International Journal of Pure and Applied Mathematics

Volume 118 No. 8 2018, 531-537

ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version)
url: http://www.ijpam.eu
Special Issue
ijpam.eu

Review on Prediction Algorithms in Educational

Data Mining
A.Dinesh Kumar1 , R.Pandi Selvam2, K.Sathesh Kumar3
1, 2
PG Department of Computer Science , Ananda College, Devakottai
3
School of Computing, Kalasalingam University, Krishnankoil, – 626126, Tamil Nadu, India.
1
[email protected] , [email protected] , [email protected]

data is a fascinating research field also known as Educational

Abstract—In present day educational system, a student’s Data Mining (EDM). EDM examines data that are generated by
performance is influenced by many factors. Students should be the educational institution such as prediction of the student
properly motivated to learn. Motivation leads to interest, interest performance, learning analytics of the student, grouping the
leads to success. Proper assessment of abilities helps the students to student according to their performance and recommending to the
perform better in their education. Data mining techniques have student.
been applied to improve the performance of the student. Predicting
student performance is one of the prominent research fields in EDM analyze data created by any type of information system
EDM. Applying data mining methods in educational data is an
interesting research area nowadays. Many studies of EDM have
supporting learning or education in schools, colleges,
focused on the data mining algorithms related with the prediction. universities and other academic or professional learning
This paper reviews the prediction algorithms and data mining tools institutions providing conventional model of teaching, as well as
used in educational data mining and future insights of better easy learning. Predicting students’ results and student modeling
prediction algorithm to be identified and new data mining tools to have been the primary goals of educational data mining. These
be used to predict the students’ performance, which helps the two issues are intensely connected with educational environment
instructor and institution to increase their study level. [2].

Keywords—Data Mining, Educational Data Mining and This paper is organized as follows. Section II introduces the
Prediction. related work of the prediction techniques. Section III provides
prediction techniques used in educational data mining. Section
IV provides a tabulated format of research works that have been
I. INTRODUCTION carried out in prediction techniques that are used in educational
data mining. Section V discusses the conclusion of this paper.
Knowledge Discovery in Databases (KDD) is an automatic,
II. BACKGROUND STUDY
exploratory analysis and modeling of large data warehouse.
KDD is the controlled process of identifying valid, new, useful,
Mustafa Agaolglu conducted the research on predicting
and understandable patterns from huge and complex data sets.
instructor performance in higher education. They have used the
Data Mining (DM) is the heart of the KDD process, concerning
artificial neural networks, classification algorithms, decision
infer of algorithm that explore the data, develop the model and
trees, support vector machines and discriminant analysis
find out previously unfamiliar patterns. The model is used for
algorithms for predicting the instructor performance. Their
understanding phenomena from the data, analysis and prediction
results shows C5.0 classifier is the best algorithm to predict the
[1,36].
performance of instructor [3].
Educational data mining is an independent research field in
Sushil and Thakur applied fuzzy association rule mining to
data mining. Applying data mining techniques in an educational
predict the student’s performance in end semester. They used the

