Você está na página 1de 8

A Fuzzy Model for Selection of QoS-Aware Web Services

Ping Wang
1
, Kuo-Ming Chao
2
, Chi-Chun Lo
3
, Chun-Lung Huang
3
, Yinsheng Li
4
1
Department of MIS, Kun Shan University, Taiwan
pingwang@mail.ksu.edu.tw
2
DSM Research Group, School of MIS, Coventry University, UK
k.chao@coventry.ac.uk
3
Institute of Information Management, National Chiao Tung University, Taiwan
{cclo@faculty, clhuang@iim}.nctu.edu.tw
4
Institute of E-Business, Software School, Fudan University, China
liys@fudan.edu.cn
Abstract
With the increasing popularity of employing web
services, quality of service (QoS) measure of service
becomes a significant concern for service consumers
and providers. However, current technologies are
poor to identify the relative importance of non-
functional QoS criteria for web services and they are
not fitting well for selecting the appropriate service. In
this paper, a new scheme for right selection of QoS-
aware of web services which exploits fuzzy logic to
locate and select the right service based on customers
preference or satisfaction degree. The aim is to
compute both functional and non-functional
weightings of QoS criteria and to assist customers to
make the right choice for web services.
1. Introduction
Tim Berners-Lee, the author of book, Weaving the
Web offers insights to understand the impact of world
by web technologies-change the way people do
business, entertain themselves, exchange ideas, and
socialize with one another [23]. There are two dreams
are depicted in this book, one is everyone receives and
shares the information thru the Internet; the other is
people commutates with computers in natural language
on the Internet. The former has happened in our life,
the latter is enabling by the Semantics Web Services
(SWS). SWS technology aims to add enough
semantics to the specifications and implementations of
Web Services to make possible the automatic
integration of distributed autonomous systems, with
independently designed data and behavior models [6].
With the widespread proliferation of electronic
commerce, wherein customers interacts with web
services providers, quality of service (QoS) has
become a significant concern. QoS determines the
service usability and utility, both of which influence
the popularity of the service [2]. In fact, QoS-aware
web service is a significant development tendency for
future web services. It is expected to adopt QoS
criteria, such as high performance and reasonable price,
to assist customers to select the appropriate services.
Referring to Zhous report [5], QoS of web services
compose both functional and non-functional properties;
functional properties is measured in terms of
throughput, latency, response time; non-functional
properties address various issues including integrity,
reliability, availability and security of web services.
The current techniques and tools for QoS measure are
more suitable to quantify functional properties, for
example, network throughput, latency, response time,
which are not fitting well for non-functional properties.
Practically, non-functional properties, such as integrity,
security, availability, might be evaluated considering
on customers QoS preference in a fuzzy way, due to
the non-functional properties are strongly influenced
by customers feeling and experience. Hence the
customers preference on QoS criteria can be
expressed in a fuzzy way and a new QoS-aware web
services analysis model in which the functional and
non-functional QoS requirements are taken into
account for the appropriate selection of required
services.
2. Related Works
IEEE International Conference on e-Business Engineering (ICEBE'06)
0-7695-2645-4/06 $20.00 2006
Many researches [3,4,10,13,15,17,18,19,23] have
studied QoS issues to implement and manage the web
services. The Web Service Description Language
(WSDL) is developed to illustrate how the service is
implemented; Web Ontology Language for Services
(OWL-S) is employed to model QoS items of web
services, which aims to enable automated web service
discovery, invocation, composition, and monitoring.
Current web services researches focus on QoS-aware
web service management with standards, such as Web
Service Level Agreement (WSLA) [15], web service
management language [3], and might neglect QoS
rating problems, which have the significant effect on
the marketing. Competitors would differentiate with
