.
The researcher wrote down the keywords according to the MESH (Medical Subject Heading)
namely "processing", "data", "students" and selected the full text to appear 100 findings, then narrowed
down to Dissertations and Theses and found the next 9 findings sorted from the most recent. Regarding
the choice of language, it was not carried out because all the journals found used Indonesian. Each of
these questions has followed the PICOT where in each of these questions there is P =
Problem/Student/Population, I/E= Implementation /Intervention/Exposure, C=Control/Comparative
Intervention, O=Results and T=Time. It is relevant that the author used to get journals about the
classification of Data Mining processing student data. The author takes all the designs in the research
that are used in identifying student data

3. RESULTS AND DISCUSSION

In this research, using history data from students from 2007 to 2011 who have graduated with a
total of 377 data with 72 attribute values of the course and 1 target class in the form of study period.
This research was conducted by following the stages of data mining work which refers to the process of
knowledge discovery in databases. The data mining application was successfully built with experimental
results showing that the best study period classification results were obtained by selecting attributes

2|Page
Billi Mahardika, Literature Review: Data …

from all elective courses with accuracy values. Based on data obtained from the Department of
Computer Science/Informatics, Diponegoro University, students from the 2007 to 2011 class year with
a total of 377 graduates obtained information that the average student study period is still over 4 years.
Inna Alvi Nikmatum (2019) and Indra are alert. There are several data mining tools including Rapid
Miner, Orange, KNIME, Weka, Keel and R. WEKA is GUI-based so that it minimizes the use of coding
which can make it easier for system users. In this study using WEKA tools. The data processed at WEKA
has ARFF and CSV formats. Based on the experimental results, it was concluded that using three model
criteria, namely the Ratio Gain, Information Gain and Gini Index. The highest accuracy results are found
in the Gini Index criterion model, namely 92.18%. The highest result of the three feature selections is
the Information Ratio Gain with a value of p = 0.6 and the accuracy results are 92.46. Feature selection
is the process of selecting the right features to be used in the classification or clustering process. The
independent variable that gives the greatest t value is taken as X1 provided that H0 is rejected. The two
taken from the independent variable are taken as X2 provided that H0 is rejected. Wiwit Supriyanti
(2018), and Miss Puspitasari in the number of questionnaire results that were successfully collected were
981 samples. Of the 981 data collected, there are 948 valid data that can be used and 33 data that are
invalid due to an unbalanced dataset, so the dataset is adjusted randomly to 360 data with an ouas value
and 350 data with a dissatisfied value. The validity and readability tests were carried out on 30 randomly
selected data samples, meaning that each statement given by the respondent was correlated with the total
score and all were declared valid..
From the results of the implementation of the confusion matrix calculation, the accuracy output
of the classification model for the three algorithms is obtained as shown in Table 1. The three algorithms,
decision tree C4.5, SVM and Naïve Bayes, show quite good accuracy in classifying correctly both for
satisfied classes and the dissatisfied class is quite good, namely above 80%, where the accuracy of the
C4.5 algorithm and the SVM algorithm is better than the accuracy provided by the Naïve Bayes
algorithm. These results are consistent with previous studies which claim that each algorithm has good
performance on the dataset used. The results of the questionnaire show that 61% of students answered
they were satisfied with the learning facilities and infrastructure and 39% answered they were
dissatisfied. C4 Decision Tree Algorithm. 98% compared to the Naïve Bayes and Support Vector
Machine algorithms when modeled from training data and tested using test data from the student
satisfaction questionnaire dataset for learning facilities and infrastructure Elga Mariati, Ariesta Lestari,
and Widiatry (2020).

Table 1 : Table literature review

No. Writer Research Title Year Method Results
1. Inna Alvi Implementation 2019 K-Nearest Student data and student graduation data can
Nikmatum of Data Mining Neighbor produce abundant information, hidden information
& Indra for can be found by managing the data. Based on data
Waspada Classification of obtained from the Department of Computer
Student Study Science/Informatics, Diponegoro University,
Period Using the students from the 2007 to 2011 class year with a
K-Nearest total of 377 graduates obtained information that the
Neighbor average student study period is still over 4 years
Algorithm [4]. [5]
2. Marina Performance 2019 Decision One of the factors affecting the quality of a higher
Windarti & Comparison of 6 Tree, education institution is student learning outcomes,
Agustinus Data Mining Bayesian which can be measured over time of study. The
Suradi Classification Network, acquisition of knowledge in a database is often
Algorithms for K-Nearest referred to as data mining or data mining. This
Prediction of Neighbors, study aims to determine the performance of the six
classification algorithms used, which are Decision