531
International Journal of Pure and Applied Mathematics Special Issue

data like attendance, midsem marks, previous semester marks III. PREDICTION TECHNIQUES
and previous academic records that collected from the previous
data base. Based on these data they analysed some hidden In educational data mining, prediction techniques are used to
patterns of student poor performance [4]. predict the performance of the student. In order predict the
performance of the student several tasks are used which are
Mukesh Kumar and A.J. Singh explained their study based classification, regression and density estimation.
on the performance of 412 students. They used Naive Bayes,
Random Forest, PART and Bayes Network for predicting the A. Classifications
student performance. They were found random forest algorithm
gave the best result comparing to other algorithms for predicting Classification is a supervised learning technique whose aim
the student performance [5]. is to create a model, in this specific case, called classifier, which
can classify the class label of unknown data. In other words, a
Kamal et al described a performance study on B.A first year classifier is created from a training set and it is then used to
students from Vikram University, Ujjain, India. They were used classify unknown data, into one of the existing classes.
ID3, C4.5 and CART algorithm to predict the first year student Classification is a two step process: learning phase and the
performance. Based on these algorithms they found a final grade classification phase. Using the mapping function one can
of student in a course [6]. classify any attribute vector in the classification phase. To
evaluate the classifier, an already classified input is considered
Bevinda Alisha et al conducted a study on predicting drop and its accuracy is calculated as the percentage of correct
out students’ performance. They have used the attributes like classification obtained [11]. Several algorithms are used under
Previous Semester Marks, Internal Grades to predict the classification tasks that have been applied to predict the
student’s final semester marks and also used different performance of the student. They are Decision tree, Bayesian
classification algorithms like ID3, C4.5, CART, CHAID. Their Classifier, Artificial Neural Network, Support Vector Machine
result shows that CHAID algorithm predicted the performance and K-Nearest Neighbor algorithms.
of drop out students with highest accuracy [7].
B. Decision Trees
Shaleena and Shaiju Paul conducted a study on predicting the
student performance by using decision trees, class imbalance Decision trees are the best known classification model. DT is
and cost sensitive classification method. They found the relevant a tree in which each branch node represents a choice between a
factors and relationships that lead to a student to pass or fail. number of alternatives, and each leaf node represents a decision.
They concluded some factors are related with the student failure Decision tree are commonly used for gaining information for the
[8]. purpose of decision- making. Decision tree starts with a root
node on which it is for users to take actions. From this node,
Dinesh and R.V.Radhika invsetigated the performance of the users will split each node recursively according to decision tree
students using feature selection method. They focused about the learning algorithm. The final result is a decision tree in which
factors that are related with the performance of the student then each branch represents a possible scenario of decision and its
they compared the environmental factor and educational institute outcome [17].
factor which most affect the performance of the student. Finally
C. Bayesian Classifier
they concluded environmental factors are affecting the
performance of the student [2]. Bayesian classifiers are statistical classifier that is
represented visually as a graph structure. This classifier predicts
Dr.N. Tajuniza et al conducted a study on the factors that the class membership by probabilities, such as the probability
affect the academic achievement of the student. They used that a given sample belongs to a particular class. Several Bayes
classification techniques and mapreduce technique to predict the algorithm have been developed, among which Bayesian and
performance of the student [9]. naïve Bayes are the two essential techniques. Naïve Bayes
algorithm assumes that the effect that an attribute plays on a
AS Galathiya et al analyzed a research on classification with given class is independent of the values of other attributes [34].
an improved decision tree algorithm using feature selection
D. Artificial Neural Network
technique. They used genetic search algorithm to improve the
classification accuracy. The classification accuracy was Artificial neural networks are much admired in pattern
improved by implemented the diversities of algorithm using recognition. It is a set of connected input/output units and each
RGUI with weka packages [10]. connection has a weight present with it. During the learning
phase, network learns by adjusting weights so as to be able to

532
International Journal of Pure and Applied Mathematics Special Issue

predict the correct class labels of the input tuples. Neural For the purpose of this study, 20 papers were chosen from
networks are well studied for continuous valued inputs and Educational Data Mining. We have analyzed the performance
outputs. These are best at identifying patterns or trends in data of the student is predicted by many classification algorithms.
and well studied for prediction of student performance [17].
The research has been done to improve the performance of the
E. Support Vector Machine student based on predicted results.x This prediction helped the
instructor and institution to know about the weak student status
Support Vector Machines are best method when the class and take proper assessment take on the student to improve their
boundaries are nonlinear but here is too little data to learn
study level. In this paper we have reviewed the papers that have
composite nonlinear models. The fundamental idea is that when
the data is mapped to a higher dimension, the classes become been published in the year from 2011 to 2017. Form their
linearly divisible. In practice, the mapping is done only research work the paper title, algorithms, tools and results are
absolutely, using kernel functions. Support Vector Machine listed in the below table. Most of the research has been done
(SVM) focus on only the class boundaries; points that are any under decision tree classification algorithms naïve bayes
way easily classified are skipped. The objective is to find the classification algorithm to predict the performance of the
“thickest hyperplane”, which splits the classes [35].
student.
F. K-Nearest Neighbor
TABLE I
K- Nearest neighbor classifier represents a totally dissimilar PREDICTION ALGORITHM AS APPLIED IN EDM
approach to classification. They do not build any clear universal
model, but estimated it only locally and implicitly. The main
idea is to classify a new object by examining the class values of No Title of Paper Algorithm Tools Results
the K most alike data points. The selected class can be either the Used
most frequent class among the neighbors or class distribution in Data Mining: A Bayesian MatLab Predicted the
neighborhood. The only learning task in K-nearest neighbor Prediction for Classification High
classifiers is to select two important parameters; the number of Performance Potential
neighbors k and distance metric d [35]. 1 Improvement Variable that
Using effect
G. Regression Classification student’s
performance
[15]
Regression analysis is a statistical methodology that is most
often used in numeric prediction [12] .The objective of this task Data Mining: A Bayesian MatLab Predicted
is to achieve a function of the independent variables that allows Prediction of Classification Student’s
performer or Final Mark
computing the conditional expectation of a dependent variable
2 under performer
for prediction and forecasting exercises based on the using
minimization of a certain type of error via an iterative procedure. classification
Nearly, Classification and Regression Trees (CART)
summarizes these tasks [13]. [16]
Mining ID3 Weka Predicted
H. Density Estimation Educational The student
Data to Analyze End
Density estimation is concerned with the estimation of 3 Students’ Semester
probability masses, univariate densities, joint densities, and Performance Performance
conditional densities. Most of the existing estimators assume
that all the data instances are available at once. With the ever [17]
increasing amounts of data and a tendency towards online Data Mining for CART Matlab Predicted
settings, however, there is an increasing demand for density Engineering overall
estimation on data streams (LINK). Some density estimation schools performance
techniques are inference, pattern mining, or outlier detection Predicting of students
4 Student
[14].
Performance and
IV. PREDICTION METHODS USED IN EDM Enrollment in
Master
Programs[18]

