Abstract. of graduate the Nation hinders wil performanc divided int methodology Making pr C & R Tr several class The outcom the univers
Keyword
1. In
1.1. Ba
Accordi No. 12 educatio the Univ who bec science interests competi graduati Universi and the of succe impact o of hig Accredit (governm require program commitm impleme (educati service) required Th
Predi
. Maintaining st tes is a success ind nal Accreditation
ll be a valuable ce and prediction to 4 segments b
y in data mining redictions, gradua ree, CHAID, an ses. Parameters u me of the study is sity may recomme
ds: graduation, se
ntroduction
ackgrounds
ing to Indon 2/2012 th on, the regu
versity goal came master
and / or tec s and i itiveness. ion rates a
ity. High ra quality of ess of a uni on public c gher educat
tation Bo ment). A Universitie ms should
ment to entation of
on, research in order to d to be a gra
he Asian Journ
icting Suc
Awan Faculty of E
tudent graduation dicator of a univ n Board as a regu input for Unive n of their gradua ased on grade p g i.e CRISP-DM ation can be done nd Logistic Regre used in addition s the student categ end actions to be
egmentation, qua
n
nesian Coun at concern ulations said ls is to prod r or specialis chnology to increase
Maintainin are the m tes of stude
graduates is iversity, whic confidence a tion and oard as Accreditation es undergr have a the qual f the acad h, and servic provide the aduate stude
nal of Technol
mail.com
ccess Stud
n Setiawan* Engineering,
n rates are the m versity, which wil ulator (governme ersity. Data min ation. Segmentati point average an M (Cross Industry
e through the mo ession tested in o to GPA also inc egory based on th taken to improv
adrant GPA, dat
ncil Regulati n on high d that one duce gradua st in branch o meet natio the natio ng stud main tasks
ent graduati s an indicati ch will have as stakehold the Natio a regula n standa raduate stu focus and lity of demic proc ce/ commun e competenc ent who is a
logy Managem
dy Using
and Kuntjah Langlangbua
main tasks of a ll have an impact
nt). Making pre ning system facili tion of student by nd the growth r ry Standard Proc odeling process by rder to find the b cluded the master heir study perform ve the student ac
ta mining, model
ion her of ates of onal on's ent of ion ion an ders onal ator ards udy a the cess nity cies able to tre pro gra ma res pro gra hin Un Da cre pe gra ou ch GP pro cre the ma ab cal sem
ment Vol. 8 No
Students
hjo S. L. Ma ana Universit
University. High t on public confid edictions of stude litates the Unive by their performa rate of students
cedure for Data M y utilizing the co best model. This
r's students data mance and predic chievement index ling algorithms compete. T eat underg ovide excel aduates to aintain the q sult, it is ne ocess. Mak aduation an nders will niversity. ata mining eate the erformance aduation. Se utcomes can hart formed PA and le oduce 4-qu eate a perfo e rate of p ade restrict ove 2.5 and lculated in mester. The
o. 1 (2015):
68-GPA Cat
argono ty, Bandung, I
h rates of studen dence as stakehol ent graduation a ersity to create th ance can be class performance in Mining) will be i ollege database. S research utilizes are expected to b ction of graduati
and graduation
This standard raduate stu llent servic
obtain high quality of e ecessary to king predic nd determin be a valua
facilitates segmentati and pre egmentation n be classi of two m evel of pr uadrant seg ormance ind progress as tions, positi d the rate o n percent
results are a
-74
egory
Indonesia
nt graduation and lders of higher ed nd determine the he segmentation sified in a quadr dex per semeste implemented in t Some algorithms student performa build the student ion. Based on thi
rates.
d also cover udy progra ce to stude h-quality res education an pass the ev ctions of ne the fact able input
the Unive ion of s ediction o n of student
fied in a q main parame
rogress wh gments, nam dex as the
y-the x-axis. tive y-axis of progress
cumulativ as follows:
d the quality ducation and e factors that of students’ rant chart is er. Standard this research. such as C5, ance data for t profile data.
is prediction,
rs how to am and ents and
sults. To nd, as a valuation student tors that for the ersity to students’ f their
Setiawan and Margono, Predicting Success Study Using Students GPA Category
Q1 Æ GPA < 2.5 and rate of progress <= 0 Q2 Æ GPA < 2.5 and rate of progress > 0 Q3 Æ GPA >= 2.5 and rate of progress <= 0 Q4 Æ GPA >= 2.5 and rate of progress > 0
GPA segmentation will be used in a predictive model of graduation. Parameters used in addition to GPA also included the master's students are expected to form the student profile data.
Making predictions, graduation can be done through the modeling process by utilizing the college database. The more parameters should be used to form a comprehensive model anyway. Some algorithms such as C5, C & R Tree, CHAID, and Logistic Regression tested to look for the best model.
