http://www.iaeme.com/IJCIET/index.asp 48 editor@iaeme.com
International Journal of Civil Engineering and Technology (IJCIET)
Volume 10, Issue 02, February 2019, pp. 48–56, Article ID: IJCIET_10_02_007
Available online at http://www.iaeme.com/ijciet/issues.asp?JType=IJCIET&VType=10&IType=2
ISSN Print: 0976-6308 and ISSN Online: 0976-6316
©IAEME Publication Scopus Indexed
A COMPLETE SURVEY ON PREDICTING
PERFORMANCE OF ENGINEERING STUDENTS
V. Sathya Durga
Research Scholar, Department of Computer Science,
Hindustan Institute of Technology and Science, Padur, India
J. Thangakumar
Associate Professor, Department of Computer Science,
Hindustan Institute of Technology and Science, Padur, India
ABSTRACT
Predictive models are quasi experimental structures used to determine the future
patterns in data. These meaningful data patterns form the building block of any
decision support system. Researchers all over the world have built many prediction
models for major industries. Research works in the educational sector has increased
steeply. This steep increase may be due to the high availability of data in the
educational domain. This survey tries to comprehend a few literary works on
academic performance prediction of engineering students with the focus on grade
predictions. Meaningful interpretations have been made and inferences are presented
at the end of this paper.
Key words: Prediction, Predictive Modeling, Predictive Analytics.
Cite this Article: V. Sathya Durga and J. Thangakumar, A Complete Survey on
Predicting Performance of Engineering Students, International Journal of Civil
Engineering and Technology (IJCIET) 10(2), 2019, pp. 48–56.
http://www.iaeme.com/IJCIET/issues.asp?JType=IJCIET&VType=10&IType=2
1. INTRODUCTION
Prediction is a powerful data analysis task used to determine the future behavior of a user or a
system. The task of prediction has aided the decision makers to take the right decision at the
right time. Predictive activities have gained wide popularity owing to the high availability of
surplus data. Computers generate nearly 2.5 quintillion bytes of data every year [1].
Hidden within those mounds of data is the knowledge that could change the life of a
patient, or change the world. [2]
- Atual Butte, Stanford University.
Making predictions from a set of data is an extremely challenging task. We need
predictive models which are simple analytical structures that work on a set of predictors and
forecast the probabilities of future occurrences.
V. Sathya Durga and J. Thangakumar
http://www.iaeme.com/IJCIET/index.asp 49 editor@iaeme.com
1.1. Predicting Students Academic performance
Predicting students performance is a significant research work in the educational sectors. Few
Years back, students waited till the end of the semester to know their GPA. With the advent
of machine learning, we can predict all semesters grades easily and we have models which
predict the likelihood of a student being placed or not.
1.2. Four Core Components of a Predictive Model
Building a predictive model is very challenging work. To Build a predictive model, we
require four important components. These components are listed below for easy reference.
 The Methodology followed to deploy the model.
 Techniques adopted to build the prediction model.
 Input attributes used by the model and
 Performance Metrics used to evaluate the model.
This survey tries to the unravel common methodology, techniques, attributes and metrics
used in predicting the performance of engineering students. Students considered for this study
are from branches of civil engineering, Mechanical, Computer Science and Information
Technology.
1.3. Research Question and Objectives of the survey
Every research work, be it descriptive or analytical, applied or fundamental, needs a set of
research questions to work on. The research question and objective of this study is:
 To find out the common methodology adopted in building a predictive model.
 To find the important techniques used in predicting the performance of students.
 To identify the commonly used attributes for building the prediction models.
 To list out the frequently used performance metrics.
1.4. Research Methodology
The methodology followed to conduct the survey is outlined below:
 First popular databases like IEEE, Scopus, ACM libraries were searched and research papers
were collected.
 Research works older than 2013 and works with accuracy percent less than 65% were
discarded and not considered for the survey.
 The research papers selected was studied in detail. In-depth analysis was carried out. Many
important interpretations were made and Research questions were answered.
 This paper title, 'A Complete survey on predicting performance ofengineering students' was
prepared and presented.
2. LITERATURE REVIEW
2.1. Performance Prediction using Naive Bayes Classification
Fahad et al. (2018) present a prediction model to predict GPA of students using Naive Bayes
technique and Kmeans clustering. The highest accuracy achieved in this work was 98.8%.
[3]. Srinivas and Ramaraju (2017) build a prediction model using a Naïve Bayes Classifier.
Dataset consisted of 257 academic records. Two feature selection technique namely
A Complete Survey on Predicting Performance of Engineering Students
http://www.iaeme.com/IJCIET/index.asp 50 editor@iaeme.com
cfsSubsetEval and GainRatioAttributeEval was used. Results reveal cfsSubsetEval perform
better with 84% accuracy. [4]
Figure 1 Accuracy of Naive Bayes Classification
2.2. Performance Prediction using Artificial Neural Network (ANN)
Bendangnuksung and Prabu (2018) implement Neural Network to predict students’
performance. Cost Function with Accuracy were the metrics used in this research work.. Their
Model attained 83.4 % accuracy [7]. Okbu et.al develop a prediction model based on
Recurrent Neural Network (RNN) . Grades were predicted from log data. Results reveals that
RNN works well with an accuracy of 90% [8].
2.4. Performance Prediction using Decision Tree
Karishma and Swati (2016) use ID3 and C4.5 algorithm to identify students who are probable