533
International Journal of Pure and Applied Mathematics Special Issue

Efficiency of C4.5 and ID3 Weka Predicted [2]

Decision Trees student
in Predicting performance
5 Student’s based on
Academic Internal and
Performance External
[19] Marks
Prediction of Support Vector Rapid Predicted the
student Machine, K- miner relationships Mining Artificial Weka Predicted
performance by Means between Educational Neural students
6 an application of Clustering student’s Data to Predict Network , C4.5 behavioral
data mining behavioral Student’s features
techniques [20] and their 13 Performance
success Using Ensemble
Data Mining: A C4.5, ID3 and Weka Predicted Methods[26]
Prediction for CART first year
performance Engineering
Improvement of students
7 Engineering performance Educational Naïve Bayes Weka Predicted
Student Using Evaluation and student
Classification Prediction of behavior of
School academic
[21] 14 Performance performance
through Data
Data Mining OneR, C4.5, Weka Predicted the Mining and
techniques in Multilayer factors that Genetic
EDM for Percepton, affect the Algorithms
8
predicting the Nearest student’s [27]
performance of Neighbour performance
Prediction of Naïve Bayes Weka Predicted
students[22] algorithm
Students student
Mining Student Naïve Bayes Weka Predicted 15 Performance performance
Academic Students using at top of the
9 Performance Academic Educational semester.
Performance Data Mining[28]
[23] Educational C4.5 ID3, Weka Predicted the
Data Mining Naïve Bayes, Weka Predicted Data Mining & CART and and students’
approach for Multilayer Student Students’ CHAID Rapid performance
Predicting Perceptron and success in a Performance Miner based on
16
Student C4.5 course and Prediction[29] related
10
Performance the personal and
[24] performance social
of learning factors.
methods Predicting Linear Weka Predicted
Data Mining: A ID3 Weka Predicted the Students Regression, student
Prediction for final grade of Performance in Multilayer performance
Student’s the student Final Perceptron in final
11 Performance 17 Examination examination
Using using Linear
Classification Regression and
Method[25] Multilayer
Mining ID3, C4.5 We ka Predicted Perceptron[30]
Educational Educational Predicting Decision Tree, Rapid Predicted
Data to and Student ANN, Random Miner students
Predicting Environment Academic Forest Tree graduation
12
Higher al Factors of 18 Performance performance
Secondary student using Data in final year
Students Mining
Performance Methods[31]