The outcome is student category based on their study performance and prediction of graduation. Based on this prediction, the college may recommend actions to be taken to improve the student achievement index and graduation rates.
1.2. Contribution and Research Output
The results of this research will generate quadrant student performance that will allow University to act help to the success of student learning, especially for quadrant 1, 2 and 3. For sharper handling required additional data that will result in the group with specific characteristics that require different handling tips.
1.3. Purpose of Research
The purpose of this study was to analyze the performance of the progress of student learning outcomes and prevent the failure of student learning to improve graduation index University. As the research data is limited to certain departments of several forces.
1.4. Research Methodology
Standard methodology in data mining i.e. CRISP-DM (Cross Industry Standard Procedure
• Business Understanding, the target is to
understand the real problem which is faced by University including business
context that influence the problem.
• Data Understanding, the target is to
determine the relevant parameters, variables and attributes related to the problem.
• Data Preparationis an activity to manipulate
data so that it’s ultimately valid. This valid data will then be inputs to the modeling process.
• Modeling, is a process of determining the
best model that will be used to solve the problems.
• Evaluation is a group of activities intended
to evaluate in detail the accuracy of the output of the chosen model inclusive of changing business context that may alter the chosen model. The model has to be able to solve the defined problem.
• Deployment the target is to use the output
of the chosen model as a business tool to solve the problems.
2. Research Outcomes
2.1. Business Understanding
Once a prospective student accepted at universities and begin the process of learning, the student data will be entered into the database University. And then based on the progress of learning outcomes shown by GPA per semester, the student performance can be classified into four quadrants. By grouping based on this quadrant, the university can more easily monitor the progress of learning outcomes and immediately act to anticipate setbacks student achievement index by holding a mentoring program so that the expected progress is
Q1
GPA < 2.5 Growth < 0
Q2
GPA < 2.5 Growth > 0
Q4
GPA >2.5 Growth > 0
Relative Growth Rate
GP
A
Q3
The Asian J
2.2. D
The tar variables problem
Collect In
Some d database • Stude • Stude • Stude seme All data five year Describe Every co that understa Explore Compili needed forecast All data year bef Verify D Verifyin avoid a extractin server .
2.3. D
In this datasets them ou we choo content modelin null and could b process accordin that it c as integ more-co processe reforma
Journal of Technol
Data Understa
rget is to d s and att m at hand.
nitial Data datasets are e comprising ents Master ents Suppor ents Perfo ester)
sets are com rs to get GP
Data
olumn in al everybody anding of th
Data
ing datasets in the mo ting process asets above a fore are com
Data Quality.
ng data qu an error t
ng and p
Data Preparati
s process, that would ut from the ose and sep
that woul ng and fore d undefined be used for continued ng to the r can be used grating som omplete file es. Include atting data. logy Management nding determine th ributes tha collected fr g of: Dataset rting Data
ormance D
mpiled for pe PA.
ll datasets ar can h he data.
s per period odelling pro
as below. and the sam mpiled.
uality is don that may rocessing d
ion
we deter d be used as servers’ da parated colu ld be used ecasting. Da d was clean r modeling
with con equired dat for the mo me files to that will us ed in this
Vol. 8 No. 1 (20
he paramete at affect rom Univers Dataset (p er semester re described have corr
d that will ocess and/
me datasets o
ne in order cause wh data from
rmined wh s well as tak atabases. Th umns and d d further ata which w ned so that
process. T structing d ta structure odeling as w make a ne sed for furth
process w 15): 68-74 ers, the sity per for d so rect be or one to hile the hich king hen, data for was t it The data so well ew-her was Fig Fig da pro ag da da Fig Mo Fig str pe str pro 2.4 Af ab da an pro gra
gure 1. DFD
gure 1 desc ata that wi
ocess and gregate of ataset, that ataset and stu
gure 2. Data odeler
gure 2 is a ream, create erform data
ream repre ocess and ou
4. Data An
fter merging ove, we fin atabase whi nalysis and
ocess can b aphic mode
Academic
D Prediction
cribe the flo ill be used
prediction student aca merge wi udent suppo
a Preparatio
group of n d using IBM a preparati esent node
utput with t
nalysis
g process nally have ich is read prediction be analyzed as follow MERG Data Proc AGGREGATE n
ow of maki d in segm process. ademic perf
ith student ort dataset
on Using IB
nodes, know M SPSS Mod ion proces s for data the several a
of three d a single co dy to be u n. This ma e and pres
Master GING a for cess ing main mentation We use formance t master BM SPSS
Setiawan an
Figure 3
Having perform
Table 1.