to fail. Experiments were done on 200 Engineering student’s dataset. Applying the Decision
Tree on students internal marks, their performance was predicted. The Accuracy of the
classifiers was calculated to be 98.5 % [9]. Sumitha and Vinothkumar (2016) present a
comparative study of classification techniques to predict academic performance. This research
work experiments data of 350 students. Out of the classifiers, it was proved J48 was highly
influential with 97% accuracy[10] .
2.3. Performance Performance Prediction using Support Vector Machine (SVM)
Eashwar and Venkatesan (2017) develop a prediction model using Support Vector Machine to
recognize postgraduate students for carrying out doctoral studies. Data were clustered and
SVM technique was applied. 96.7% accuracy was achieved [13]. Jamuna and Shoba (2017)
compare various machine learning algorithms. Students data from a private institute was
taken. Results unravel the Polynomial Kernel achieved 97.62 % accuracy [14].
98.8
84.23
93.17
68.5
0
20
40
60
80
100
Fahad Razaque
2018 [3]
Srinivas A
2017 [4]
Ihsan A. Abu
Amra
2017 [5]
Azwa Abdual
Aziz
2014 [6]
AccuracyoftheWork
Author Name and Year of Publication
Accuracy of Naive Bayes Classification Works
Accuracy
V. Sathya Durga and J. Thangakumar
http://www.iaeme.com/IJCIET/index.asp 51 editor@iaeme.com
Figure 2 Accuracy of Decision Tree
2.5. Performance Performance Prediction using fuzzy system
Ramjeet, et al. propose a new fuzzy expert system to evaluate the performance of students.
First semester exam results and output values are fuzzified. Application rules were
determined, performance value was defuzzified and final marks were calculated [15]. Shilpa
and Bakal use three nodes fuzzy logic system in their research work. The Triangular
membership function was used in this research work. For Defuzzification Center of gravity is
used. Scores obtained was calculated with the score matrix and accuracy matrix [16].
Table 1 Fuzzy Functions used in Literary works
S.No Author Fuzzification
Rule
Inference
De
fuzzificatin
1.
Ramjeet
Singh
Yadav [19]
Triangular
Membership
function
Mamdani
max–min
inference
Center of
area
2.
Shilpa N
[20]
Triangular
Membership
function
Mamdani
max–min
inference
Ceter of
gravity
2.6. Performance Prediction using optimization technique
Hind Almayan and Waheeda Al Mayyan (2016) use Swarm Optimization method to build
their prediction model. Using Dimensionality reduction technique, feature space was reduced.
PSO is used for feature selection. C4.5 Classifier achieves 93.5% accuracy [17]. Ramanathan
et al. (2016) implements a Neural Network with a Lion – Wolf optimization technique for
weight selection. The model predicts one to eight semester exam performance of college
students. The trained network performances well compared to other prediction models with
RMSE - 2.3 [18].
98.5 97
92 94.41
0
20
40
60
80
100
Karishma B.
Bhegade 2016
[9]
R. Sumitha
2016 [10]
Tribhuvan A.P
2015 [13]
Tripti Mishra
2014 [14]
Accuracyofthework
Author Name and Year of Publication
Accuracy of Decision Tree Works
Accuracy
A Complete Survey on Predicting Performance of Engineering Students
http://www.iaeme.com/IJCIET/index.asp 52 editor@iaeme.com
Table 2 Accuracy Comparison of Hybrid ANN
2.9. Attributes used in prediction model
This section lists the commonly used attributes in students performance prediction works.
Total of eight attributes is listed in the following table. The most commonly occurring
attributes are identified and presented in the inferences. section.
Table 3 Attributes used in the Literary Works
S.N
o
Author Model
Accur
acy
RMSE
1.
Md. Fahim
Sikder [17]
ANN 99.8 % 0.1765
2.
Ramanathan L
[18]
Hybrid -
ANN
94 % 2.3
S.No Author 1 2 3 4 5 6 7 8
Naive Byaes
1. A. Srinivas [4]
Previous
marks
High
School
grades
Senior
secondrary
grades
Family
annual
income
Medium of
instruction
General
proficiency
Fathers'
education
Mothers'
education
2.
Ihsan A. Abu
Amra [5]
Previous
marks
Gender
Date of
birth
Students
branch
Location
School
name
Marital status
Father's
job
Neural Network
3. Bendangnuksu
ng [7]
Previous
marks
Nationality Stages
Parents
responsible
Raised
hands
Visited
resource
Parent
answers
Views
updates
4.
Md. Fahim
Sikder [19]
Previous
semester
marks
Class Test
Marks
Class
attendance
Lab
performance
Study time
Family
education
Living area
Extra
curicular
activity
SVM
5.
Huda Al-
Shehri [20]
Previous
marks
Family type
Parents'
status
Parent's job
Extra paid
classed
Extra
curricular
activities
Family
education
support
Internet at
home
6
M. Jamuna
[14]
GPA
Parents
working in
university
Students
discount
Transport Family size
Family
income
Parental
status
Parental
occupation
DECISION TREE
7.
R. Sumitha
[10]
Twelth
total Mark
Medium of
education
Type of
board
Engineering
cut off
Current
CGPA
No of
arrears
Attendance
percentage
8.
Tripti Mishra
[11]
Previous
marks
Loan
Type of
graduation
Year gap Location
Empathy
of student
Decision maker
Leaders
hip
ability
FUZZY
9.
Ramjeet Singh
Yadav [15]
Semester 1
marks
Semester 2
marks
- - - - -
10.
Meenakshi
[21]
Student
attendance
Internal
marks
External
marks - - - -
OPTIMIZATION TECHNIQUES
13.
Ramanathan L
[18]
Individual
details
Environme
ntal factors
Schooling
factors
Family
factors
14.
Hind Almayan
[17]
Three
Period
grades
Number of
absentees
Mother
education
Family
income
Number of
past failure
Alchol
usage
-
V. Sathya Durga and J. Thangakumar
http://www.iaeme.com/IJCIET/index.asp 53 editor@iaeme.com
2.10. Performance Measures used in prediction model
Performance of the model is assessed by a set of standard metrics. There are many metrics
used by researchers to evaluate the developed model. This section presents a table indicating
the presence or absence of nine metrics from selected research works. Presence is indicated by
a tick symbol () and absence of that metrics is indicted by a cross symbol ( ).
Table 4 Performance Metrics used in the Literary works
S.N
o
Author
Yea
r
Accurac
y
Confusio
n Matrix
F -
Measur
e
Preciso
n
Recal
l
RMS
E
True
Positiv
e
False
Positiv
e
1.
Fahad Rafiq [3] 2018
       