3|Page
 Volume 1 Number 1
September 2023

Student Study Naïve Trees C4.5, Bayesian Networks, Nearest

Period Bayes, Neighbors, Naïve Bayes, Neural Networks, and
Neural SVMs. The acquisition of knowledge in a database
Network, is often referred to as data mining. Decision Tree
Svm C4. Bayesian Network has the best performance
with an accuracy value of 80.615%, precision and
recall values of 0.785 and 0.806, for AUC values
included in the good category, namely 0.837 [5].
3. Wiwit Implementation 2018 K-Nearest The experimental results concluded that using
Supriyanti of Forward Neighbor three criteria models namely gain ratio,
& Nona Selection information gain and Gini index, the highest
Puspitasari Feature accuracy results were found in the Gini index
Selection criterion model, namely 92.18%. The highest result
Techniques in of the three feature selections is the information
the Minig Data gain ratio with a value of p=0.6 and the accuracy
Classification results are 92.46. Feature selection is the process of
Algorithm for selecting the right features to be used in the
Predicting Study classification or clustering process. The
Period of independent variable that gives the greatest t value
Indonusa is taken as X1 provided that H0 is rejected. The two
Surakarta taken from the independent variable are taken as
Polytechnic X2 provided that H0 is rejected [6]
Students
4. Elga Engineering 2020 Decision Of the 16 attributes used, only 8 attributes affect
Mariati, Student Tree the level of customer satisfaction because they
Ariesta Satisfaction have a higher presentation of answers, the rest have
Lestari & Classification relatively small presentations. The use of the naïve
Widiatry Model for Bayes algorithm in this study resulted in an
Learning accuracy of 74.46% which stated that the amount
Facilities Using of correct data was greater than the wrong data.
Data Mining Input attributes of customer satisfaction used in
this study include price, facilities, service and
loyalty. The implementation of the support vector
machine algorithm gives a fairly good accuracy
value of 80%. Each of the previous studies only
implemented one algorithm, then stated that the
algorithm used had given good results in terms of
accuracy. The results of the questionnaire showed
that 61% of students answered that they were
satisfied with the learning facilities and
infrastructure and 39% answered that they were not
satisfied. C4 Decision Tree Algorithm. 98%
compared to the Naïve Bayes and Support Vector
Machine algorithms when modeled from training

4|Page
Billi Mahardika, Literature Review: Data …

data and tested using test data from the student

satisfaction questionnaire dataset for learning
facilities and infrastructure [7] .
5. Eka Sabna Data Mining 2021 Naïve The results of implementing data mining using the
& Yuda With 2 (Two) Bayes, Rapidminer software were carried out on two
Irawan Student classification algorithm models, namely C4.5 and
Performance NBC, then entering datasets as test material for the
Prediction two models, which contained experimental data
Classification and test data. By using the Associa on Rule
Models Method. in the Naive Bayes Classifier (NBC)
algorithm model is 80% while in the C4.5
algorithm model it is 60%. It can be concluded that
the best accuracy value from the results of the
comparison of the two algorithm models is
obtained by the Naïve Bayes algorithm model with
an accuracy value of 80% [8].
6. Resti, Dodo Application of 2021 Naïve Bayes This research used 100 student data for 2015 and
Zaenal Outstanding 2016 respectively, so that the total data used was
Abidin, & Classification 200. These attributes will be applied to the Naive
Errissya Data Mining at Bayes method for students majoring in Informatics
Rasywir Stikom Engineering class of 2015-2016 as many as 200
Dinamika data to predict student potential. achievers, which
Bangsa Jambi are categorized into 3 namely Very Potential,
Using the Naïve Potential, and Potential Enough. The results of this
Bayes Method study used the Use Training Set with a Correctly
Classified Instances accuracy percentage of 87%
and 13% Incorrectly Classified Instances. 5-cross
validation Correctly with a percentage accuracy of
Correctly Classified Instances of 77% and
Incorrectly Classified Instances of 23%. 10-Fold
Cross Validation with a Correctly Classified
Instances percentage of 78% accuracy, 22%
Incorrectly Classified Instances [9].
7. Anggi Application of 2022 Algoritma Algorithm C4.5 Calculation Process Calculation of
Trifani, C4.5 C4.5 Algorithm C4.5 to determine the dominant factors
Agus Classification of causing Stress can be described as follows: Step 1:
Perdana Data Mining in Count the number of cases, the number of cases for
Windarto, Determining Highly Stressful decisions, the number of cases for
Hendry Stress Levels of Not Stressed decisions. Step 2: Calculating the
Qurniawan Final Students entropy of all cases and cases divided by attribute
class with equation (1) then calculating the gain for
each attribute with equation (2). C4.5 Algorithm.
Generates 20 rules and the accuracy rate produced
by this method is 87.88%. Based on calculations