534
International Journal of Pure and Applied Mathematics Special Issue

Is Alcohol BFTree, C4.5, Weka Predicted

Affect Higher RepTree and student
Education Simple CART performance
Students based on
Performance: alcohol
19 Searching and consumption
Predicting student data
pattern using set
Data Mining
Algorithms
[32]
A Comparative ID3, C4.5, Weka Predicted
Analysis of CART, Student
Decision Tree CHAID Performance
20 Algorithms for in University
Predicting Results
Student’s
Performance[33] Fig. 2 Tools used in EDM
Based on table we have listed the prediction algorithm as a The Fig.2 illustrate most of the research has been applied in
figure that is most often used by the researchers in educational weka tool for prediction. Matlab and Rapid miner tools are
data mining for prediction. The Fig.1 Shows the prediction rarely used to predict the performance of the student by the
algorithm mostly used in EDM. researchers.
V. CONCLUSION AND FUTURE WORK

In this paper different prediction techniques and prediction

tools are discussed in educational scenario to predict student
performance. Prediction of student academic performance helps
both the teacher and parents to predict about their success and
failure in examinations. From this paper we concluded that most
of the research has been done on same prediction algorithms,
student variables and data mining tools. Their work was only
different from to improve the classification accuracy and
different factors related to the student performance. Symbiotic
structure learning algorithm and feed forward neural-network-
aided grey model is not yet applied for predicting student
performance. In future this algorithm and model can be applied
for predicting student performance or new algorithm; new
student variable and new data mining tools can be also identified
for better prediction based on this study.
REFERENCES
[1] Oded Maimon Lior Rokach, “Data Mining and Knowledge Discovery
Fig.1 Prediction Algorithm of EDM
Handbook”,Second Edition, Springer.
The Fig.1 says that the C4.5 (J48) is often used by most [2] A.Dinesh Kumar ,Dr.V.Radhika, “Mining Educational Data to Predicting
researchers to predict the performance of the student then ID3, Higher Secondary Students Performance”, International Journal of
CART and Naïve Bayes algorithms are frequently used by the Computational Intelligence and Informatics(IJCII), Vol 6 ,September 2016.
researchers. The other algorithms are rarely used by the [3] Mustafa Agaoglu,” Predicting Instructor Performance Using Data Mining
Techniques in Higher Education”, IEEE, Vol 4, 2016.
researchers. Some researchers used combination of these
[4] Sushil Kumar Verma, R.S. Thakur, “Fuzzy Association Rule Mining based
algorithms to predict the performance of the student in
Model to Predict Students’ Performance”, International Journal of
educational data mining. Electrical and Computer Engineering (IJECE) Vol. 7, No. 4, August 2017
[5] Mukesh Kumar Prof. A.J. Singh, “Evaluation of Data Mining
Techniques for Predicting Student’s Performance”, Modern
Education and Computer Science, 2017, 8, 25-31.