2.
Bendangnuksun
g [7]
2018
       
3
A.Srinivas [4] 2017
       
4.
Ihsan A. Abu
Amra [5]
2017
       
5.
F.Okubo [8] 2017
       
6.
K.B.Eashwar
[13]
2017
       
7.
M.Jamuna [14] 2017
       
8.
Huda Al-Shehri
[20]
2017
       
9.
Omar Augusto
[22]
2017
       
10.
Hind Almayan
[17]
2016
       
11.
Ramanathan.L
[18]
2016
       
12.
Karishma B.
Bhegade [9]
2016
       
13.
Raj Kumar [14] 2015
       
14.
Tribhuvan A.P
[11]
2015
       
15.
Tripti Mishra
[12]
2015
       
16.
Amirah [25] 2015
       
17.
Azwa Abdul
Aziz [6] 2014        
18.
Fahim Sikder
[19]
2014
       
20.
Ruhi Kabra [23] 2014
       
A Complete Survey on Predicting Performance of Engineering Students
http://www.iaeme.com/IJCIET/index.asp 54 editor@iaeme.com
3. INFERENCES FROM THE LITERATURE
In this section, answers to the research question framed in the introduction part can be found.
These questions are answered after a thorough analysis of the literature.
Question 1: What is the common methodology used to build student academic
prediction model.
The common methodology observed from literature for predicting students’ academic
performance is as follows.
 Collect data from Educational Institutes. Pre-process the data and form the data set. Split the
data into a training set and testing set.
 Select an appropriate technique for prediction. Using the training set of data, build the model.
 With the built model, predict result with test data. Measure the accuracy of classifiers using
standard metrics.
Question 2: What are the common techniques used for academic performance
prediction of students.
It can be inferred from the survey, that machine learning techniques are widely used for
prediction. Their simple nature and wide availability of development tools make these
techniques a preferred one by the researchers.
Question 3: What are the commonly used attributes in predicting the
performance of students.
Almost all research works use the same attributes like personal details, academic details and
details from Socio-Economic factors. It was noted that the selection of the right subsets of
attributes increased the accuracy of the prediction works radically.
Question 4: What are the common metrics used in evaluating the prediction
model.
Accuracy is the most preferred performance metrics by all researcher. Next to Accuracy,
widely used metrics by researchers were RMSE, Precision and Recall.
4. DISCUSSIONS
We wrap up by discussing briefly on how to build a better prediction model after an elaborate
literature review. To build a prediction model the following steps can be followed.
Steps for building a predictive model
 Collect the required data and preprocess it.
 Select standard Machine Learning techniques instead of resorting to hybrid approaches.
 Attributes mentioned in the attribute matrix can be used to build the model.
 Split the data into training and testing set. Built the model with the training set and predict
grades with the test data.
 The model can be evaluated with metrics like Accuracy, RMSE, Precision and Recall.
V. Sathya Durga and J. Thangakumar
http://www.iaeme.com/IJCIET/index.asp 55 editor@iaeme.com
5. CONCLUSIONS
This survey intends to compile research works on students performance prediction. The result
of the survey was elaborately presented. To Make Every Man a Success and No Man a Failure
[24] is the vision of the K. C. G. Verghese, Founder, Hindustan Institute of Technology and
Science. If every educational institute implements predictive models to assess their students
regularly, definitely we can make every student success and no student a failure.
REFERENCES
[1] Bernard Marr, How Much Data Do We Create Every Day? The Mind-Blowing Stats
Everyone Should Read, 2018.
https://www.forbes.com/sites/bernardmarr/2018/05/21/how-much-data-do-we-create-
every-day-the-mind-blowing-stats-everyone-should-read/#692eade360ba
[2] Bruce Goldman, King of the mountain - Digging data for a healthier world, 2012.
http://sm.stanford.edu/archive/stanmed/2012summer/article3.html
[3] Fahad Razaque., Nareena Soomro., Shoaib Ahmed Shaikh., Safeeullah Soomro., Javed
Ahmed Samo., Natesh Kumar. and Huma Dharejo. Using Naïve Bayes Algorithm to
Sudents' Bachelor Academic Performances Analysis. Proceedings 4th IEEE International
Conference on Engineering Technologies and Applied Sciences, Salmabad, 2017, pp. 1-5.
[4] Srinivas,A and Ramaraju,V. Detection of Failures by Analysis of Student Academic
