IRJET- Predictive Analytics for Placement of Student- A Comparative Study

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 06 | June 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 3423
Predictive Analytics for Placement of Student- A Comparative Study
Sonali Rawat
Assistant Professor, Department of Computer Science Chandigarh University Mohali, Punjabi
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Predictive Analytics is science of withdrawing information from the existing data for the purpose of determining
existing patterns and future trends. Scope of this review paper is to study different data mining techniques that can be used to
analyze the academic performance of students to predict their chances of getting placed through campus placement. Different
attributes such as academic results, technical skills, training and projects done, are considered to be desirable for prediction
purpose. This paper presents an outlook to differentiate data mining techniques for predictive analytics that can be used in the
process of predicting campus placement.
Key Words: Predictive analytics, Data Mining, Classification, Decision Tree
1. INTRODUCTION
Predictive Analytics presently has many applications includingbanking,health,social media,retail etc.Duetoitspositive result
new areas are also emerging to use it, one such area is “EDUCATION”. Predictive Analytics canbeusedinEducational Institute
to predict placements of students as employability has become one of the crucial businessesinthepresentworld.Monumental
amount of students takes admission in professional college with the hope of acquiring their dream job. So, it would be a great
deal if institute as well as student can get idea of placement beforehand .The result of prediction is informed to the studentsso
that they can upgrade and refine their skill in order to increase the graph of recruitment.
2. Literature Review
Classification Approach: - Classification is used to classify each dataitemintooneofthepredictedtargetclassorgroupandto
accurately predict categorical labels. Classificationusesclassificationmodelstopredicttheclasslabel. Usinga setofpredefined
classes, class label of each object is determined. Training set is provided as an input to algorithm to build model, which can be
used for classification of new object. For example, a bank starts credit policy for his customers; manager by the behavior of
customer can classify them under three categories: “safe”, “risky”, “very risky”. So classification will help us to draw a model
that could be used to accept or reject future request for the credits.
Sudheep Elayidom et al. [1] aim to use data mining technique for the benefit of students in future. They uses different
techniques like decision tree, naive bayes and artificial neural networks and declare four class labels excellent, good, average
and poor for each branch. A student needs to enter his entrance rank, gender (M/F), sector (rural/urban) and reservation
category, and then using data mining techniques, he or she may know which branch is suitable for him or her. Then with the
help of above information a student enters his branch, locationetc.andontherootofwhichtheplacementchancesfordifferent
streams of study is calculated. Hence student may opt for thebranchprovidingchancesofexcellentplacement.Attheendof the
paper the three techniques are compared and it shows that decision tree is slightly good in terms of accuracy however the
difference is unimportant. And therefore there is no universally accepted best model.
Tripti Mishra et al. [2] provides classificationtechniquelikeBayesianmethod,multilayerperceptronsandsequential minimal
optimization (SMO), Ensemble methods, decision tress using WEKA and emotional skill like assertion, empathy, decision
making, leadership and stress management to predict placement of MCA students. ROC curve and F measure are used to
compare these algorithms. Emotional skill parametersareassessedthroughEmotional skill assessmentprocess(ESAP)tool.All
the models are compared and J48 is suggested as the best technique among all with the best accuracy and least time to build.
T.Jeevalatha et al. [3] used decision tree algorithm like ID3, CHAID, and C4.5 using Rapid Miner tool to predict student
performance in placement activity using two year student records.All threealgorithms arecomparedandtheresultshowsthat
ID3 has highest accuracy rate of 95.33 percent.
Neelam Naik et al. [4] provides prediction to gain the knowledge about students of Master of Application before admitting
them to the course. Sample of around 325 students is taken among which 195 records are used to create model for prediction
of result and placement and remaining are used for validation purpose. Data mining algorithm are applies using XL minertool
and a classification tree is developed, on the basis of which decisionrulesaremade.TheserulesareimplementedusingASP.net