5|Page
 Volume 1 Number 1
September 2023

using the C4.5 Algorithm, the most dominant

factor is Interpersonal with a gain value of
0.328180116 [10].
8. Arief Main Student 2020 Algoritma Student main data at the preparatory stage was
Jananto,Sula Data as a CART taken from the Academic Administration Bureau
stri, Eko Nur Predictor of of the Faculty who had graduated in the graduation
Wahyudi Timeliness of data in the form of student name, gender, city of
Dan Sunardi Graduation birth, date of birth, name of school of origin for 5
Using the CART years distributed in each semester. From this
Algorithm for process the data obtained amounted to 1151
Data Mining records. By implementing categorical data mining
Classification techniques and the CART algorithm using repeated
Data Mining binary partitioning, it is expected to obtain a
decision tree that can be used to predict class
timeliness of graduation from active students. By
using graduation data and student main data of
1018 records, a decision tree model is obtained
with an accuracy rate of 63% from the data testing
[11].
9. Ni Luh Optimization of 2019 Algoritma The results of the classification on the training data
Ratniasih Data Mining C4.5 Dan are the GPA attribute as the root of the decision
Using Naïve KNN tree, while the other attributes are as child nodes.
Bayes and C4.5 From the training data with a total of 50 data
Algorithms for produced 5 rules. The rules that have been obtained
Classification of from the training data can be used as rules to
Student determine whether graduation is on time or not for
Graduations STMIK STIKOM Bali students. 4 predictor
attributes and 1 target attribute produce 5 rules in
the decision tree so that these rules can be used in
determining timely graduation for STMIK
STIKOM Bali students. The results of the analysis
using the Naïve Bayes method obtained an
accuracy of 89.27% where the performance results
are accurate [12].

Based on the data in Table 1, it can be said that the emphasis on the problems raised in Inna Alvi
Nikmatum and Indra Waspada's research has problems with the timeliness of students in completing
their studies and the proportion of students who complete their studies within the study period are
included in the element of accreditation assessment. In Marina Windarti and Agustinus Suradi's
research, the problem that can affect the quality of a tertiary institution is student performance which
can be measured through the length of study period. In Wiwit Supriyanti's research, and Miss Puspitasari
have problems processing student data needs to be done to find out important information in the form
of new knowledge (knowledge discovery), for example information about classifying student data based
on profiles and academic data. This new knowledge can help universities to classify student graduation
rates in order to determine strategies to increase graduation in the following years. In the research by
Elga Mariati, Ariesta Lestari, and Widiatry, assessment of student satisfaction with facilities and

6|Page
Billi Mahardika, Literature Review: Data …

learning at the Faculty of Engineering had been carried out before, but the assessment was still carried
out partially and the results of data collection on satisfaction assessment had not been evaluated before.
This study uses data mining techniques in classifying.
Next, namely the research steps or methods used, based on Table 1. In the research of Inna Alvi
Nikmatum and Indra Waspada using the K-Nearest Neighbor. Marina Windarti and Agustinus Suradi
used this study to understand the performance of the six classification algorithms used, namely C4.5
decision tree, Bayesian network, KNearest neighbor, Naïve Bayes, neural network and SVM. There are
several data mining tools including Rapid Miner, Orange, KNIME , Weka, Keel and R. WEKA is GUI-
based so that it minimizes the use of coding which makes it easier for system users. In this study using
WEKA tools. Data processed at WEKA has ARFF and CSV formats. In the research by Wiwit
Supriyanti, and Nona Puspitasari using the Information Gain Feature Selection Technique for Predicting
Student Academic Performance » argues that the main problem in the process of discovering knowledge
from data in the field of education is identifying representative data. J48, RandomForest, MLP, SVM to
predict students' academic performance in mathematics. In the research of Elga Mariati, Ariesta Lestari,
and Widiatry. In this study, the method from the data mining approach was applied to classify whether
students were satisfied or not with the quality of learning facilities at the Faculty of Engineering. This
study compares three data mining algorithms, namely Decision Tree C4.5, Support Vector Machine,
and Naïve Bayes to get the best algorithm for prediction systems.