other service providers by offering distinct service
quality.
Few works on selection of QoS-aware web service
have been carried out. Service matchmaking
techniques [4,16,17] based on QoS have developed to
match the needs between consumers and providers.
Zeng [18] addressed the issue of selecting web services
by maximizing user satisfaction expressed as utility
functions over QoS attributes, while satisfying the
constraints set by the user and by the structure of the
composite service; Lin [19] regarded QoS-driven web
service with dynamic composition as a fuzzy constraint
satisfaction problem and applied optimal search
approach with adjustments to web services
composition; Sirine [13] developed a goal-oriented,
interactive composition approach that uses
matchmaking algorithms to help users filter and select
services while building the composition; Ran [23]
developed a service discovery model in which
functional and non-functional requirements are
considered to evaluate QoS metric. Hwang and Chao
[4,10,16] proposed several multi-agent approaches
with fuzzy group decision making methods and
semantic web technologies to assist service providers
and consumers in discovering appropriate services
with consideration of their expectations and
preferences.
However, traditional matchmaking and composition
approaches have no discussed that how to analyze the
relative importance of QoS criteria systemically which
might influence on the customers decision.
Furthermore, linear programming and optimal research
technologies are more suitable to quantify functional
properties, which are not fitting well for measuring the
non-functional metrics, because some of non-
functional properties of QoS are not easy to be
quantified in numerical form. Meanwhile, consumers
QoS preference often remain imprecise, uncertain or
ambiguous due to various human mental states, and the
preferences over the criteria are hard to be quantified
especially in distinguishing the priority order among
QoS criteria. As a result, the traditional approaches
(i.e., linear programming, optimal search) are too
complex to compute in some cases; for example, 1)
many evaluators are involved, this issue becomes a
multi-person decision making problem; 2) The form of
QoS criteria are fuzzy variables which are decided by
customers preference; 3) numerous criteria mutually
influence on each others.
Hence a new web services analysis model in which
both functional and non-functional QoS requirements
are taken into account for the appropriate selection of
required services in this research. This approach can
deal with the consumers imprecision preference
relations with fuzzy sets, determine the weighting of
QoS criteria base on group preferences and assist
service consumers to get dynamic QoS ranking on
available web services.
The contribution of our research is to present a
fuzzy decision making model to locate and select the
right web service based on customers preference or
satisfaction degree. Furthermore, we hope it will be
utilized by independent third-party in an experimental
QoS forum in the Internet to differentiate web service
level (WSL) for web service providers and to assist
consumers to make the right choice. By contrast, the
service providers also can improve their services
referring to evaluation results assessed by customers.
A new resolution process for evaluation of QoS-aware
web service problems based on customers preferences
among a group is introduced in Section 3. Section 4
takes an example to illustrate our method and discusses
the results; Section 5 presents a discussion of
comparison with other approach. Section 6 draws the
conclusion and points out future work.
3. The Proposed Model
In this research, we present a new scheme by
revising a traditional entropy weighting approach,
called Linguistic Entropy Method (LEM), to prioritize
the relative importance of QoS criteria considering
consumers fuzzy preferences and confidence interval
through the use of fuzzy logic as follows:
3.1. Basic definitions and notations
In this section, we review some arithmetic
operations on fuzzy numbers for representing the
proposed algorithm in Sec. 3.3 and Sec. 3.4.
Definition 1. Triangular fuzzy number. A triangular
fuzzy number A
~
can be defined by a triplet ) , , ( c b a .
The membership function is defined as follow [1]:
IEEE International Conference on e-Business Engineering (ICEBE'06)
0-7695-2645-4/06 $20.00 2006