software. Error experienced during the experiment inthe paperis38.46%forMCAresultpredictionand45.38%forplacement
prediction. A keen observation is made in the paper that classification tree do not take gender as a trait for taking decision.
Pratiyush Guleria et al. [5] uses application of Bayesian theorem to predict the result of placements. The traits used are
attendance, GPA, reasoning, quantitative, communication, and technical skill. Two tools are used for the classification of the
record – WEKA & RAPID MINER. At last the result is represented with the help of a graph which shows that student with good
technical skill have comparatively more chances of getting placed.
Prof. Savita Bakare et al. [6] defines education data mining as the use of data mining technique in the process of education.
Database of students is analyzed and studied using mining technique to predict whether student enlists during the campus
placement. The attributes taken for prediction include academics as well as co-curricular activitiesalongwithcommunication
skills etc. “Fuzzy logic” and “K nearest Neighbor (KNN) are used for building model. Data of900studentsistakenamong which
600 are used as training set for model building and remaining as test data for validating themodel.Atthe end result ofboththe
method is calculated and observed that fuzzy has accuracy of 92.67% with running time of 450msec and KNN has accuracy of
97.33% with running time of 13458msec. So KNN is concluded as better method.
Revathy S et al. [7] focuses on the importance of Education data mining (EDM) technique which can be used to traverse
concealed information of students from their resumes. The paper uses data mining techniques to get an idea of students
composing for the coming placement activity. A reliable framework is designed to locate students to be placed in whole
database. Classification technique is used to categorize student according to their academicdocumentation. Thespecifications
used for categorization are academic detail, technical skill, programming skill, quantitative and reasoning skill. To forecast
about the company student is likely to be placed. C5.0 algorithm is used for classification, which result in decision tree
formation using Quinlan. The prediction is done using R, where data is divided into two parts one is training data other is test
data. The output predicted is displayed using a pie chart and accuracy of 75% is observed.
Namita puri et al. [8] intent to design a model for the prediction of students placementbyidentifyingappropriatetraitsbased
on academics skills and curricular of final year student. The model is designed using a classification technique based on
decision tree. The paper suggests use of ID3 algorithm for the implementation.Theresultstipulatesaccuracyofalgorithm with
95%.
Mosima Anna Masethe et al. [9] notifies that large databases of educational institutes can be studied with the help of data
mining techniques to discover hidden pattern in it in order to use them for the purpose of decision making. Various
classification algorithms like j48, simple cart, bayes net, naïve bayes and reptree algorithm are applied to database so as to
predict placement in work integrated learning.Confusion matrixforall algorithmsisformedandwasobservedthat naïvebayes
and bayes net gave best result.
Ravi Kumar Rathore et al. [10] targeted to use fuzzy inference system from Matlab tool box in order to differentiate the
information of student based on their performance. Data of 31 students from M-tech CS including X th marks, XII th marks, B-
tech cgpa, M-tech cgpa, and data of backlog is collected, then membershipfunctionisassignfor eachinputset whichshowsthat
student with membership function greater than 2.1 is eligible for the placement process.
Ajay Shiv Sharma et al. [12] develops a predictor system for placement of student using logistic regression. The score in
matriculation, senior secondary, subject in various semesteroftechnical education&demographicofGNDEC’sstudentistaken
as input by the system. For the generation of result programming tool GNU octave & for optimization of classifier gradient
descent algorithm, is used. It was revealed that 91% of total placement is from urban background, where gender played an
important role as 59.09% of females are placed.
Ajay Kumar pal et al. [13] collected data of 65 students from MCA and evaluated using classification algorithm like Naïve
Bayesian, C4.5 tree, and multilayer perception; for the prediction of training and placement. The attribute chosen from the
database for evaluation include sex, MCA result, seminar performance, lab work, communication skill, and graduation
background. Attribute assessment is done using chi-square test, information-gain and gain-ratio test. Average of these
assessments is taken for each attribute and it was observed that sex has most impact on the output. Naïve Bayes classifier has
highest accuracy rate of 86.15% with 0 time to build and lowest error 0.28.
Ramanathan.L et al. [14] uses EDM to predict the placement of final year student by using traits such as gender, category,
academic gap, 10th -12th result, grade in B-Tech, communication skill, technical skill and grade in M-Tech. Similarity measure
“Sum of Difference” is implemented on C# to analyze the pattern of data set and carry out prediction process. As a result of
experiment a graph is obtained which shows that if value on y-axis goes higher than 4, it designate “yes” value of placement.

