You are on page 1of 10

Optimization Method of AHP with SAW for Recruitment New Employee :

Case Study Verint System

RINGKASAN TESIS

Oleh:
Muhammad Saipul Rohman
1411600818

PROGRAM STUDI MAGISTER ILMU KOMPUTER(MKOM)


PROGRAM PASCA SARJANA
UNIVERSITAS BUDI LUHUR
JAKARTA
2016

Optimization Method of AHP with SAW for Recruitment New Employee :


Case Study Verint System
Muhammad Saipul Rohman1, Dana Indra Sensuse2
Program Studi Magister Ilmu Komputer , Program Pascasarjana, Universitas Budi Luhur
Ciledug Raya Street, Petukangan Utara, Kebayoran Lama, Jakarta Selatan 12260 Telp. (021) 5853753, Fax.
(021) 5869225
E-mail :msaipulrx@gmail.com1,dana@cs.ui.ac.id 2
Abstrack
Recruitment selection for new employees is important for the company. By choosing the right candidate as an
employee so the company's operations can run smoothly so that it can compete with other companies. Human
Resource Section is one part that is vital for the recruitment selection of new employees. Possibility of the risk of
human error, the assessment factors subjectivity and checks incoming files one by one so it takes a long time a
major issue when the selection process for new employees at Verint Systems.AHP is one of the decision support
system method to calculate the weight of each criterion and SAW method is one of the decision support system
method to rank each alternative based on any criteria.Method of Exact Match is used to check the accuracy of
the word if there are same word then it exact and if there is none then the word not match, while the flag is a
marker if true then given flag 1(true) Exact Match and if not true then given flag 0(false) not match .The method
used in this research is the method of AHP and optimization method of AHP with SAWalso method of calculate
the accuration of them use Exact Macth. Based on test result that has been processed and analyzed, the result of
optimization method of AHP with SAW produces accuracy percentage about 70% better than accuracy
percentage of AHP about 10% which proves that the optimization method of AHP with SAW make the prototype
of system better than by using AHP.
Keywords: SAW, AHP, Decision Support System, employee recruitment, Exact Match, Human Resource

1. Introduction
One factor could be the survival of a company is
the owned human resources (HR). HR management
of an enterprise is one important aspect to
determine the success of the work of the company.
To get quality human resources, the company
implemented a strict selection on hiring applicants
for employment to suit the needs of the company
[5].
During the process of checking and selection is
done manually by looking one by one file submitted
by the applicant. So it become inconvenient also
inefficient and takes a long time, maybe there are
the possibility of human error and assessments are
subjective so that the selected candidates are not
the best candidate.
In writing this thesis, the author proposes a
decision support system using AHP and SAW that
will help solve the problem of the selection process
of new employees. SAW method was taken because
the method is simple, easy to understand,
computationally efficient and has the ability to

measure the relative performance of the alternative


- the alternative [1].
In a previous study using SAW method of decision
support system that will help to select employees
[1]. In a another study conducted research for
selecting new teaching staff using AHP [13].
Before doing the study, researchers used the AHP
method but the results are less accurate with the
original data in the field. Therefore, researchers
conducted a study of optimization methods AHP
added by SAW so the results are more accurate
than the method of AHP alone in selecting new
employees.
2. Theoritical Basis
2.1 Analytic Hierarchy Process (AHP)
AHP is a decision support models developed by
Thomas L. Saaty. Decision support models will
elaborate multi-factor problem or a complex multicriteria into a hierarchy. According to [8], the
hierarchy is defined as a representation of a
complex problem in a multi-level structure where

the first level is a goal, followed by level factors,


criteria, sub-criteria, and so on down to the last
level of the alternatives.
With a hierarchy, a complex problem can be
decomposed into groups - the group who then
arranged into a form of hierarchy so that the
problem would appear more structured and
systematic.
In resolving problems with AHP there are certain
principles that must be understood as follows [6]:
1) Make hierarchy
A
complex
system
can
be
understood by breaking it down into
supporting elements, arrange and
combine elements hierarchically
2) Assessment criteria and alternatives
Criteria and alternatives conducted by paired
comparison. According to [16], for a variety
of problems, a scale of 1 to 9 is the best scale
for expressing opinions. Values and opinions
qualitative definition of scale can be
measured using the Saaty comparison
analysis table as shown in Table 1 below:
Table 1 : comparison scale of AHP(Saaty)
Values
1
3
5
7
9
2,4,6,8
1/(1-9)