>
s <

s <

<
=
c x for
c x b for
b c
x c
b x a for
a b
a x
a x for
x u
A
0
0
) ( ~
(1)
Definition 2. o -cuts. The o -cuts of triangular
fuzzy number
o
~
A
is defined as follow [1]:
] ) ( , ) [( ] , [
~
c b c a a b a a A
R L
+ + = = o o
o o o
(2)
Definition 3. Fuzzy arithmetic operations. The
arithmetic operations of the positive triangular fuzzy
numbers are expressed below: [14,22]
Addition operation :
) , , ( ) , , ( ) , , (
2 1 2 1 2 1 2 2 2 1 1 1
c c b b a a c b a c b a + + + =
,
Subtraction operation -:
) , , ( ) , , ( ) , , (
2 1 2 1 2 1 2 2 2 1 1 1
a c b b c a c b a c b a =
,
Multiplication operation :
) , , ( ) , , ( ) , , (
~ ~
2 1 2 1 1 1 2 2 2 1 1 1
c c b b a a c b a c b a B A = =
,
) , , ( ) , , (
~
1 1 1 1 1 1
kc kb ka c b a k A k == = , R k e
Division operation / :
) / , / , / ( ) , , /( ) , , (
~
/
~
2 1 2 1 2 1 2 2 2 1 1 1
a c b b c a c b a c b a B A == =
Log operation ln :
). ln , ln , (ln ) , , ln(
~
ln
1 1 1 1 1 1
c b a c b a A = =
(3)
3.2. Problem formulation
Consider the problem of ranking web services
i
s
(i=1,..,m). A committee of decision makers (DMs)
k
d
(k=1,..,q) is formed to identify n evaluation criteria,
say ) ,.., 1 ( n j c
j
= . Each decision maker is to evaluate
the alternatives individually, and then assign
performance rating to the alternatives for each criterion.
The performance ratings assigned by the decision
makers are linguistic terms, as shown in Table 1. In
Table 1, the membership function of linguistic terms
for the rating of each alternative is given by
) 2 , , 2 ( + x x x for
) 7 , 5 , 3 (
~
=
ij
x
; (1,1,3) for 1
~
=
ij
x ;
and (7,9,9) for 9
~
=
ij
x . Next, the linguistic terms are
transformed to triangular fuzzy numbers of the form as
Eq. (1).
Table 1. Linguistic terms for the rating of each
alternative
Linguistic Terms Triangular fuzzy number
Very Poor (VP) ) 3 , 1 , 1 (
Poor (P) ) 5 , 3 , 1 (
Fair (F) ) 7 , 5 , 3 (
Good (G) ) 9 , 7 , 5 (
Very Good (VG) ) 9 , 9 , 7 (
3.3. Weights of QoS criteria
In this article, we introduce the entropy method [20]
which is derived from the classical maximum entropy
theory [12] to the fuzzy environment, called LEM. The
LEM is used to evaluate the fuzzy weights of QoS
criteria. Instead of the Sattys pairwise comparison
matrix [8,25], entropy method has become a significant
concept in social science, information theory as well as
physics sciences. However, it can be only applied in a
crisp form. Hence we attempt to revise it for attaining
the relative importance for each QoS criterion from the
average intrinsic information by a set of alternatives
(web services) using fuzzy terms or fuzzy variables.
Assume that a set of web services
i
s (i=1,..,m) have
the same functional and non-functional properties. A
set of DMs
k
d (k=1,..,q) evaluate service
i
s based on
both crisp and fuzzy data with respect to criteria
j
c
) ,.., 1 ( n j = .The performance rating matrix X
~
is shown
as Eq.(4) [11].
n
c c c ...
2 1

= =
mn m m
n
n
m
ij
x x x
x x x
x x x
s
s
s
x X
~ ~ ~
~ ~ ~
~ ~ ~
]
~
[
~
2 1
2 22 21
1 12 11
2
1

,
(4)
where
)
~
...
~
(
1
~ 1 q
ij ij ij
x x
q
x =
,
k
ij
x
~
represents the
fuzzy rating of web service
i
s with respect to criterion
j
c assessed by
k
d ;
k
ij
x
~
can be crisp or fuzzy form
depending on QoS attribute.
In order to rank the web services compatibly
between evaluation criteria, the matrix X
~
has to be
normalized for transforming the distinct scales of
criteria into a numerically comparable scale. For QoS
criteria, there exist two basic types: utility-oriented and
price-oriented criteria. Both are sometimes mutually
conflict and inconsistent and needs to tradeoff. To
avoid the complicated normalization formula used in
[26], a linear scale transformation is applied to Eq.(4)
for forming the normalized fuzzy rating matrix R
~
as
follows [11].
),
~
,
~
,
~
( ) , , (
~
~
~
,
, ]
~
[
~
u
ij
u
ij
u
ij
j
ij
j
ij
j
ij
j
ij
ij
mxn ij
c b a
c
c
c
b
c
a
x
x
r
criteria utility j
r R
= = =
e
=
+ + + +
(5)
IEEE International Conference on e-Business Engineering (ICEBE'06)
0-7695-2645-4/06 $20.00 2006
)
~
,
~
,
~
( ) , , (
~
~
~
,
p
ij
p
ij
p
ij
ij
j
ij
j
ij
j
j
i
ij
c b a
a
a
b
a
c
a
x
x
r
criteria price j
= = =
e