V.Ramesh et al. [15] scrutinize various data mining technique in order to use them for thepurposeofplacement prediction.A
dataset of 300 students from computer science is taken and cleaned, then suitable attribute like English marks, Math marks,
programming language, lab marks are chosen on which technique like Naïve bayes, SMO, j48, multilayer perception, Rep tree
algorithms are applied. Data set is applied on Weka tool and it is observed thatmultilayerperceptionshowsanaccuracyrateof
87.395%.
Vikas Chirumamilla et al. [16] uses CGPA, backlog, aptitude, technical articulatecore skill level &achievementsasevaluation
attributes, which are preprocessed and cleaned before applying naïve bayes and C4.5. Fortheanalysisrapidminer tool isused.
Using confusion matrix it is observed that Naïve bayes has accuracy of 77.78% and C4.5 has 88.89 %.
Ravina Sangha et al. [17] aims to develop a model which can extract informationfrom thedatabaseof educational institution.
Rule based classification using fuzzy is applied on the datasetofMESCOEpune.There areseveral advantagesofusingfuzzyasit
gives various prospects rather than a single value, with its high efficiency in handling erroneous input.
G.Vadivu et al. [19] uses a database of educational institute containing large amount of student data which can be used to
forecast student performance in forthcoming semester and likelihood to get placed in coming recruitment session. For this
purpose two algorithm KNN and naïve bayes are used. Dataset of 250 students with 59 attribute is taken for experiment
purpose & accuracy of 95.33 was observed in case of KNN & 97.67% in case of Naïve bayes.
Dammalapati Rama Krishna et al. [20] aimed to pinpoint whether thestudent will be placedornot.Inordertoimprovetheir
performance, improved decision tree algorithm is used. Attribute include city, communication skill, technical skill, grade,
attitude, economical background, written test. Paper also includes privacy preserving approach. Improved C4.5 decision tree
based algorithm is applied on the eligible students as it can handle both missing data and continuous data.
Clustering Approach: - Similar to classification it sorts data into classes; however unlike classification here thelabelsof class
are not defined; therefore known as unsupervised classification. Clustering is based on maximum resemblanceamongobjects
in same class and minimum among different class. It is not specific to single algorithm and can be achieved through many
algorithms depending upon the scenario where training set is not needed. Clustering is also used for exploring data as it finds
natural grouping and because of this reason it is a useful data preprocessing step as it identify homogeneous groups.
Dr Rajan vohra et al. [11] aims to find the eligible candidate for the placement by segregating the studentsintothreeclusters
i.e. group 1 contain student who will face difficulty in completing their degree, there score range is 0.55-0.6, count of such
student is 170. Group 2 contain above average student they are eligible for placement, score range 0.7-0.8; count of such
student is 230. Last group contain 200 students with a score range of 0.625-0.675 this group need to improve a bit to become
eligible. The data related to academic detail, technical skill, project detail,trainingandsoftskill actasinputforweka tool where
K-mean clustering & j-48 algorithm are applied. Here classification algorithm is used to predict the cluster for pre-final year
student which results in some errors.
Karan Pruthi et al. [18] uses data mining techniques to extract information of student from previously recorded data, which
can be used for prediction purpose. The paper not only tells about placement but also whether the student gets placed in Core
Company or consultancy, as well as predictcompany’sname.Aftertheexperimentitwasobservedthatclustering,classification
and naïve bayes has accuracy of 95.52%, 82.39%, 62.3% respectively while predicting company type , also classification and
naïve bayes has accuracy of 62.1%, 45.78% respectively in case of predicting company’s name.
3. A Comparative Study of Papers
Sr
No.
Papers name
(year)
Method/Algorithm Best Method/Method :
Accuracy
1 Paper 1
(2011)
1.Decision Tree
2.Naive bayes
3.Artificial neural network
Decision Tree
2 Paper 2
(2016)
1.Bayesian method
2.Multilayer perceptrons
3.Sequential minimal optimization
4.Decision Tree
Decision Tree (j48)