Description
A same important (Equal) with B
A little important (Moderate) than B
A more important (Strong) than B
A Very important (Very Strong) than B
A absolute important (Extreme) than B
When hesitating between two adjacent
values
Contrary value of the interest rate on a
scale of 1-9

3) Synthesis of Priority
For each criteria and alternatives, should be
paired comparisons (Pairwise Comparison).
Values of the relative ratio of the entire
alternative criteria can be adjusted to a
predetermined judgment to make weight and
priority. Weights and priority calculated by
manipulating the matrix or through the
completion of a mathematical equation.
4) Logical Consistency
Consistency has two meanings. First, similar
objects can be grouped according to the
uniformity and relevance. Second, regarding
the level of relationships between objects
based on specific criteria. Calculation of
logical consistency is done by following the
steps - steps as follows:
a. Multiplying matrices with corresponding
priority.
b. Summing up the results of multiplications
per line.
c. The sum of each row is divided priorities
concerned and the results summed.

d. Results c divided by the number of


elements, will be maks.
e. Consistency Index (CI) = (maks-n) / (n-1).
f. Consistency ratio = CI / RI, where RI is a
random index consistency. If the consistency
ratio 0.1, the calculation result data can be
justified.
Table 2 : Random Index Values
n
2
3
4
5
6
7
8

RI
0.00
0.58
0.90
1.12
1.24
1.32
1.41

2.2 Simple Additive Weighting (SAW)


Method of simple additive weighting (SAW) is a
weighted summation method. The basic concept is
simple additive weighting method (SAW) is
looking for a weighted sum of rating performance
on each alternative on all criteria. method of
simple additive weighting (SAW) requires a
decision matrix normalization process (X) to a
scale that can be compared with all the ratings of
existing alternatives. SAW method know their two
(2) attributes that criterion advantages (benefits)
and cost criteria (cost). The fundamental
difference of this criterion is in the selection
criteria when making decisions 19].
Step calculation methods [9]:
1) Determine the alternative, namely Ai.
2) Determine the criteria that will be used as a
reference in the decision, namely Cj.
3) Provide rating matches the value of each
alternative on each criterion.
4) Determine the weight of preference or level of
interest (W) each criterion.

5) Make a table rating the suitability of each


alternative on each criterion
6) Make a decision matrix X formed from rating
table suitability of each alternative on each
criterion. {x} value of each alternative (Ai) on
each criterion (Cj) that have been determined
where, i = 1,2, ..., m and j = 1,2, ..., n.

7) Perform normalized decision matrix X by


calculating the value of normalized performance
rating (rij) of alternative (Ai) on the criterion (Cj).

calculate the accuracy rate method or algorithm


after using exact match technique using confusion
matrix equations 2:11 as follows [2]:

accuracy ( )=

Description :
a) It said benefit criteria if value of
give
benefit for decision makers, otherwise cost
criteria if
have cost for decision makers
b) If benefits criteria then value of
divided

of data that true prediction by system x 10


of all sample data

2.4 Conceptual Framework


The conceptual framework of this research as
follows:

with value of
from each column,
while for cost criteria, value of
from each column divided with value of
8) Hasil dari nilai rating kinerja ternormalisasi
(rij) membentuk matrik ternormalisasi (R).

9) The final result preference value (Vi) obtained


from the sum of the normalized rows of the matrix
multiplication element (R) with a weight of
preferences (W) corresponding element column
matrix (W).

Description :
Vi = rangking for each alternatif
Wj = the weights of each criterion
rij = value normalized performance rating
The calculation result Vi greater value indicates
that the alternative Ai is the best alternative [4].

Picture 1 : Conceptual framework

2.3 Accuracy Testing methods / algorithms


To test the accuracy of the method or algorithm
there are two stages, by matching the output data
from the system with the rangkings of data from
users use exact match and calculating the
percentage for accuracy.Exact Match is the steps
being taken to check the accuracy of the word if the
same word is appropriate and if there is none then
the word Not Match, while the flag is a marker
when actually given the flag is 1 (true) Exact
Match. if it is not true then given a flag / mark of 0
(not match) in preparation flag 0 is not included.
Then is a similary calculations add up all the flags
are worth 1 (true) derived from the first flag of the
index 1 and also the sum flag 2 from index flag 2.
The sum is taken the smallest value [3].Then to

