Академический Документы
Профессиональный Документы
Культура Документы
A = (x; m
A
(x))[x X
m
A
(x) is the membership function or grade of membership of x in
A
that maps X to the membership space M(when Mcontains only the
two points 0 and 1,
A is nonfuzzy and m
A
(x) is identical to the
characteristic function of a nonfuzzy set). The range of the
Table 1
Key performance criteria for promotion screening.
Key performance criteria Performance indices
Service rating 1. Completeness of job objectives
2. Service rating on job assignment
Multi-area aptitude 1. Creativity
2. Organizational contribution
3. Management capability
4. Achievement
5. Job expertise
6. Teamwork
Growth potential 1. Planning
2. Communication
3. Group discussion
4. Foreign language
5. Information systems usage
Innovativeness 1. Innovation score
C. Moon et al. / Applied Soft Computing 10 (2010) 512519 513
membership function is a subset of the nonnegative real numbers
whose supremum is nite. Elements with a membership of zero
degrees are normally not listed.
Denition 2. The (crisp) set of elements that belong to the fuzzy
set
A at least to the degree a is called the a-level set:
A
a
= x X[m
A
(x) _a
A
/
a
= x X[m
A
(x) >a is called strong a-level set or strong a-
cut.
Denition 3. A fuzzy number
Mis a convex normalized fuzzy set
M
of the real line R such that
1. It exists exactly one x
o
R with m
M
(x
0
) = 1.
2. m
M
(x
0
) is piecewise continuous.
In this paper, trapezoidal fuzzy numbers are used. A fuzzy
number
Mcan be dened as (a, b, c, d) as shown in Fig. 1. In addition
its membership function is dened as in Eq. (1).
m
M
=
0; x <a
x a
b a
; a _ x _ b
1; b _ x _ c
x d
c d
; c _ x _ d
0; x >d
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
(1)
For example, a fuzzy number, which could be approximately
5, would normally be dened as the quadruple (3, 4, 6, 7). If b = C
in a fuzzy number
M = (a; b; c; d), then
Mis called a triangular fuzzy
number. Additionally, a nonfuzzy number k can be expressed as (k,
k, k).
The basic operations of the fuzzy numbers used in this paper are
dened as follows:
M
1
M
2
= (m
1l
m
2l
; m
1lm
m
2lm
; m
1um
m
2um
; m
1u
m
2u
) (2)
M
1
M
2
= (m
1l
m
2l
; m
1lm
m
2lm
; m
1um
m
2um
m
1u
m
2u
) (3)
M
1
= (m
1u
; m
1um
; m
1lm
; m
1l
) (4)
1
M
1
=
1
m
1u
;
1
m
1um
;
1
m
1lm
;
1
m
1l
(5)
where
M
1
= (m
1l
; m
1lm
; m
1um
; m
1u
) and
M
2
= (m
2l
; m
2lm
; m
2um
; m
2u
)
represent two trapezoidal fuzzy numbers with lower, lower modal,
upper modal and upper values.
Denition 4. A linguistic variable is characterized by a quintuple
(x; T(x); U; G;
= (a
, b
, c
, d
), and y
i
= (a
i
, b
i
, c
i
, d
i
) be
the positive-ideal solution, negative-ideal solution, and ith
candidates score, respectively. Also, let m
*
, m
, and m
i
be the
mean values of the positive-ideal solution, negative-ideal solution,
and ith candidate.
We dene the following distance measurements:
d
+
=
(a
+
)
2
(b
+
)
2
(c
+
)
2
(d
+
)
2
q
(10)
d
(a
)
2
(b
)
2
(c
)
2
(d
)
2
q
(11)
d
i
=
a
2
i
b
2
i
c
2
i
d
2
i
q
(12)
Using Eqs. (9)(12), we suggest the following ranking measures.
M
+
i
=
1
2
d
i
d
d
+
d
m
i
m
m
+
m
(13)
where d
_ d
i
_ d
*
and m
_ m
i
_ m
*
.