, , ) , , ( min
~
, , ) , , ( max
~
criteria price j if a c b a x
criteria utiltiy j if c c b a x
where
j ij ij ij
i
j
j ij ij ij
i
j
e = =
e = =

+ +
and u , p represents a set of utility-oriented criteria
and price-oriented criteria, respectively. Once matrix
R
~
is obtained, we calculate the weighting of each
criterion based on the entropy theory as follows:
Referring to the maximum entropy theory, the
entropy measure of the j
th
criterion is given by [6]

=
=
m
i
ij ij j
P P m e
1
ln ) ln( / 1 (6)
where
ij
P
is the probability of information which
can be explained as the relative strength of information
ij
r
~
. In [20],
ij
P
is defined as
,
1

=
=
m
i
ij
ij
ij
r
r
p
(7
)
From the definition of Eq. (7), we see that
ij
P
is
given by the ratio of rating of criterion
j
c to sum of
ratings for all web services; Due to fuzzy feature of
ij
r
~
,
ij
P
has to be transformed into Eq. (8) , that is,
)
~
...
~
/(
~ ~
1 mj j ij ij
r r r p =
(8
)
There are two approaches ( I and II ) to get different
forms of entropy
j
e depending on the form of
weighting, as follows:
Crisp form: Transform
ij
p
~
into crisp value and
substitute it into Eq.(6). Here we select o -cuts as
defuzzification method to form Eq. (9).
), ... /(
1
o o o
mj j ij ij
r r r p + + =
(9
)
where , 2 / ) (
o o o
ijR ijL ij
r r r + = ), (
* * *
ij ij ij ijL
a b a r + = o
o
and ) (
* * *
ij ij ij ijR
b c c r = o
o
[1] , * may represent either
u or p in (5). Then, the normalized crisp weighting of
criterion
j
w
is computed by Eq.(10).

=
m
j
j
j
j
e
e
w
1
) 1 (
1
(10)
Fuzzy form: From Eq.(8), we employ fuzzy rating
ij
r
~
to compute
ij
p
~
and substitute directly
ij
p
~
into
Eq.(6) and generate the fuzzy form of
j
e .
)]
~
ln
~
( ... )
~
ln
~
[(
) ln(
1
~
1 1 mj mj j j j
p p p p
m
e

=
(11)
The normalized fuzzy weighting of criterion
j
w
can be computed by Eq.(12).
)]
~
1 ( .. )
~
1 /[( )
~
1 (
~
1 n j j
e e e w =
(12)
3.4 Fuzzy synthetic evaluation
Once the weighting for each QoS criterion is
identified, the fuzzy simple additive weighting method
[22] is employed to attain the priorities of alternatives,
which derives the synthetic evaluations of alternatives
by multiplying the weighting of a criterion (
j
w
~
) with
the normalized fuzzy rating of alternatives (
ij
r
~
). The
decision matrix Z
~
is given by.
ij
n
j
j ij
r w z Z
~ ~
]
~
[
~
1
= =

=
, i=1,2,,m, j=1,2,,n.
(13)
However, the aggregation results Z
~
are still fuzzy
numbers, which cannot be applied directly to decision
making. The use of interval arithmetic and o -cuts of
fuzzy numbers provides a ranking for the alternatives
and can transform them into numerical values. The
transformation uses the preference values that are
located within a given boundary for which L and R
represents left and right side of interval for fuzzy
number, respectively. This leads to the fact that the less
fuzziness, the larger o value. A crisp value exists
when o =1, that is, L=R.