3. Research Methodology
3.1 Research Method
Research optimization methods with case studies
on
enterprise software house is a kind of applied
research (Applied Research). The results of the
research conducted can be directly implemented to
solve the problems [7].
In this study, the method used is quantitative
method in which the calculation process in
accordance with the existing formula based on the
methods used to obtain a decision [7].
3.2 Sample Selection Method
In this study, the sample data used are the data of
candidates applying to Verint System for

department Support within the period of the month


of January to June 2015 with details of the column
as follows: Name of prospective employees,
Background education, work experience, test
scores of English , the score of the technical test
and the results passed or not at this stage of the
interview call
3.3 Data Collection Method
The data collection was conducted to obtain
information and data related to this study. In
collecting data and information, do methods of
collecting primary data and secondary .
1) The primary data collection methods.
By collecting data directly to the data
source which is owned Verint System
also data collection is done by
observation,
interviews
and
questionnaires.
2) Secondary data collection methods.
Reading, observing and studying data from sources
associated with this research.
3.4 Analyst and Testing Method
A. Analyst Method
In this study, researchers conducted the
optimization of the calculation method with
AHP added SAW method to determine the
rank of the alternatives that already exist so
that the expected results of the calculations
are more accurate than just using the AHP. To
test the accuracy AHP and AHP-SAW adopt
methods Exact Match. The final result of
optimization methods AHP with SAW aims to
obtain the best alternative candidates that
could help to determine the candidates will be
recruited new employees.
B. Testing Method
Testing here, the author uses testing to see the
level of accuracy of the algorithm which
produces better accuracy percentage. The
algorithm is the AHP and AHP - SAW. For the
testing methods use
Exact Match for
calculating the level of accuracy and
calculate the percentage of accuracy using
the confusion matrix.

3.5 Step by step of Research


The step by step of Research for optimization
methods of AHP with SAW can see as bellow :

Picture 2 : step by step of research

4. Research Result and discussion


In this study discusses the results of the analysis of
the calculation method for weighting and ranking
the AHP then compared with the results of the
analysis of the calculation method for weighting
AHP and SAW method for perangkingan. So the
results of this comparison method is more
accurate to the original data from the object of
research.
4.1 Weighting Process
In this study the first user will be given a
questionnaire to discover the importance of the
existing criteria then used AHP to determine the
weight of each criteria that will be used as a
benchmark assessment. The following tables
summarize the results of the level of interest form
questionnaire given by the user according to Saaty
scale:

Table 3 : the level of criteria importance


according to user
LBP
PK
NTB
NTT
HW

LBP
1

PK
3
1

NTB
2
1/5
1

NTT
5
2
3
1

HW
7
7
7
5
1

Description :
LBP = Latar
Belakang
Pendidikan
(Background Education)
PK
= Pengalaman Kerja (Work of
Experience)
NTB = Nilai Tes Bahasa Inggris (Score
of English Test)
NTT = Nilai Tes Teknikal (Score of
Technical Test)
HW
=
Hasil Wawancara (Interview
Result)
4.2 Weighting the criteria with AHP
Steps by steps determining criteria weights using
AHP method as follows:
1. In this method, first made pairwise comparison
matrix taken from the table values of importance
criteria by user. To simplify the matrix inserted
into table. Once the value of each importance
criteria included then add up each criteria.
Table 4 : pairwise comparison matrix
LBP
PK
NTB
NTT

LBP
1
0,33
0,50
0,20

HW

0,14

sum

2,18

PK
3
1
5
0,5
0
0,1
4
9,6
4

NTB
2
1/5
1
0,33

NTT
5
2
3
1

HW
7
7
7
5

0,14

0,20

3,68

11,2
0

27,0
0

2. Then each value importance criteria divided by


the number of their importance criteria
(normalization matrix). After that summed to the
right value divided by the number of criteria to
calculate the weighting of criteria.
Table 5 : Calculating normalization matrix and
weighting criteria
LB
P
PK
NT
B
NT
T
HW