The values of M
+
i
are between 0 and 1. If the value of M
+
i
is larger,
the candidate is closer to the ideal solution and farther from the
negative solution. In (13), each candidates scores are simulta-
neously compared with the positive (negative)-ideal solution in
terms of the distance and fuzzy mean. The greater the distance
from the negative-ideal solution and the larger the fuzzy mean
value compared to the negative-ideal solution, the higher the
candidate ranks.
Eq. (13) is the mixture of the ranking methods in Hwang and
Yoon [22] and Goetshel and Voxman [26]. The rst termof Eq. (13)
shows the distance from the ideal solution and the second term
shows the difference fromthe ideal solutions fuzzy mean. The fact
that one candidates score is closer to the ideal solution does not
necessarily mean that it has a greater fuzzy mean than the others.
We consider the distance fromthe ideal solution and fuzzy mean at
the same time. We derive Eq. (13) from the concept that the ideal
solution has to be the collection of the best solutions in each
criterion. This is the difference between our approach and TOPSIS
introduced by [22]. In the promotion screening process, it is not
y
i
=
X
l
k=1
w
k
1
n
X
n
j=1
x
i jk
0
@
1
A
8
<
:
9
=
;
=
X
l
k=1
w
k
l
P
l
h=1
w
h
u
;
w
k
l
P
l
h=1
w
h
um
;
w
k
l
P
l
h=1
w
h
lm
;
w
k
l
P
l
h=1
w
h
l
!
P
n
j=1
x
i jk;l
n
;
P
n
j=1
x
i jk;lm
n
;
P
n
j=1
x
i jk;um
n
;
P
n
j=1
x
i jk;u
n
! ( )
=
X
l
k=1
w
k
l
P
l
h=1
w
h
u
P
n
j=1
x
i jk;l
n
;
w
k
l
P
l
h=1
w
h
um
P
n
j=1
x
i jk;um
n
;
w
k
l
P
l
h=1
w
h
lm
P
n
j=1
x
i jk;lm
n
;
w
k
l
P
l
h=1
w
h
l
P
n
j=1
x
i jk;u
n
! ( )
=
X
l
k=1
w
k
l
P
l
h=1
w
h
u
P
n
j=1
x
i jk;l
n
( )
;
X
l
k=1
w
k
l
P
l
h=1
w
h
um
P
n
j=1
x
i jk;um
n
( )
;
X
l
k=1
w
k
l
P
l
h=1
w
h
lm
P
n
j=1
x
i jk;lm
n
( )
;
X
l
k=1
w
k
l
P
l
i=h
w
h
l
P
n
j=1
x
i jk;u
n
( ) !
(8)
C. Moon et al. / Applied Soft Computing 10 (2010) 512519 515
important that one candidate has better grades than the others. It
is more important that the grades of candidates could exceed the
level necessary for promotion.
4. Performance appraisal and promotion ranking system
Performance appraisal and promotion ranking (PAPR) is the
main primary role of human resource management in military
organization. Therefore, to support the PAPR function, we develop
a system called the PAPR system. The system is effective in
performance appraisal and ranking, and can help organizations
transform the employee evaluation process into a well-dened,
fair, and transparent process. This developed system is composed
of three main functions: (1) determining the weights of evaluation
criteria, (2) rating and data monitoring, and (3) data aggregation
and ranking.
4.1. Determining the weights of evaluation criteria
In order to aggregate decision data, the nominal group
technique plays an important role in generating ideas, prioritizing
them, and coming to a group consensus among organizational
members. An effective electronic nominal group technique is
developed for generating evaluation criteria and providing a
prioritized list of criteria through voting by group members. The
weight of evaluation criteria can be determined by aggregating the
decisions of external and internal experts in the eld of human
resource management. The process for determining weights using
the electronic technique is shown in Fig. 3.
In Fig. 3, the process for determining the weights of evaluation
criteria consists of three parts as follows:
(1) The department of human resource management develops a
list of key criteria. In this appraisal and ranking system for
promotion screening, four criteria, such as, service rating,
multi-area aptitude, growth potential and innovativeness, are
considered.
(2) Decision makers evaluate the criteria through the electronic
nominal group technique. The technique provides an advanta-
geous decision-making environment to equalize participation,
encouraging the free ow of ideas in a nonthreatening setting
and enabling participaints to reach nal decisions.