=
] , [ ] , [
] , [ ] , [
~
1 1
1 1 11 11
o o o o
o o o o
o
mnR mnL R m L m
nR nL R L
z z z z
z z z z
Z

(14)
where
o o o
ijL jL ijL
x w z = ,
o o o
ijR jR ijR
x w z = .
Finally, decision makers can estimate the degree of
satisfaction of the judgment Z by giving the value of
o and the index of optimism

. The index of
optimism

indicates the degree of optimism


(confidence) of a decision maker. A large value
of

implies a high degree of optimism whereas a small


value of

means a low degree of optimism. Thus we


have:
mxn ij
z Z ]
~
[
o
= ,
(15)
IEEE International Conference on e-Business Engineering (ICEBE'06)
0-7695-2645-4/06 $20.00 2006
where Z is a precise judgment matrix, and
] 1 , 0 [ , ) 1 (
~
e + =
o o o
ijR ijL ij
z z z
The decision makers can decide the priority among
the alternatives through the use of
o
ij
z
~
and index of
optimism

. In summary, the procedure for the


proposed method is described as follows:
Step 1. Form a committee, identify the evaluation
criteria, and recognize a set of possible alternatives.
Step 2. Select the appropriate linguistic terms for
representing weights of criteria as shown in Table 1.
Step 3. Evaluate the performance rating of
alternative
i
s for each criterion
j
c by Eq. (4).
Step 4. Calculate the normalized linguistic rating
ij
r
~
by Eq. (5).
Step 5. Determine the entropy of criterion
j
c by
two approaches:
5.1. Employ Eqs.(6)~(9) and generate crisp form of
entropy
j
e , or
5.2. Apply Eq.(11) and obtain fuzzy form of
entropy
j
e
~
.
Step 6. Calculate the weightings of criteria
j
c by
two approaches:
6.1. Use Eq. (10) to resolve crisp form of weighting
j
w , or
6.2. Utilize Eq.(12) to determine fuzzy form of
weighting
j
w
~
.
Step 7. Use Eq.(13) to compute the weighted fuzzy
decision matrix, Z
~
.
Step 8. Calculate the judgment matrix (
o
Z
~
) using
fuzzy interval arithmetic operations via Eq.(14).
Step 9. According to Z derived from Eq.(15), the
decision makers can prioritize the ranking of the
alternatives with the given index of optimism

and
value of o .
4. An example
In this section an example of selecting web services
is used as an illustration of the application of the
proposed method. Three alternatives of web services
for travel agents, ) 3 ,.., 1 ( = i s
i
, are performed by a set
of decision makers
k
d (k=1,..,4) based on five category
and 17 criteria of QoS criteria
) 5 ,..., 1 ( = l C
l
.
The evaluation criteria, excerpted from research
report of Ran [23], were pre-selected for the evaluation
of available web services. The DMs has to perform the
QoS evaluation to all candidates and select the best
one from the three candidates according to the
following five main categories. The hierarchical
structure of this evaluation is shown in Figure 1.
Goal
C

Runntime
QoS
C
2
Transaction
QoS
C
3
Cost
QoS
C
4
CM
QoS
C
5
Security
QoS
S1 S3 S3
Goal
C