3 Paper 3
(2014)
1.ID3
2.CHAID
3.C4.5
ID3: 95.33%
4 Paper 4
(2017)
1.Classification
5 Paper 5
(2015)
1.Bayesian theorem
6 Paper 6
(2016)
1.Fuzzy logic
2.K nearest neighbor
Fuzzy logic: 92.67%
KNN : 97.33%
7 Paper 7
(2017)
Classification(C5.0) C5.0 : 75%
8 Paper8
(2015)
Decision Tree Decision Tree(ID3):95%
9 Paper 9
(2014)
1.J48
2.Simple cart
3.Bayes net
4Rep tree algorithm
5.Naive bayes
Naïve bayes
Bayes net
10 Paper 10 Fuzzy inference system
11 Paper 11
(2015)
1.K-mean 2.clustering
J-48
12 Paper 12
(2014)
1.GNU octave
2.Gradient descent algorithm
13 Paper 13
(2013)
1.Naive Bayesian
2.C4.5
3.Multilayer perception
Naïve Bayesian:86.15%
14 Paper 14
(2014)
Sum of difference
15 Paper 15
(2011)
1.Naive Bayes
2.SMO
3.j48
4.Multilayer perception
5.Rep tree algorithm
Multilayer perception:87.395%
16 Paper 16
(2014)
1.C4.5
2.Naive Bayes
Naïve bayes: 77.78%
C4.5: 88.89%
17 Paper 17 Rule based classification using
fuzzy

(2016)
18 Paper 18
(2015)
1.Clustering
2.Classification
3.Naive bayes
Company
type
Company
name
Clustering
:95.52%
Clustering:
55%
Classification
:82.93%
Classification:
62.1%
Naive
bayes:62.3%
Naïve
Bayes
:45.78%
19 Paper 19
(2017)
1.K nearest neighbor
2.Naive bayes
KNN: 95.33%
Naïve bayes: 97.67%
20 Paper 20
(2014)
Improved C4.5 decision tree
 Techniques used in papers
SR Number Technique Paper used
1 Decision Tree Paper 1,2,8,20
2 Naïve Bayes Paper 1
3 Artificial Network Paper 1, 5, 15, 16, 18, 19
4 Bayesian Method Paper 2, 5, 13
5 Multilayer Perception Paper 2, 13
6 Sequential min optimal Paper 2
7 ID3 Paper 3
8 CHAID Paper 3
9 C4.5 Paper 3, 13, 16, 20
10 Fuzzy Logic Paper 17, 10, 6
11 K nearest neighbor Paper 6
12 C5.0 Paper 7
13 J48 Paper 9, 11
14 Rep Tree Algorithm Paper 9
15 K mean Paper 11
16 Gradient Descent Algorithm Paper 12
17 Sum of difference Paper 14
18 Clustering Paper 20