LB
P
0,4
6
0,1
5
0,2
3
0,0
9
0,0

PK
0,
31
0,
10
0,
52
0,
05
0,

NT
B
0,5
4
0,0
5
0,2
7
0,0
9
0,0

NT
T
0,4
5
0,1
8
0,2
7
0,0
9
0,0

H
W
0,
26
0,
26
0,
26
0,
19
0,

01

04

843

Examples of value matrix normalization line to


1:
Column vs row
LBP vs LBP : 1 / 2,18 = 0,46
PK vs LBP : 3 / 9,64 = 0,31
NTB vs LBP : 2 / 3,68 = 0,54
NTT vs LBP : 5 / 11,20 = 0,45
HW vs LBP : 7 / 27,00 = 0,26
Then value of weighting are : (0,46 + 0,31 +
0,54 + 0,45 + 0,26)/5 = 0,404072
3. The next step is to calculate the lambda max. To
calculate lambda max with 2 steps as below: Step
1 is the value of the importance of each criteria is
multiplied by the weight of each criteria then sum
and divided by the each weight. Step 2 sum the
scores in step 1 divided by the number of the
criteria.
4. And the last stage is to calculate the value of CI
(Consisten Index), set the value RI (Ratio Index)
and CR (Consisten Ratio). RI value already has a
default value specified by the appropriate number
of criteria Saaty. In this study has five criteria
then its value RI is 1.12. The formula for
calculating the value of CI are: (lambda max
value - criteria amt) / (total criteria - 1). Then the
value of CI is (5.41 to 5) / (5-1) = 0.10. The
formula for calculating the value of CR are: CI /
RI. Then the value of CR is 0.10 / 1.12 = 0.09. CR
consistent value for under 0.1.
4.3 Accuracy testing methods
Total records have 50 data and for sample
calculation used 10 data were applying for work
in the support department. For all total of data
can be seen in appendix 5. for testing the accuracy
of AHP and AHP - SAW using exact match. The
trick is the rangking result of output 10 from the
system AHP and AHP - SAW matched with the
sample data in the field as much as 10 data. If the
result is exactly the same as was given a value of
1, if not rated 0 then calculate the percentage of
accuracy.For a sample of data used by user
(dataset) as many as 10 data with the following
details:
Table 6: sample data that used by user

Bobot

Code

Name

LBP

0,404
072
0,149
822
0,309
483
0,101
78
0,034

CK1
3
CK1
4
CK1
5
CK1
6
CK1

Arief

NTB

NTT

HW

S1

P
K
5

85

86

0,9

Hafid

S1

90

87

0,9

Winda

S1

85

75

0,9

Gede Seri

S1

10

88

85

0,9

Meidy

S1

15

92

88

0,9

7
CK1
8
CK3
0
CK3
4
CK4
4
CK4
0

Giasi
Alex
Sandro
M Solihin

S1

88

87

0,9

S1

89

84

0,9

Dhian Putra

S2

85

80

0,9

Sofyan H

S1

86

84

0,9

Liliani

D3

57

53

0,9

The results of testing the accuracy \using exact


match to the results on ranking AHP and AHP SAW are as follows:

Method of AHP
The result of top 10 rangking method of AHP as
below
Table 7 : The result of top 10 rangking method of
AHP
Alternatif
Code
CK1
CK24
CK4
CK3
CK5
CK40
CK8
CK36
CK43
CK23

M=1
N = 10 people
Accuracy (%) = M/N x 100%
Accuracy (%) = 1/10 x 100 % = 10%
So value of accuracy AHP method is 10%
Description :
M = total people that true prediction by system that
matched with exact match method
N = total all of sample data (10 people)

Method od AHP - SAW


The result of top 10 rangking method of AHP SAW as below
Table 9 : The result of top 10 rangking method of
AHP - SAW
Alternatif
Code
CK17
CK16

Alternatif Name

Result

Rangking

Fariz Fadlan
Andry
Febriansyah
Luthfan
Areka
Ardi
Khristian Wijaya
Sirin Shafwati
Liliani
Caesar Putranto
Fabianus Hendy
Evan
Hadi Gunawan
Ditra Alfilia

0,036152
0,033194

1
2

CK14

0,027506

0,026765
0,026309
0,025654
0,025451
0,025373

4
5
6
7
8

CK50
CK34
CK44

0,025295
0,023088

9
10

Here below are the results of testing the accuracy


using exact match to the results of AHP method.
Method Exact is matching user data with the data
from the calculation of AHP. If there is the same
then the calculated one if not then computed 0.
How to match the data using exact match is not
bound by the order or ranking.
Table 8 : accuracy testing for AHP method
Comparison
Data User
CK13
CK14
CK15
CK16
CK17
CK18
CK30
CK34
CK44
CK40

AHP
CK1
CK24
CK4
CK3
CK5
CK40
CK8
CK36
CK43
CK23

Exact Match
0
0
0
0
0
0
0
0
0
1

Result from accuracy testing using exact match


above show that there are only 1 data that match
that is CK40 between data from user and calculate
result of AHP. So the value of accuracy percentage
as bellow

CK13
CK18
CK30
CK15

Alternatif Name

Result

Rangking

Meidy Giasi
Gede
Serikastawan
Hafid
Inggiantowi
Robert Tjahjadi
Dhian Putra
Sofyan
Hasanuddin
Arief Hidayat
Alex Sandro
Muhammad
Sholihin
Winda Giam