Runntime
QoS
C
2
Transaction
QoS
C
3
Cost
QoS
C
4
CM
QoS
C
5
Security
QoS
S1 S3 S3
Figure 1. A hierarchical structure for the
decision problem
The details for these criteria are listed as follows:
(1) Runtime Related QoS (
1
c )-contains 5
performance criteria: Scalability (
11
c ), Capability (
12
c ),
Performance (
13
c ), Reliability (
14
c ), Availability (
15
c ).
(2) Transaction Support Related QoS (
2
c )-consists
of only one criterion: Data Integrity (
21
c ).
(3) Cost Related QoS (
3
c )-focus on Acceptable
Price (
31
c ), It is charged by US dollars ($) /month.
(4) Configuration Management (CM) Related
QoS (
4
c )-this category is judged by Regulatory ) (
41
c ,
Supported Standard (
42
c ), Stability/Change Cycle
(
43
c ), and Completeness (
44
c ).
(5) Security Related QoS (
5
c )-this category
emphases Identity Authentication
51
c , Confidentiality
(
52
c ), accountability (
53
c ), Auditability (
54
c ), Data
Encryption (
55
c ) and Non-Repudiation (
56
c ).
The proposed method is applied to solve this
problem according to the following procedures:
Step 1: Form a committee
} , , , {
4 3 2 1
d d d d D =
, and
determine the evaluation criteria
} , , , , {
5 4 3 2 1
c c c c c C=
and
possible web services
} , , {
3 2 1
s s s S =
.
Step 2: After communicating and discussing, DMs
synthesize all of sub-criteria and use the linguistic
terms and produce fuzzy or crisp performance ratings
of evaluation criteria to each web service as Table 2.
Step 3: Aggregate the individual ratings shown in
Table 3 into group ratings and produce the fuzzy
decision matrix as shown in Table 3.
Step 4: Construct the fuzzy normalized decision
matrix as Shown in Table 4.
Step 5: For sake of gaining fuzzy weights from
Table 4, users may select approach II for calculating
IEEE International Conference on e-Business Engineering (ICEBE'06)
0-7695-2645-4/06 $20.00 2006
the fuzzy entropy of criterion
j
c by Eq (11) and the
result is shown as Table 5.
Step 6: The fuzzy weightings of criteria
j
c can be
carried out using Eq.(12) as shown in Table 6.
Table 2. Ratings by DMS with respect to
criteria
DMs
Criteria
Web
Service
s
1
d
2
d
3
d
4
d
S
1
F VG G G
S
2
G F VG VG C
1
S
3
F G F G
S
1
G F G F
S
2
VG G G G C
2
S
3
G G F G
S
1
5.0 5.5 7.0 6.0
S
2
8.5 7.5 6.5 6.5 C
3
S
3
4.5 4.0 7.5 3.5
S
1
G F G F
S
2
G G P F C
4
S
3
P G F F
S
1
F G G P
S
2
F G G VG C
5
S
3
P F G G
Table 3. Fuzzy decision matrix
C
1
C
2
C
3
C
4
C
5
S
1
(5.00, 7.00, 8.75) (4.00, 6.00, 8.00) (5.88, 5.88, 5.88) (4.00, 6.00, 8.00) (3.50, 5.50, 7.50)
S
2
(5.50, 7.50, 9.00) (5.50, 7.50, 9.25) (7.25, 7.25, 7.25) (3.50, 5.50, 7.50) (5.00, 7.00, 8.75)
S
3
(4.00, 6.00, 8.00) (4.50, 6.50, 8.50) (5.13, 5.13, 5.13) (3.00, 5.00, 7.00) (3.50, 5.50, 7.50)
Table 4. Fuzzy normalized decision matrix
C
1
C
2
C
3
C
4
C
5
S
1
(0.56, 0.78, 0.97) (0.43, 0.65, 0.86) (0.87, 0.87, 0.87) (0.50,0.75,1.00) (0.40, 0.63, 0.86)
S
2
(0.61, 0.83, 1.00) (0.59, 0.81, 1.00) (0.71, 0.71, 0.71) (0.48,0.69,0.94) (0.57, 0.80, 1.00)
S
3
(0.44, 0.67, 0.89) (0.49, 0.70, 0.92) (1.00, 1.00, 1.00) (0.38,0.63,0.88) (0.40, 0.63, 0.86)
Table 5. Fuzzy entropy of the criteria
C
1
C
2
C
3
j
e
~
(0.9989,0.9962,0.9923) (0.9984,0.9961,0.9920) (0.9985,0.9973,0.9940)
C
4
C
5
j
e
~
(0.9986,0.9975,0.9938) (0.9975,0.9938,0.9863)
Table 6. Fuzzy weights of the criteria
C
1
C
2
C
3
i
w
~
(0.090,0.200,0.208) (0.129,0.206,0.216) (0.041,0.141,0.477)
C
4
C
5
i
w
~
(0.107,0.131,0.167) (0.197,0.322,0.369)
Step 7: Construct the weighted fuzzy decision
matrix by employing fuzzy weighting additive rule.
This step generates necessary values to form Table 7.
C
1
C
2
C
3
C
4
C
5
S
1
(5.00, 7.00, 8.30) (2.67, 5.00, 7.00) (2.67, 5.00, 7.00) (1.00, 3.00, 5.00) (1.00, 3.00, 5.00)
S
2
(5.00, 7.00, 8.30) (6.33, 8.67, 9.00) (6.33, 8.67, 9.00) (1.67, 3.67, 5.67) (1.67, 3.67, 5.67)
S
3
(3.67, 5.67, 7.67) (4.33, 6.33, 8.33) (4.33, 6.33, 8.33) (1.00, 2.33, 4.33) (1.00, 2.33, 4.33)
Step 8: Calculate the judgment matrix (
o
V
~
) using
Eq. (14) as follows:
o
o
216 . 0 458 . 0
11
+ =
L
z ,
o
o
229 . 0 903 . 0
11
=
R
z
and
]. 229 . 0 903 . 0 , 216 . 0 458 . 0 [ ] , [
11 11 11
o o
o o o
+ = =
R L
z z z
Other
elements of the judgment matrix can be retained by
following the same procedures.
Step 9: We can attain a set of crisp score for each
alternative through the selection of o and index of
optimism