535
International Journal of Pure and Applied Mathematics Special Issue

[6] Kamal Bunkar Rajesh Bunkar,” Data Mining: Prediction for Performance Classification Method”, World Journal of Computer Application
Improvement of Graduate Students using Classification”, IEEE, 2012. and Technology, 2014.
[7] Bevinda Alisha Pereira, Anusha Pai,” A Comparative Analysis of [26] Elaf Abu, Amrieh, Thair Hamtini, Ibrahim Aljarah, “Mining
Decision Tree Algorithms for Predicting Student’s Performance”, Educational Data to Predict Student’s Performance Using
Ensemble Methods”, International Journal of Database and
International Journal of Engineering Science and Computing,
Theory and Application,Vol 9, 2016.
Vol 7, 2017. [27] Patinon Galvan,” Educational Evaluation and Prediction of School
[8] Shaleena K.P Shaiju Paul,” Data Mining Techniques for Performance through Data Mining and Genetic Algorithms”, Future
Predicting Student”, International Conference on Engineering Technologies conference IEEE,2016.
and Technology (ICETECH), IEEE 2015. [28] Ms.Tismy Devasia,Ms.Vinushree , “Prediction of Students Performance
[9] Dr.N.Tajunisha and M.Anjali,” Predicting Student Performance using Educational Data Mining “, Data Mining and Advanced Computing
Using Mapreduce”, International Journal of Engineering and (SAPIENCE) IEEE-2016.
Computer Science (IJECS) ,Vol.4, Issue1-2015. [29] Amjad Abu Saa , “Educational Data Mining & Students’ Performance
[10] As.Galathiya and AP.Ganatra, “Classification with an improved Prediction,International Journal of Advanced Computer Science and
Decision Tree Algorithm”, International Journal of Computer Applications (IJACSA), Vol. 7, 2016
Application (IJECS), Vol 46, 2012. [30] Febrianti Widyahastuti , Viany Utami Tjhin , “Predicting Students
[11] Ricardo Mendes And Joao P.Vilela, “ Privacy- Preserving Data Performance in Final Examination using Linear Regression and Multilayer
Mining: Methods, Metrics, and Applications”, IEEE, 2017. Perceptron”, IEEE, 2017.
[12] Jiawei Han and Micheline Kamber ,” Data mining concepts and [31] Raheela Asif , Saman Hina and Saba, “Predicting Student Academic
Performance using Data Mining Methods”, International Journal of
techniques”, Second Edition.
Computer Science and Network Security (IJCSNS), VOL.17, 2017.
[13] Chady EI Moucary “Data mining for Engineering Schools”, International
[32] Saurabh Pal, Vikas Chaurasia, “Is Alcohol Affect Higher
Journal of Advanced Computer Science and Applications(IJACSA) Vol.2,
Education Students Performance: Searching and Predicting
2011. pattern using Data Mining Algorithms”, International Journal of
[14] www.datamining.informatik.uni-mainz.de Innovations & Advancement in Computer Science IJIACS,
[15] Birijesh Kumar Bharadwaj, Saurabh Pal,” Data Mining: A Volume 6, April 2017.
Prediction for performance improvement using classification”, [33] Bevinda Alisha, Anusha Pai, “A Comparative Analysis of
International Journal of Computer Science and Information Decision Tree Algorithms for Predicting Student’s Performance”,
Security, Vol 9, 2011. International Journal of Engineering Science and Computing
[16] Umesh Kumar Pandey S.Pal, “Data Mining:A Prediction of (IJESC) ,Vol 7,2017."
performer or underperformer using classification”, International [34] Dorina Kabakchieva, “Predicting Student Performance by Using
Journal of Computer Science and Information Technologies Data Mining Methods for Classification”, Cybernatics and
(IJCSIT) Vol 2(2),2011. Information Technologies, Vol 13, 2013.
[17] Birijesh Kumar, Saurabh Pal, “ Mining Educational Data to [35] C Romero, Ventura, Pechenizkiy, and Baker Rsjd,“ Handbook of
Analyze Students’ Performance”, International Journal of Educational Data mining”, 2010.
Advanced Computer Science and Applications Vol 2, 2011. [36] Shankar, K. “Prediction of Most Risk Factors in Hepatitis
[18] Chady EI Moucary, “Data Mining for Engineering schools Disease using Apriori Algorithm.” RESEARCH JOURNAL OF
Predicting Student Performance and Enrollment in Master PHARMACEUTICAL BIOLOGICAL AND CHEMICAL
Programs”, International Journal of Advanced Computer Science SCIENCES 8.5 (2017): 477-484.
and Applications (IJACSA), Vol. 2, 2011.
[19] S.Anupama Kumar, Dr.Vijayalskhmi, “Efficiency of Decision
Tress in Predicting Student’s Academic Performance”,
CCSEA,2011.
[20] Sajadin Sembiring and Zarlis ,”Prediction of student
performance by an application of data mining
techniques”,International Conference on Management and
Artificial Intelligence IPEDR vol.6 2011.
[21] Surjeet Kumar Yadav, Saurabh Pal, “ Data Mining: A Prediction
for Performance Improvement of Engineering Students using
Classification”, World of Computer Science and Information
Technology Journal (WCSIT), Vol 2, 2012.
[22] Ajay Kumar Pal, Saurabh Pal, “ Data Mining Techniques in
EDM for Predicting the Performance of Students”, International
Journal of Computer Science and Information Technology,
Volume 2, 2013.
[23] Azwa Abdul, Nur Hafieza, Fadhilah Ahmad, “ Mining Students’
Academic Performance”, Journal of Theoretical and Applied
Information Technology”, Vol 53, 2013.
[24] Edin Osmanbegovic and Mirza Suljic, “Data Mining approach for
Predicting Student Performance”, Journal of Economics and
Business, May 2012.
[25] Abeer Badr El Din Ahmed, Ibrahim Sayed Elaraby, “Data
Mining: A Prediction for Student’s Performance Using

536
537
538