(3) The weights of key criteria are determined by aggregating the
decision makers results.
This technology can be used to collect and assess the relative
importance of evaluation criteria at various organizational levels.
The linguistic weighting variables, seen in Fig. 2, are used to assess
the weight of each criterion.
4.2. Rating and data monitoring
The second functionconsists of rating, data monitoring and data
processing. The committee for decision making involves a number
of evaluators with equal authority who assess each candidate and
implement individual analyses. The candidates performance
scores recorded at their respective past organizations are provided
to the human resource management department for rating.
Fig. 3. Process of determining the weights of criteria.
Fig. 4. Example of data monitoring for abnormality and incorrectness. Fig. 5. Monitoring procedure.
C. Moon et al. / Applied Soft Computing 10 (2010) 512519 516
The monitoring process is employed to examine abnormal
or incorrect data received from the evaluators. The degree
of scattering is used to control abnormal data as shown in Fig. 4.
The procedure for data monitoring is shown in Fig. 5.
This procedure utilizes all performance evaluation data without
excluding the maximum and minimum scores from the nal
performance evaluation results. If all input is perceived as normal,
the linguistic assessments received from evaluators are converted
intotrapezoidal fuzzynumbers toconstruct a fuzzy-decisionmatrix,
andfromthis, the fuzzyweight of eachcriterionis determined. Then,
a weighted-normalized fuzzy-decision matrix is constructed.
4.3. Data aggregation and ranking
The appraisal results of individual evaluators should be
aggregated to compute the nal appraisal scores to determine
the ranking of each candidate. The data aggregation and ranking
Fig. 6. Overall schematic ow diagram of the PAPR system.
Fig. 7. Assigning the weights of evaluation criteria.
Table 2
The weights of evaluation criteria.
Criteria Fuzzy number
C
1
0.2056 0.2551 0.2766 0.3415
C
2
0.1963 0.2449 0.2553 0.3293
C
3
0.2056 0.2551 0.2766 0.3415
C
4
0.1589 0.2041 0.2340 0.2927
Table 3
The aggregated scores for each candidate.
Candidate Fuzzy number
y
1
5.5140 7.8571 8.9574 12.0813
y
2
5.3330 7.6293 8.7305 11.8780
y
3
4.9003 7.6905 8.6028 11.8577
y
4
5.5296 7.8741 8.9858 12.0976
y
5
5.5885 8.0782 9.2199 12.2927
y
6
5.6667 8.0544 9.4397 12.3821
y
7
4.8162 6.9830 8.0993 11.3252
y
8
5.6760 8.0646 9.2057 12.2805
y
9
5.5140 7.8571 8.9574 12.0813
y
10
5.2617 7.5408 8.5319 11.7602
C. Moon et al. / Applied Soft Computing 10 (2010) 512519 517
function transforms the several individual multi-criteria scored
lists of candidates into one aggregated rank-ordered list. Then,
using the ranking measures to determine the ranking of each
candidate, nalists can be selected for promotion.
The overall schematic ow diagram of the PAPR system is
shown in Fig. 6.
5. Numerical example
An example of ranking promotion candidates is introduced and
demonstrates the effectiveness of the PAPR system. In this
example, there are three evaluators, four criteria, and 10
candidates. A detailed procedure for ranking is shown in Section
3. Eqs. (2)(5) are used for the fuzzy operations. First, the weights
of evaluation criteria are determined at by aggregating the
decisions of external and internal experts in the eld of human
resource management, as shown in Fig. 7. Then, the aggregated
weights of the criteria are summarized in Table 2.
Then, the candidates scores are evaluated for each criterion.
Appendix A shows the scores that each candidate received from
each evaluator. The rst candidates score from the second
evaluator for the third criterion is G, so x
123
= (7, 8, 8, 9). From
Eq. (8), the aggregated scores of each candidate can be expressed
by a fuzzy number, y
i
. Table 3 displays the fuzzy numbers
representing the aggregated scores of each candidate.