Performance using Naïve Bayes Classifier .International Journal of Computer &
Mathematical Sciences, 7(3), 2017, pp. 277-283.
[5] Ihsan Abu Amra,A. and Ashraf Maghari, Y.A. Students Performance Prediction Using
KNN and Naïve Bayesian. Proceedings 8th International Conference on Information
Technology.Amman, 2017, pp. 412-422.
[6] Azwa Abdul Aziz, Nur Hafieza Ismail, Fadhilah Ahmad and Hasni Hassan. A Framework
for Students’ Academic Performance Analysis using Naïve Bayes Classifier. Jurnal
Teknologi, 2013, 80(5), pp. 13-19.
[7] Bendangnuksung and Prabu, P. Students' Performance Prediction Using Deep Neural
Network. International Journal of Applied Engineering Research, 13(2), 2018, pp. 1171-
1176.
[8] Okubo.,F, Yamashita.,T, Shimada.,A, and Ogata.,A. A Neural Network Approach for
Students' Performance Prediction. Proceedings Seventh International Learning Analytics
& Knowledge Conference, Canada, 2017, pp. 598-599.
[9] Karishma Bhegade, B., and Swati Shinde,V. Student Performance Prediction System with
Educational Data Mining. International Journal of Computer Applications, 146 (5), 2016,
pp. 32-35.
[10] Tribhuvan,A.P., Tribhuvan ,P.P., and Gade, J.G. 2015. Applying Naive Bayesian
Classifier For Predicting Performance Of A Student Using Weka. 7(1), 2015, pp. 239-242.
[11] Eashwar,K.B, and Venkatesan, R, Student Performance Prediction Using Svm.
International Journal of Mechanical Engineering and Technology (IJMET). 8(11), 2017,
pp .649-662.
A Complete Survey on Predicting Performance of Engineering Students
http://www.iaeme.com/IJCIET/index.asp 56 editor@iaeme.com
[12] Jamuna,M., and Shoba,S.A. Educational Data Mining & Students Performance Prediction
Using Svm Techniques. 2017.International Research Journal of Engineering and
Technology. 4(8), 2017,pp. 1248-1254.
[13] Tribhuvan, A.P., Tribhuvan, P.P., and Gade J.G. Applying Naive Bayesian Classifier For
Predicting Performance Of A Student Using Weka. 7(1),2015, pp. 239-242.
[14] Tripti Mishra, Dharminder Kumar and Sangeeta Gupta. Mining Students’ Data for
Performance Prediction. Proceedings of Fourth International Conference on Advanced
Computing & Communication Technologies, Washington, 2014, pp. 255-262.
[15] Ramjeet Singh Yadav, Soni,A.K and Saurabh Pal,A.M. Study of Academic Performance
Evaluation Using Fuzzy Logic Techniques. In Proceedings of the International
Conference on Computing for Sustainable Global Development NewDelhi,2014, pp. 48-
53.
[16] Shilpa Ingoley,N., and Bakal, J.W. Students’ Performance Evaluation Using Fuzzy
Logic. Proceedings nirma university international conference on engineering. Nuicone
2012, pp.1-3.
[17] Hind Almayan and Waheeda Al Mayyan. Improving Accuracy of Students' Final Grade
Prediction Model Using PSO . Proceedings 6th International Conference on Information
Communication and Management,Hatfield,2016, pp. 35-39.
[18] Ramanathan,L., Angelina Geetha, Khalid and Swarnalatha. Student Performance
Prediction Model Based on Lion-Wolf Neural Network. International Journal of
Intelligent Engineering & System, 10(1), 2016, pp. 114-123.
[19] Fahim Sikder,M.D., Jamal Uddin,M.d. and Sajal Halder. Predicting Students Yearly
Performance using Neural Network: A Case Study of BSMRSTU. Proceedings 5th
International Conference on Informatics, Electronics and Vision (ICIEV).2016.pp.524-
529
[20] Huda Al-Shehri, Amani Al-Qarni, Leena Al-Saati and Arwa Batoaq. Student Performance
Prediction Using SupportVector Machine and K-Nearest Neighbor. Proceedings of IEEE
30th Canadian Conference on Electrical and Computer Engineering. 2017. pp.1-4.
[21] Meenakshi and Pankaj. Application of Fuzzy Logic for Evaluation of Academic
Performance of Students of Computer Application Course, 3(10), pp.260-267.
[22] Omar Augusto Echegaray-Calderon and Dennis Barrios-Aranibar. Optimal selection of
factors using Genetic Algorithms and Neural Networks for the prediction of students’
academic performance. 2015 Latin America Congress on Computational Intelligence (LA-
CCI).pp.1-6.
[23] Ruhi Kabra,R., and Bichkar,R,S. 2017. Students’ performance prediction using genetic
algorithm. International Journal of Mechanical Engineering and Technology, 8(11) , pp.
649–662.
[24] Tribute to Our Founder, https://hindustanuniv.ac.in/founder.php
[25] Amirah Mohamed Shahiria., Wahidah Husaina, and Nur’aini Abdul Rashida. A Review
on Predicting Student’s Performance using Data Mining Techniques. Proceedings The
Third Information Systems International Conference, 2015.Malaysia. pp. 412-422.