using Eq. (15). Three kind of optimism


(pessimistic, moderate, and optimistic decision makers;

=0, 0.5, 1.0, respectively) DMs are considered and


its corresponding o =0.05, to construct the matrix of
integrated judgments shown in Table 8.
Table 8. Weighted fuzzy decision matrix
Pessimistic DM

=0
Moderate DM

=0.5
Optimistic DM

=1.0
S
1
0.468 0.680 0.891
S
2
0.498 0.729 0.960
S
3
0.452 0.659 0.866
Furthermore, varying o from 0.0 to 1.0, the
experimental results indicate that the ranking order of
three web services will be consistent with different
confidence level of DMs.
From Table 8, a decision of selecting
2
s can be
reached as its score is greater than the other two. The
order of the preferences of DMs on three web services
can be stated as
3 1 2
s s s > > regardless the value of
optimism

and o .
5. Discussion
For the purpose of comparison, we will use
Zelenys method [8,19] to treat the same problem.
From Table 4, the weightings of criteria are obtained
by setting =1.0, as shown in Table 9. By applying (2-
20)~(2-23) in [8] (p.53~p54), we deduce the
weightings for all criteria
5 ,... 1 , ] 246 . 0 , 094 . 0 , 357 . 0 , 158 . 0 , 145 . 0 [ ] [ = = = j w W
j
.
Table 9. The weightings of evaluation criteria
1
s
2
s
3
s
Criteria
Weightings
1
c 0.78 0.83 0.67 0.145
2
c 0.65 0.81 0.70 0.158
3
c 0.87 0.71 1.0 0.357
4
c 0.75 0.69 0.63 0.094
5
c 0.63 0.80 0.63 0.246
IEEE International Conference on e-Business Engineering (ICEBE'06)
0-7695-2645-4/06 $20.00 2006
The crisp rating (
ij
r ) is obtained by setting o =1.0
and referencing the averaged values assessed by four
decision makers in Table 2.
.
50 . 5
00 . 7
50 . 5
00 . 5 88 . 4 50 . 6 00 . 6
50 . 5 88 . 6 00 . 7 50 . 7
00 . 6 88 . 5 00 . 6 00 . 7
] [

= =
ij
r D
,
(16)
Here we use a Simple Additive Weighting (SAW)
method to treat the synthetic evaluation. Therefore the
outcome is