In this case the positive and negative-ideal solutions are
y
*
= (6.1308, 8.6327, 10.4255, 13.0488), and y
= (0, 0, 1.0426,
2.6098) which are derived from Table 2, the weights for criteria,
and Eq. (6), fuzzy membership functions.
Lastly, the candidates are ordered by the proposed ranking
method. The distance and fuzzy means are derived using Eqs. (9)
(12), and the candidates are ranked using Eq. (13). In Table 4, the
value represents the degree of separation fromthe ideal solution. A
larger value means that the candidate is closer to the ideal solution,
in this case, the sixth candidate attains the highest ranking. Fig. 8
shows the visual ranking results from the PAPR system.
From the experimental result, we know that the developed
PAPR system can handle applicant records that contain both
qualitative and quantitative information. Finally, the system
provided performance evaluations of all candidates.
6. Conclusions
This paper develops an efcient performance appraisal and
ranking systemfor the promotion screening of candidates applying
for a particular commission in a military organization. The system
uses fuzzy theory and electronic nominal group technique to
produce fair ranking decisions through a multi-criteria perfor-
mance appraisal process. The electronic nominal group technique
is adopted to collect and assess the relative importance of various
performance evaluation criteria collected at different organiza-
tional levels. This technology can prevent any particular organiza-
tions self-interests from dominating the selection and weighting
of performance evaluation criteria.
A new ranking procedure considering the metric distance and
fuzzy mean value is also proposed, which makes it possible to rank
order the performance of candidates by aggregating the scores of
multiple evaluators. The systemalso has a monitoring function that
uses all performance evaluation data without any removal. This
function is to prevent abnormal evaluation data which could occur
when there is an overly inuential member in the evaluators group,
or when a particular evaluator gives an incorrect evaluation result.
The system developed was applied to a military organization in
Korea. The results of this example have shown that this systematic
approach with a fuzzy procedure is a suitable method to produce
transparent and fair multi-criteria performance evaluations in
military organizations.
Appendix A
We assess 10 candidates scores for four criteria by three
evaluators. In the actual system, the evaluators score the candidates
using a computer program and the rawdata are not shown. However,
for the purpose of demonstrating the procedure used in the example,
it is necessary to show the raw scores here.
Criteria Candidates Evaluators Criteria Candidates Evaluators
D
1
D
2
D
3
D
1
D
2
D
3
C
1
A
1
VG G G C
3
A
1
VG G G
A
2
VG G VG A
2
G VG G
A
3
G VG G A
3
G G G
A
4
G G VG A
4
G VG VG
A
5
G G VG A
5
VG G VG
A
6
VG VG G A
6
G MG G
A
7
VG G MG A
7
MG G G
A
8
G G G A
8
VG VG VG
A
9
G VG G A
9
VG G G
A
10
G G G A
10
G VG G
C
2
A
1
G VG G C
4
A
1
G VG MG
A
2
G G MG A
2
MG G G
A
3
G G G A
3
MG G VG
A
4
VG G G A
4
MG G G
A
5
G G VG A
5
G VG G
A
6
VG VG VG A
6
G VG VG
A
7
G MG G A
7
MG G MG
A
8
G G G A
8
G VG VG
A
9
G G VG A
9
MG G VG
A
10
G MG VG A
10
MG G G
Table 4
The nal ranking.
Candidate M
+
1
M
+
2
M
+
3
M
+
4
M
+
5
M
+
6
M
+
7
M
+
8
M
+
9
M
+
10
Value 0.8853 0.8610 0.8502 0.8876 0.9107 0.9197 0.7932 0.9092 0.8853 0.8467
Ranking 5 7 8 4 2 1 10 3 6 9
Fig. 8. . Visual ranking results from PAPR system.
C. Moon et al. / Applied Soft Computing 10 (2010) 512519 518
References
[1] P. Allan, Avoiding common pitfalls in performance appraisal, Industrial Manage-
ment (1992) 3032.
[2] H.G. Chambers, M. Foulon, H. Haneld-Jones, S.M. Hankin, E.G. Michaels III, The
war for talent, McKinsey Quarterly (1998).