0,997738
0,930949

1
2

0,889998

0,875668
0,87525
0,873149

4
5
6

0,872047
0,863294
0,863265

7
8
9

0,849619

10

Here below are the results of testing the accuracy


using exact match to the results of AHP method.
Table 10 : accuracy testing for AHP - SAW method
Comparison
Data
User
CK13
CK14
CK15
CK16
CK17
CK18
CK30
CK34
CK44
CK40

With
-SAW
CK17
CK16
CK14
CK50
CK34
CK44
CK13
CK18
CK30
CK15

AHP

Exact Match
1
1
1
1
1
1
1
1
1
0

Result from accuracy testing using exact match


above show that there are 9 data that match
between data from user and calculate result of
AHP SAW. So the value of accuracy percentage
as below.
M=9
N = 10 people
Accuracy (%) = M/N x 100%
Accuracy (%) = 9/10 x 100 % = 90%

So value of accuracy AHP - SAW method is 90%.


From these results it can be concluded that the
optimization results with SAW AHP method
produces a better accuracy rate than using AHP
alone.

[2]

[3]

5. Conclusion
5.1 Conclusion
Based on the analysis and the testing of this study
was concluded to address existing problems that
are the percentage level of accuracy of decision
support systems using AHP - SAW 90% better than
the level of accuracy of AHP method in which the
percentage level of accuracy of 10% which proves
that AHP - SAW made the system being better than
using AHP.
5.2 Suggestion
There are some suggestions for further studies of
this research as follows:
1. It is expected that further research to add
modules related recruitment process so that
system can be used to support the recruitment of
new employees.
2. It is expected that further research, recruitment
of data where in the HRD expected reduction in
duplicate data, incomplete or there are blank.
3. It is expected that further research add others
variables to support decision support system
changes according to user needs in the future or
developed with others algorithm of decision
support system such as algorithm C4.5, TOPSIS,
Fuzzy AHP, PROMETHEE or the other.

References
[1] Afshari, A., Mojahed, M. & Yusuff, R., 2010,
Simple Additive Weighting Approach To
Personnel Selection Problem. International
Journal of Innovation, Management and
Technology, 1(5), pp.511515. Diakses pada
27
Januari
2016,
<
http://www.researchgate.net/publicat
ion/256031272_Simple_Additive_Wei

[4]
[5]

[6]

[7]
[8]

[9]

ghting_Approach_to_Personnel_Selec
tion_Problem/file/e0b49524c34debf7
b5.pdf>.
Ahmad, F., M. Aziz, M dan Hadi Suyono,
2014, Komparasi Fuzzy AHP dengan AHP
pada Sistem Pendukung Keputusan Investasi
Property. Jurnal EECCIS Vol. 8, No.2
Heriyanto. 2011, Penggunaan Metode Exact
Match Untuk Menentukan Kemiripan Naskah
Dokumen Teks, Universitas Pembangunan
Nasional Veteran, Jurnal Telematika Vol. 8
No.1, pp.43-52.
Kusumadewi, S. 2006, Fuzzy Multi Atribut
Decision Making, Penerbit Graha Ilmu.
Lorincova, S. 2015, The Improvement of the
Effectiveness in the Recruitment Process in
the Slovak Public Administration, Business
Economics
and
Management
2015
Conference, 34(2015): 382-389. Diakses pada
2
Februari
2016,
<https://www.researchgate.net/publi
cation/291337176_The_Improvemen
t_of_the_Effectiveness_in_the_Recrui
tment_Process_in_the_Slovak_Public_
Administration>.
Marimin dan Maghfiroh, N. 2013, Aplikasi
Teknik Pengambilan Keputusan dalam
Manajemen Rantai Pasok, cetakan Ketiga, PT
Penerbit IPB Press, Bogor.
Moedjiono. 2012, Pedoman Penelitian,
Penyusunan dan Penelitian Tesis. V5,
Program Pascasarjana Universitas Budi Luhur.
Saaty, T.L. 1991. Pengambilan Keputusan
bagi para Pemimpin, Proses; Hirarki Analitik
untuk Pengambilan Keputusan dalam situasi
yang Kompleks. Seri Manajemen no.134,
PPM, Jakarta.
Usito, N.J., 2013, Sistem Pendukung
Keputusan Penilaian Proses Belajar Mengajar
Menggunakan Metode Simple Additive
Weighting (Saw). Journal of Chemical
Information and Modeling, 53(9), pp.1689
1699.
Available
at:
core.ac.uk/download/pdf/18605548.pdf.

You might also like