4. CONCLUSION
Based on the research conducted by the researcher, it can be concluded that from the 9 studies
presented, static classification techniques include logistic regression and discriminant analysis. Both
assume that the relationship between input and output variables is basically linear, the data is normally
distributed, and the variables are not interrelated and independent of one another. Assessment of student
satisfaction with facilities and learning at the Faculty of Engineering has been carried out before, but the
assessment is still carried out partially and the results of data collection on satisfaction assessment have
never been evaluated before. This research uses data mining techniques in classifying. The research that
has been carried out is considered insufficient so it is advisable to classify student data in order to be
able to manage the data as a whole consisting of student personal data in the form of name, name, student
address, parents' names and have student GPA data so that it can be processed appropriately with the
like that, it can be seen how student performance, student accuracy, understanding of lessons and can
see graduation in a timely manner or not.

REFERENCES

[1] D. Purwandani, C. Sutarsih, “Pengaruh Mutu Layanan Sarana Dan Prasarana Terhadap Kepuasan
Mahasiswa Di Fakultas Pendidikan Teknologi Dan Kejuruan Universitas Pendidikan Indonesia,”.
[2] F. Gorunescu, “Data Mining Concept, Models And Techniques. Berlin Heidelberg: Springer.,”
2011.
[3] M. Alghobiri, “A Comparative Analysis Of Classification Algorithms On Diverse Datasets.
Engineering, Technology & Applied Science Research,” 2018.
[4] Inna Alvi Nikmatun, “Implementasi Data Mining Untuk Klasifikasi Masa Studi Mahasiswa
Menggunakan Algoritma K-Nearest Neighbor”.
[5] A. S. Mariana Windarti, “Perbandingan Kinerja 6 Algoritme Klasifikasi Data Mining Untuk
Prediksi Masa Studi Mahasiswa”.
[6] N. P. Wiwit Supriyanti1), “Implementasi Teknik Seleksi Fitur Forward Selection Pada Algoritma
Klasifikasi Data Mining Untuk Prediksi Masa Studi Mahasiswa Politeknik Indonusa Surakarta,”
2018.
[7] W. C Elga Mariati A,1,*, Ariesta Lestari B,2, “Model Klasifikasi Kepuasan Mahasiswa Teknik
Terhadap Sarana Pembelajaran Menggunakan Data Mining”.
[8] Y. I. Eka Sabna1, “Data Mining Dengan 2 (Dua) Model Klasifikasi Untuk Prediksi Kinerja
Mahasiswa,” Http://Doi.Org/10.33060/Jik/2021/.
[9] E. R. Resti1, Dodo Zaenal Abidin2, “Penerapan Data Mining Klasifikasi Untuk Memprediksi

7|Page
 Volume 1 Number 1
September 2023

Potensi Mahasiswa Berprestasi Di Stikom Dinamika Bangsa Jambi Dengan Metode Naive
Bayes,” Vol. Vol.3, 2021.
[10] Anggi Trifani, “Penerapan Data Mining Klasifikasi C4.5 Dalam Menentukan Tingkat Stres
Mahasiswa Akhir,” 2022.
[11] S. Arief Jananto, Sulastri, Eko Nur Wahyudi, “Data Induk Mahasiswa Sebagai Prediktor
Ketepatan Waktu Lulus Menggunakanalgoritma Cart Klasifikasi Data Mining,” Doi
10.32736/Sisfokom.V10i1.991.
[12] Ni Luh Ratniasih, “Optimasi Data Mining Menggunakan Algoritma Naïve Bayes Dan C4.5 Untuk
Klasifikasi Kelulusan Mahasiswa”.

8|Page