[3] G. Lowenberg, K.A. Conrad, Current Perspectives in Industrial/Organizational
Psychology, Allyn & Bacon, Boston, MA, 1998.
[4] A.G. Momeyer, Why no one likes your performance appraisal system? Training 23
(1986) 9598.
[5] G. Dressler, Human Resource Management, 6th ed., Prentice-Hall, Englewood
Cliffs, NJ, 1994.
[6] S.W. Gilliland, J.C. Langdon, Creating performance management systems that
promote perceptions of fairness, in: J.W. Smither (Ed.), Performance Appraisal:
State of the Art in Practice, Jossey-Bass, San Francisco, CA, 1998.
[7] E.W. Duggan, C.S. Thachenkary, Integrating nominal group technique and joint
application development for improved systems requirement determination,
Information & Management 41 (2004) 399411.
[8] J. Perry, S. Linsley, The use of the nominal group technique as an evaluative tool in
the teaching and summative assessment of the inter-personal skills of student
mental health nurses, Nurse Education Today 26 (2006) 346353.
[9] P.L. Williams, N. White, R. Klem, S.E. Wilson, P. Bartholomew, Clinical education
and training: using the nominal group technique in research with radiographers
to identify factors affecting quality and capacity, Radiography 12 (2006) 215224.
[10] A.L. Delbecq, A.H. Van de Ven, Nominal versus interacting group processes for
committee decision making effectiveness, Journal of Academy Management 14
(1971) 203212.
[11] G. Bortolan, R. Degani, A reviewof some methods for ranking fuzzy subsets, Fuzzy
Sets and Systems 15 (1985) 119.
[12] J.J. Buckley, Ranking alternatives using fuzzy numbers, Fuzzy Sets and Systems 15
(1985) 2131.
[13] J.J. Buckley, Fuzzy hierarchical analysis, Fuzzy Sets and Systems 17 (1985) 233
247.
[14] S.H. Chen, Ranking fuzzy numbers with maximizing set and minimizing set, Fuzzy
Sets and Systems 17 (1985) 113129.
[15] L.S. Chen, C.H. Cheng, Selection IS personnel use fuzzy GDSS based on metric
distance method, European Journal of Operational Research 160 (2005) 803820.
[16] C.T. Chen, C.T. Lin, S.F. Huang, A fuzzy approach for supplier evaluation and
selection in supply chain management, International Journal of Production
Economics 102 (2006) 289901.
[17] D. Dubois, H. Prade, The mean value of a fuzzy number, Fuzzy Sets and Systems 24
(1987) 279300.
[18] M. Modarres, S. Sadi-Nezhad, Ranking fuzzy numbers by preference ration, Fuzzy
Sets and Systems 118 (2001) 429436.
[19] E. Triantaphyllou, C.T. Lin, Development and evaluation of ve fuzzy multi-
attribute decision-making methods, International Journal of Approximate Rea-
soning 15 (1996) 281300.
[20] E.S Lee, R.L. Li, Comparison of fuzzy numbers based on the probability measure of
fuzzy events, Computers and Mathematics with Applications 15 (1988) 887896.
[21] C.H. Cheng, A newapproach of ranking fuzzy numbers by distance method, Fuzzy
Sets and Systems 95 (1998) 307317.
[22] C.L. Hwang, K. Yoon, Multiple attribute decision making, Lecture Notes in Eco-
nomics and Mathematical Systems, vol. 198, Springer, Berlin, 1981.
[23] Korean Military, Guide Book for the Human Resource Management in Korea
Military, Korea Military, Seoul, 2005.
[24] F. Herrera, E. Herrera-Viedma, Linguistic decision analysis: steps for solving
decision problems under linguistic information, Fuzzy Sets and Systems 115
(2000) 6782.
[25] C.G.E. Boender, J.G. de Graan, F.A. Lootsma, Multi-criteria decision analysis with
fuzzy pairwise comparisons, Fuzzy Sets and Systems 29 (1989) 133143.
[26] R. Goetschel Jr., W. Voxman, Elementary fuzzy calculus, Fuzzy Sets and Systems
18 (1986) 3143.
C. Moon et al. / Applied Soft Computing 10 (2010) 512519 519