Q4 to Q4 Q4 to Q3 Q4 to Q2 Q4 to Q1
2.5. Mo
Having we then some av by usin powerfu compare
nd Margono, Predi
3. Student Q
known t mance catego
Student Qu
2009 2 20093 90 53 74 38
0 0
1 1
odelling & Its
prepared d n must choo vailable mod ng auto-clas ul techniqu es a numbe
icting Success Study
Quadrant Dis
the profile ory per quad
uadrant Mov
9
3 20101 20 3 12 8 99 0 0 1 0
s Output
data needed ose the bes dels in IBM ssifier node ue that e er of diffe
dy Using Students G
stribution
of stud drant, we nee
vement From
010
2 20103 2 22 37 26 48 0 0 4 2
for modeli st model fro
SPSS mode e that can estimates a rent model
GPA Category
ent
ed tooth be
m Q4 To Ot
SEMES 2011
1 20112 17 30 78 11 0 0 4 1
ing, om eler be and ling
me acc the co ma
know the h her quadra elow.
ther Quadra
STER 2011
3 20121 45 6 74 4 0 0 5 2
ethods, rank curacy. The en be use
mpletion o anner.
historical mo nts, which
ants
2012 2 20123
1 0
1 2
0 0
0
king them chosen mo ed for pr of the curr
ovement fro can be d
2
3 20131 20 0 0 2 1 0 0 1 1
in order of odel is C.5 redicting su riculum in
m Q4 to described
013 2 20133 1 0 0 0 0 0 0 1
The Asian J
Data m period. accuracy
Figure 5
The pr process average amounts
Figure 6
Figure 7
Journal of Technol
modeling usin The focus y of stu
5. Predictor
redictor im is GPA 6th
score 6th s as shown i
6. Algorithm
7. Accuration
logy Management
ng the data s of the a udents com
Importance
mportance semester, s
semester in figure 5.
m of Predicti
n of Predict
Vol. 8 No. 1 (20
a of one cl analysis is
mpleting
e
for model student gend
and ‘A sc
ion Model
tion Model
15): 68-74
lass the the
cu ma
ling der,
core’
Mo alg
urriculum. M ain paramete
odeling p gorithm:
Modeling pr ers that influ
process ha
rocess gener uence, as fol
as the f
rates the llows:
Setiawan an
Figure 8
2.6. E
The pre of stud evaluate
Table 2.
The abo predicte that com in a tim confiden
nd Margono, Predi
8. Model Cl
Evaluation
ediction mo dent for ne e the accu
Accuracy D
ove table sh ed vs actual mpleted the ely manner nt with the r
icting Success Study
ustering
odel is appli ext period. uracy of p
Data, Predict
hows that th l success of study progr
is 88.034 % results.
dy Using Students G
ed to the d In order prediction,
ted versus A
he accuracy f the stude ram curriculu %. We are qu
GPA Category
data to we
co act
Actual
of ents um uite
to stu ref kn be
mpare betw tual data wh
solve the udent perf
flected thro nowledge, U est effort to
ween the p hich is shown
e problem formance b
ough its qu niversity wil
maintain th
prediction w n below.
i.e. under behavior w uadrant. W ll be able to he students t
with the
rstanding which is With that
The Asian Journal of Technology Management Vol. 8 No. 1 (2015): 68-74
3. Research Constraints
Complete supporting data will be greatly helpful to track the characteristics of the learning process of students inhibitor category. Unfortunately, such data cannot be obtained easily because it is in many cases very confidential. However with the available data, this study should be able to make the positive contribution.
4. Conclusion and Recommendation
The accuracy of the predictions against the actual data completion of the curriculum is from 87 % to 88.034 %. This proves that the resulting model can be used for the data set used. Model results from IBM SPSS Modeler streams will produce different accuracy when directly applied to the different data groups, for example to other universities, because the resulting parameter values will be different. Nevertheless forming method applied models have been tested, so that the same method, and different data sets should be re-modeling, to form the corresponding model. The prediction results and recommendations can be used by the counselor in a motivating student to intensify efforts to achieve learning success. For prediction accuracy, external information like activities, academic activities of students and family background data, should be included as a part of student demographic data.
References
Bramer, M. (2007). Principles of Data Mining.
Springer, London.
Giudici, P. (2003). Applied Data Mining
Statistical Methods for Business and Industry,
Wiley.
Han, J. and Kamber, M. (2006). Data Mining
Concepts and Techniques Second Edition.
Morgan Kauffman, San Francisco. Pyle, D. (2003). Business Modeling and Data
Mining. Morgan Kaufmann Publishers.
Rud, O. P. (2000). Data Mining Cook Book,
John Wiley and Sons Nov.
Susanto, S., and Suryadi, D. (2010). Pengantar Data Mining, Menggali Pengetahuan dari
Bongkahan Data, Andi Yogyakarta
Santosa, B. (2007). Data Mining Teknik
Pemanfaatan Data untuk Keperluan Bisnis,