Ijciet 10 02_007

  • 1.
    http://www.iaeme.com/IJCIET/index.asp 48 editor@iaeme.com InternationalJournal of Civil Engineering and Technology (IJCIET) Volume 10, Issue 02, February 2019, pp. 48–56, Article ID: IJCIET_10_02_007 Available online at http://www.iaeme.com/ijciet/issues.asp?JType=IJCIET&VType=10&IType=2 ISSN Print: 0976-6308 and ISSN Online: 0976-6316 ©IAEME Publication Scopus Indexed A COMPLETE SURVEY ON PREDICTING PERFORMANCE OF ENGINEERING STUDENTS V. Sathya Durga Research Scholar, Department of Computer Science, Hindustan Institute of Technology and Science, Padur, India J. Thangakumar Associate Professor, Department of Computer Science, Hindustan Institute of Technology and Science, Padur, India ABSTRACT Predictive models are quasi experimental structures used to determine the future patterns in data. These meaningful data patterns form the building block of any decision support system. Researchers all over the world have built many prediction models for major industries. Research works in the educational sector has increased steeply. This steep increase may be due to the high availability of data in the educational domain. This survey tries to comprehend a few literary works on academic performance prediction of engineering students with the focus on grade predictions. Meaningful interpretations have been made and inferences are presented at the end of this paper. Key words: Prediction, Predictive Modeling, Predictive Analytics. Cite this Article: V. Sathya Durga and J. Thangakumar, A Complete Survey on Predicting Performance of Engineering Students, International Journal of Civil Engineering and Technology (IJCIET) 10(2), 2019, pp. 48–56. http://www.iaeme.com/IJCIET/issues.asp?JType=IJCIET&VType=10&IType=2 1. INTRODUCTION Prediction is a powerful data analysis task used to determine the future behavior of a user or a system. The task of prediction has aided the decision makers to take the right decision at the right time. Predictive activities have gained wide popularity owing to the high availability of surplus data. Computers generate nearly 2.5 quintillion bytes of data every year [1]. Hidden within those mounds of data is the knowledge that could change the life of a patient, or change the world. [2] - Atual Butte, Stanford University. Making predictions from a set of data is an extremely challenging task. We need predictive models which are simple analytical structures that work on a set of predictors and forecast the probabilities of future occurrences.
  • 2.
    V. Sathya Durgaand J. Thangakumar http://www.iaeme.com/IJCIET/index.asp 49 editor@iaeme.com 1.1. Predicting Students Academic performance Predicting students performance is a significant research work in the educational sectors. Few Years back, students waited till the end of the semester to know their GPA. With the advent of machine learning, we can predict all semesters grades easily and we have models which predict the likelihood of a student being placed or not. 1.2. Four Core Components of a Predictive Model Building a predictive model is very challenging work. To Build a predictive model, we require four important components. These components are listed below for easy reference.  The Methodology followed to deploy the model.  Techniques adopted to build the prediction model.  Input attributes used by the model and  Performance Metrics used to evaluate the model. This survey tries to the unravel common methodology, techniques, attributes and metrics used in predicting the performance of engineering students. Students considered for this study are from branches of civil engineering, Mechanical, Computer Science and Information Technology. 1.3. Research Question and Objectives of the survey Every research work, be it descriptive or analytical, applied or fundamental, needs a set of research questions to work on. The research question and objective of this study is:  To find out the common methodology adopted in building a predictive model.  To find the important techniques used in predicting the performance of students.  To identify the commonly used attributes for building the prediction models.  To list out the frequently used performance metrics. 1.4. Research Methodology The methodology followed to conduct the survey is outlined below:  First popular databases like IEEE, Scopus, ACM libraries were searched and research papers were collected.  Research works older than 2013 and works with accuracy percent less than 65% were discarded and not considered for the survey.  The research papers selected was studied in detail. In-depth analysis was carried out. Many important interpretations were made and Research questions were answered.  This paper title, 'A Complete survey on predicting performance ofengineering students' was prepared and presented. 2. LITERATURE REVIEW 2.1. Performance Prediction using Naive Bayes Classification Fahad et al. (2018) present a prediction model to predict GPA of students using Naive Bayes technique and Kmeans clustering. The highest accuracy achieved in this work was 98.8%. [3]. Srinivas and Ramaraju (2017) build a prediction model using a Naïve Bayes Classifier. Dataset consisted of 257 academic records. Two feature selection technique namely
  • 3.