Table -1: Sample Table format
Preparation of Manuscript
Margins : Top 0.5” Bottom 0.5”
Left 0.5” Right 0.5”
Margin : Narrow Font Cambria / 10 pt
Title of paper : 16 Point Heading 13 Point
Sub Heading :12 Point Spacing Single line spacing
4. CONCLUSION
Various classification and clustering techniques are inspected to evaluate the performance of students in the recruitment
procedure. Using the comparative study amongst these techniques ID3 with accuracy 95.33%, KNN with 97.33%, C4.5 with
88.89%, Naïve bayes with 86.15 %, Multilayer perception having 87.395% accuracy is suggested to be the best one. If
placement cell conduct workshop in advance for pre-final year to train them according to the result of these techniques,
students will be able to identify their scope of improvement and will be able to refine themselves accordingly.
5. Future Scope
It would of great help if we revise and update our curriculum and other extra activities for each semester in accordance with
the public, private and government sector requirement. We can also predict which company picks which categoryofstudents.
Make a list of skill a particular company looking for, then on the basis of that we can train our student. These traits will make
prediction process more accurate.
REFERENCES
1. Sudheep Elayidom, Summan Mary Idikkula, Joseph Alexander, ” A generalized data mining framework for placement
chance prediction problems” International journel of computer applications(0975-8887) volume 31- No.3,October
2011.
2. Tripti Mishra, Dharminder Kumar, Sangeeta Gupta, “Students’ employability prediction model through data mining”
International journal of applied engineering research ISS0973-4562 volume 11, November 4 2016.
3. T.jeevalatha, N.Ananthi, D.Saravana, “Performance analysis of undergraduate students placement selection using
decision tree algorithms” International journal of computer application (0975-8887) volume 109-No.15, December
2014.
4. Neelam Naik, Seema Purohit, “Prediction of final result and placement of students using classification algorithm”
International journal of computer application (0975-8887) volume 56-No.12, October 2017.
5. Pratiyush Guleria, Manu Sood, “Predicting student placement using Bayesian classification” Third international
conference on image processing 978-5090-0148, 2015 IEEE.
6. Mangasuli Sheetal B, Prof. Savita Bakare, “Predictionofcampusplacementusingdata miningalgorithm-fuzzylogicand
K nearest neighbor” International Journal of advance research in computer and communication engineering (2278-
1021) vol-5, June 2016.
7. Revathy S, Roopika G, Rishitha R, Revathy P, “An approach to suggest companyspecificplacementopportunitiesusing
data mining techniques” IJCSMC (2320-088X) vol-6, March 2017.
8. Namita Puri, Deepali Khot, Pratiksha Shinde, Kishori Bhoite, Prof Deepali Maste “Student placement prediction using
ID3 algorithm” IJRASET (2321-9653) vol-3,March 2015.
9. Mosima Anna Masethe, Hlaudi Daniel Masethe, “Prediction of work integrated learning placement using data mining
algorithm” WCECS volume 1 22-24 October 2014.
10. Ravi Kumar Rathore, J.Jayanth, “Student predictionsystemforplacementtrainingusingfuzzyinferencesystem”ICTAT
journal on soft computing (2229-6959) volume 07

11. Praveen Ram, Dr. Rajan Vohra, “Generating placement intelligence in higher education using data mining”
International journal of computer science and information technologies (0975-9646) volume 6 2015.
12. Ajay Shiv Sharma, Swaraj Prince, Shubham Kapoor, Keshav Kumar, “PPS- placement prediction system using logistic
regression” International conference on MOOC, Innovation and technologyineducation(978-4799-6876)IEEE2014.
13. Ajay Kumar Pal, Saurabh Pal, “Classification model of prediction for placement of students” Modern education and
computer science (49-56) November 2013.
14. Ramanathan.L, Swamalatha P, D. Ganesh Gopal, “ Mining educational data for students’ placement prediction using
sum of difference method” International journal of computer application (0975-8887) volume 99 18, August 2014.
15. V. Ramesh, P.Parkavi, P.Yosodha, “Performance analysis of data mining techniques for placement chance prediction”
International journal of scientific & engineering research (2229-5518) volume 2 , 8 August 2011.
16. Vikas Chirumamilla, Bhagya Sruthi T, Sasidhar Velpula, Indira Sunkara, “A novel approach to predict student
placement chance with decision tree induction” International journal of system and Technologies (ISSN0974-2107)
volume 7 , Issue 1,2014, pp 78-88.
17. Ravina Sangha, Akshay Satras, Lisha Swamy, Gopal Deshmukh, “Student’s placement eligibilitypredictionusingfuzzy
approach” International journal of engineering and techniques(ISSN:2395-1303)-volume 2, November-December
2016 pp 148-152.
18. Getaneh Berie Tarekegan, Dr. Vuda Sreenivasarao “Application of data mining in predicting placement of students”
International Journel of Research Studies in Computer Science and Engineering (IJRSCE) (ISSN 2349-4840(print) &
ISSN 2349-4859(Online)) Volume 3, Issue 2, 2016, pp 10-14
19. G.Vadivu, K.sornalakshmi, “Applying machine learning algorithms for student employability prediction using R”
International journal of pharmaceutical sciences review and research (ISN 0976-044X) March-April 2017,ArticleNo.
11, pages 38-41.
20. Krishna, Dammalapati Rama, Bode Prasad, and Teki Satyanarayana Murthy. "Placement Prediction Analysis in
University Using Improved Decision Tree Based Algorithm."

IRJET- Predictive Analytics for Placement of Student- A Comparative Study

More Related Content

What's hot (18)

Similar to IRJET- Predictive Analytics for Placement of Student- A Comparative Study (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Predictive Analytics for Placement of Student- A Comparative Study