    A Complete Surveyon Predicting Performance of Engineering Students http://www.iaeme.com/IJCIET/index.asp 50 editor@iaeme.com cfsSubsetEval and GainRatioAttributeEval was used. Results reveal cfsSubsetEval perform better with 84% accuracy. [4] Figure 1 Accuracy of Naive Bayes Classification 2.2. Performance Prediction using Artificial Neural Network (ANN) Bendangnuksung and Prabu (2018) implement Neural Network to predict students’ performance. Cost Function with Accuracy were the metrics used in this research work.. Their Model attained 83.4 % accuracy [7]. Okbu et.al develop a prediction model based on Recurrent Neural Network (RNN) . Grades were predicted from log data. Results reveals that RNN works well with an accuracy of 90% [8]. 2.4. Performance Prediction using Decision Tree Karishma and Swati (2016) use ID3 and C4.5 algorithm to identify students who are probable to fail. Experiments were done on 200 Engineering student’s dataset. Applying the Decision Tree on students internal marks, their performance was predicted. The Accuracy of the classifiers was calculated to be 98.5 % [9]. Sumitha and Vinothkumar (2016) present a comparative study of classification techniques to predict academic performance. This research work experiments data of 350 students. Out of the classifiers, it was proved J48 was highly influential with 97% accuracy[10] . 2.3. Performance Performance Prediction using Support Vector Machine (SVM) Eashwar and Venkatesan (2017) develop a prediction model using Support Vector Machine to recognize postgraduate students for carrying out doctoral studies. Data were clustered and SVM technique was applied. 96.7% accuracy was achieved [13]. Jamuna and Shoba (2017) compare various machine learning algorithms. Students data from a private institute was taken. Results unravel the Polynomial Kernel achieved 97.62 % accuracy [14]. 98.8 84.23 93.17 68.5 0 20 40 60 80 100 Fahad Razaque 2018 [3] Srinivas A 2017 [4] Ihsan A. Abu Amra 2017 [5] Azwa Abdual Aziz 2014 [6] AccuracyoftheWork Author Name and Year of Publication Accuracy of Naive Bayes Classification Works Accuracy
  • 4.
    V. Sathya Durgaand J. Thangakumar http://www.iaeme.com/IJCIET/index.asp 51 editor@iaeme.com Figure 2 Accuracy of Decision Tree 2.5. Performance Performance Prediction using fuzzy system Ramjeet, et al. propose a new fuzzy expert system to evaluate the performance of students. First semester exam results and output values are fuzzified. Application rules were determined, performance value was defuzzified and final marks were calculated [15]. Shilpa and Bakal use three nodes fuzzy logic system in their research work. The Triangular membership function was used in this research work. For Defuzzification Center of gravity is used. Scores obtained was calculated with the score matrix and accuracy matrix [16]. Table 1 Fuzzy Functions used in Literary works S.No Author Fuzzification Rule Inference De fuzzificatin 1. Ramjeet Singh Yadav [19] Triangular Membership function Mamdani max–min inference Center of area 2. Shilpa N [20] Triangular Membership function Mamdani max–min inference Ceter of gravity 2.6. Performance Prediction using optimization technique Hind Almayan and Waheeda Al Mayyan (2016) use Swarm Optimization method to build their prediction model. Using Dimensionality reduction technique, feature space was reduced. PSO is used for feature selection. C4.5 Classifier achieves 93.5% accuracy [17]. Ramanathan et al. (2016) implements a Neural Network with a Lion – Wolf optimization technique for weight selection. The model predicts one to eight semester exam performance of college students. The trained network performances well compared to other prediction models with RMSE - 2.3 [18]. 98.5 97 92 94.41 0 20 40 60 80 100 Karishma B. Bhegade 2016 [9] R. Sumitha 2016 [10] Tribhuvan A.P 2015 [13] Tripti Mishra 2014 [14] Accuracyofthework Author Name and Year of Publication Accuracy of Decision Tree Works Accuracy
  • 5.
    A Complete Surveyon Predicting Performance of Engineering Students http://www.iaeme.com/IJCIET/index.asp 52 editor@iaeme.com Table 2 Accuracy Comparison of Hybrid ANN 2.9. Attributes used in prediction model This section lists the commonly used attributes in students performance prediction works. Total of eight attributes is listed in the following table. The most commonly occurring attributes are identified and presented in the inferences. section. Table 3 Attributes used in the Literary Works S.N o Author Model Accur acy RMSE 1. Md. Fahim Sikder [17] ANN 99.8 % 0.1765 2. Ramanathan L [18] Hybrid - ANN 94 % 2.3 S.No Author 1 2 3 4 5 6 7 8 Naive Byaes 1. A. Srinivas [4] Previous marks High School grades Senior secondrary grades Family annual income Medium of instruction General proficiency Fathers' education Mothers' education 2. Ihsan A. Abu Amra [5] Previous marks Gender Date of birth Students branch Location School name Marital status Father's job Neural Network 3. Bendangnuksu ng [7] Previous marks Nationality Stages Parents responsible Raised hands Visited resource Parent answers Views updates 4. Md. Fahim Sikder [19] Previous semester marks Class Test Marks Class attendance Lab performance Study time Family education Living area Extra curicular activity SVM 5. Huda Al- Shehri [20] Previous marks Family type Parents' status Parent's job Extra paid classed Extra curricular activities Family education support Internet at home 6 M. Jamuna [14] GPA Parents working in university Students discount Transport Family size Family income Parental status Parental occupation DECISION TREE 7. R. Sumitha [10] Twelth total Mark Medium of education Type of board Engineering cut off Current CGPA No of arrears Attendance percentage 8. Tripti Mishra [11] Previous marks Loan Type of graduation Year gap Location Empathy of student Decision maker Leaders hip ability FUZZY 9. Ramjeet Singh Yadav [15] Semester 1 marks Semester 2 marks - - - - - 10. Meenakshi [21] Student attendance Internal marks External marks - - - - OPTIMIZATION TECHNIQUES 13. Ramanathan L [18] Individual details Environme ntal factors Schooling factors Family factors 14. Hind Almayan [17] Three Period grades Number of absentees Mother education Family income Number of past failure Alchol usage -
  • 6.
    V. Sathya Durgaand J. Thangakumar http://www.iaeme.com/IJCIET/index.asp 53 editor@iaeme.com 2.10. Performance Measures used in prediction model Performance of the model is assessed by a set of standard metrics. There are many metrics used by researchers to evaluate the developed model. This section presents a table indicating the presence or absence of nine metrics from selected research works. Presence is indicated by a tick symbol () and absence of that metrics is indicted by a cross symbol ( ). Table 4 Performance Metrics used in the Literary works S.N o Author Yea r Accurac y Confusio n Matrix F - Measur e Preciso n Recal l RMS E True Positiv e False Positiv e 1. Fahad Rafiq [3] 2018         2. Bendangnuksun g [7] 2018         3 A.Srinivas [4] 2017         4. Ihsan A. Abu Amra [5] 2017         5. F.Okubo [8] 2017         6. K.B.Eashwar [13] 2017         7. M.Jamuna [14] 2017         8. Huda Al-Shehri [20] 2017         9. Omar Augusto [22] 2017         10. Hind Almayan [17] 2016         11. Ramanathan.L [18] 2016         12. Karishma B. Bhegade [9] 2016         13. Raj Kumar [14] 2015         14. Tribhuvan A.P [11] 2015         15. Tripti Mishra [12] 2015         16. Amirah [25] 2015         17. Azwa Abdul Aziz [6] 2014         18. Fahim Sikder [19] 2014         20. Ruhi Kabra [23] 2014        
  • 7.
    A Complete Surveyon Predicting Performance of Engineering Students http://www.iaeme.com/IJCIET/index.asp 54 editor@iaeme.com 3. INFERENCES FROM THE LITERATURE In this section, answers to the research question framed in the introduction part can be found. These questions are answered after a thorough analysis of the literature. Question 1: What is the common methodology used to build student academic prediction model. The common methodology observed from literature for predicting students’ academic performance is as follows.  Collect data from Educational Institutes. Pre-process the data and form the data set. Split the data into a training set and testing set.  Select an appropriate technique for prediction. Using the training set of data, build the model.  With the built model, predict result with test data. Measure the accuracy of classifiers using standard metrics. Question 2: What are the common techniques used for academic performance prediction of students. It can be inferred from the survey, that machine learning techniques are widely used for prediction. Their simple nature and wide availability of development tools make these techniques a preferred one by the researchers. Question 3: What are the commonly used attributes in predicting the performance of students. Almost all research works use the same attributes like personal details, academic details and details from Socio-Economic factors. It was noted that the selection of the right subsets of attributes increased the accuracy of the prediction works radically. Question 4: What are the common metrics used in evaluating the prediction model. Accuracy is the most preferred performance metrics by all researcher. Next to Accuracy, widely used metrics by researchers were RMSE, Precision and Recall. 4. DISCUSSIONS We wrap up by discussing briefly on how to build a better prediction model after an elaborate literature review. To build a prediction model the following steps can be followed. Steps for building a predictive model  Collect the required data and preprocess it.  Select standard Machine Learning techniques instead of resorting to hybrid approaches.  Attributes mentioned in the attribute matrix can be used to build the model.  Split the data into training and testing set. Built the model with the training set and predict grades with the test data.  The model can be evaluated with metrics like Accuracy, RMSE, Precision and Recall.
  • 8.
    V. Sathya Durgaand J. Thangakumar http://www.iaeme.com/IJCIET/index.asp 55 editor@iaeme.com 5. CONCLUSIONS This survey intends to compile research works on students performance prediction. The result of the survey was elaborately presented. To Make Every Man a Success and No Man a Failure [24] is the vision of the K. C. G. Verghese, Founder, Hindustan Institute of Technology and Science. If every educational institute implements predictive models to assess their students regularly, definitely we can make every student success and no student a failure. REFERENCES [1] Bernard Marr, How Much Data Do We Create Every Day? The Mind-Blowing Stats Everyone Should Read, 2018. https://www.forbes.com/sites/bernardmarr/2018/05/21/how-much-data-do-we-create- every-day-the-mind-blowing-stats-everyone-should-read/#692eade360ba [2] Bruce Goldman, King of the mountain - Digging data for a healthier world, 2012. http://sm.stanford.edu/archive/stanmed/2012summer/article3.html [3] Fahad Razaque., Nareena Soomro., Shoaib Ahmed Shaikh., Safeeullah Soomro., Javed Ahmed Samo., Natesh Kumar. and Huma Dharejo. Using Naïve Bayes Algorithm to Sudents' Bachelor Academic Performances Analysis. Proceedings 4th IEEE International Conference on Engineering Technologies and Applied Sciences, Salmabad, 2017, pp. 1-5. [4] Srinivas,A and Ramaraju,V. Detection of Failures by Analysis of Student Academic Performance using Naïve Bayes Classifier .International Journal of Computer & Mathematical Sciences, 7(3), 2017, pp. 277-283. [5] Ihsan Abu Amra,A. and Ashraf Maghari, Y.A. Students Performance Prediction Using KNN and Naïve Bayesian. Proceedings 8th International Conference on Information Technology.Amman, 2017, pp. 412-422. [6] Azwa Abdul Aziz, Nur Hafieza Ismail, Fadhilah Ahmad and Hasni Hassan. A Framework for Students’ Academic Performance Analysis using Naïve Bayes Classifier. Jurnal Teknologi, 2013, 80(5), pp. 13-19. [7] Bendangnuksung and Prabu, P. Students' Performance Prediction Using Deep Neural Network. International Journal of Applied Engineering Research, 13(2), 2018, pp. 1171- 1176. [8] Okubo.,F, Yamashita.,T, Shimada.,A, and Ogata.,A. A Neural Network Approach for Students' Performance Prediction. Proceedings Seventh International Learning Analytics & Knowledge Conference, Canada, 2017, pp. 598-599. [9] Karishma Bhegade, B., and Swati Shinde,V. Student Performance Prediction System with Educational Data Mining. International Journal of Computer Applications, 146 (5), 2016, pp. 32-35. [10] Tribhuvan,A.P., Tribhuvan ,P.P., and Gade, J.G. 2015. Applying Naive Bayesian Classifier For Predicting Performance Of A Student Using Weka. 7(1), 2015, pp. 239-242. [11] Eashwar,K.B, and Venkatesan, R, Student Performance Prediction Using Svm. International Journal of Mechanical Engineering and Technology (IJMET). 8(11), 2017, pp .649-662.
  • 9.
    A Complete Surveyon Predicting Performance of Engineering Students http://www.iaeme.com/IJCIET/index.asp 56 editor@iaeme.com [12] Jamuna,M., and Shoba,S.A. Educational Data Mining & Students Performance Prediction Using Svm Techniques. 2017.International Research Journal of Engineering and Technology. 4(8), 2017,pp. 1248-1254. [13] Tribhuvan, A.P., Tribhuvan, P.P., and Gade J.G. Applying Naive Bayesian Classifier For Predicting Performance Of A Student Using Weka. 7(1),2015, pp. 239-242. [14] Tripti Mishra, Dharminder Kumar and Sangeeta Gupta. Mining Students’ Data for Performance Prediction. Proceedings of Fourth International Conference on Advanced Computing & Communication Technologies, Washington, 2014, pp. 255-262. [15] Ramjeet Singh Yadav, Soni,A.K and Saurabh Pal,A.M. Study of Academic Performance Evaluation Using Fuzzy Logic Techniques. In Proceedings of the International Conference on Computing for Sustainable Global Development NewDelhi,2014, pp. 48- 53. [16] Shilpa Ingoley,N., and Bakal, J.W. Students’ Performance Evaluation Using Fuzzy Logic. Proceedings nirma university international conference on engineering. Nuicone 2012, pp.1-3. [17] Hind Almayan and Waheeda Al Mayyan. Improving Accuracy of Students' Final Grade Prediction Model Using PSO . Proceedings 6th International Conference on Information Communication and Management,Hatfield,2016, pp. 35-39. [18] Ramanathan,L., Angelina Geetha, Khalid and Swarnalatha. Student Performance Prediction Model Based on Lion-Wolf Neural Network. International Journal of Intelligent Engineering & System, 10(1), 2016, pp. 114-123. [19] Fahim Sikder,M.D., Jamal Uddin,M.d. and Sajal Halder. Predicting Students Yearly Performance using Neural Network: A Case Study of BSMRSTU. Proceedings 5th International Conference on Informatics, Electronics and Vision (ICIEV).2016.pp.524- 529 [20] Huda Al-Shehri, Amani Al-Qarni, Leena Al-Saati and Arwa Batoaq. Student Performance Prediction Using SupportVector Machine and K-Nearest Neighbor. Proceedings of IEEE 30th Canadian Conference on Electrical and Computer Engineering. 2017. pp.1-4. [21] Meenakshi and Pankaj. Application of Fuzzy Logic for Evaluation of Academic Performance of Students of Computer Application Course, 3(10), pp.260-267. [22] Omar Augusto Echegaray-Calderon and Dennis Barrios-Aranibar. Optimal selection of factors using Genetic Algorithms and Neural Networks for the prediction of students’ academic performance. 2015 Latin America Congress on Computational Intelligence (LA- CCI).pp.1-6. [23] Ruhi Kabra,R., and Bichkar,R,S. 2017. Students’ performance prediction using genetic algorithm. International Journal of Mechanical Engineering and Technology, 8(11) , pp. 649–662. [24] Tribute to Our Founder, https://hindustanuniv.ac.in/founder.php [25] Amirah Mohamed Shahiria., Wahidah Husaina, and Nur’aini Abdul Rashida. A Review on Predicting Student’s Performance using Data Mining Techniques. Proceedings The Third Information Systems International Conference, 2015.Malaysia. pp. 412-422.