=
= = =
5
1
]. 46 . 5 89 . 6 98 . 5 [ . ] [
j
ij j i
r w z Z
,
(17)
Obviously, the applicant 2 is the best choice and the
ranking order is
3 1 2
s s s . The solution of
Zelenys method concludes the same result as our
proposed model. From Table 9, the traditional entropy
method is not able to model the uncertainty of the
decision makers judgment. In addition, the synthetic
evaluation is neither flexible nor can it illustrate the
degree of optimism of the decision makers. We can
conclude that the proposed method is more flexible
than the traditional entropy method when QoS criteria
or score are fuzzy variables.
6. Conclusions
This paper presents a fuzzy model to solve the
selection of QoS-aware of web services problems. The
proposed linguistic entropy method could objectively
analyze the weights of QoS criteria from the evaluation
information and is capable of dealing with incoherent
subjective judgments in the group and prioritizing
alternatives. Consequently, our approach enables
customers to get a dynamic ranking of the available
web services with respect to various QoS criteria, in
order to select the services of market place. The future
work in this research will focus on evaluating the
performance of service discovery in multi-agent
systems. This is particularly relevant to situations
where autonomous and distributed agents, with
automatic negotiations capability, need to reach an
assigned mission.
References
[1] A. Kaufmann, and M. M. Gupta, Intro. to Fuzzy
Arithmet. Theory and App., New York, 1991, pp. 235 and
6972.
[2] A. Mani, and A. Nagarajan, "Understanding quality of
service for web services," http://www-128.ibm.com/library/
developerworks/ws-quality.html, Jan 2002.
[3] A. Sahai, A. Durante and V. Machiraju, "Towards
Automated SLA Management for Web Services,"
http://www.hpl.hp.com/techreports/2001/HPL-2001-310
R1.pdf, 2001.
[4] C-L Huang, K-M Chao and C-C Lo, "A Moderated Fuzzy
Matchmaking for Web Services," Proceedings of the Fifth
International Conference on Computer and Information
Technology (CIT'05), IEEE CS, Shanghai, China, Sep. 2005,
pp.11161122.
[5] C. Zhou; L-T Chia and B-S Lee, "Semantics in service
discovery and QoS measurement," IT Professional, Vol. 7,
Issue. 2, MarApr 2005, pp.2934.
[6] C. Jorge and S. Amit P. (Eds.), "Semantic Web Services,
Processes and Applications," Ch.3, Springer, 2006
[7] C.E. Shannon and W. Weaver, The Math Theory of
Communica., the University of Illinois Press, Urbana, 1947.
[8] C.H. Cheng, "Evaluating Naval Tactical Missile Systems
by Fuzzy AHP based on the Grade value of membership
function," European Journal of Operation Research, Vol. 96,
1996, pp. 343350.
[9] C.L. Hwang and K. Yoon, Multiple Attribute Decision
Making: Methods and Applications, Springer-Verlag, Berlin,
Heidelberg, New York, 1981, pp. 4158 and 153154.
[10] C-L, Huang, C-C Lo, Y Li, K-M Chao, J-Y Chung, and
Y Huang, "Service discovery through multi-agent
consensus," Proceedings of IEEE International Workshop on
Service-Oriented System Engineering (SOSE 2005), IEEE
CS, Oct. 2005, pp.3744.
[11] C.T. Chen, "Extensions of the TOPSIS for Group
Decision-Making under Fuzzy Environment," Fuzzy set and
Systems, Vol.114, 2000, pp. 19.
[12] E.T. Janes, Info. Theory and Statistical Mechanics,
Physical Review, vol. 106, No. 4, 1957, pp. 620630.
[13] E. Sirine, B. Parsia and J. Hendler, "Filtering and
selecting semantic web services with interactive composition
techniques," IEEE Intelligent Systems, 2004, pp.4249.
[14] G. L. Klir and B. Yuan., Fuzzy sets and fuzzy logic,
Prentice Hall, Singapore, 1988, pp. 103107.
[15] H. Ludwig, A. Keller, A. Dan, R. P. King, and R.
Franck, "Web Service Level Agreement (WSLA) Language
Specification v1.0," http://www.research.ibm.com/wsla/
WSLASpecV1-20030128.pdf, January 2003.
[16] K-M Chao, M. Younas, C-C Lo, and T-H Tan, "Fuzzy
Match- making for Web Services," Proceedings of 19 IEEE
Conf. on Advanced Network and Information Application,
IEEE CS, 2005, pp.721726
IEEE International Conference on e-Business Engineering (ICEBE'06)
0-7695-2645-4/06 $20.00 2006
[17] K. Hogg, P. Chilcott, M. Nolan, and B. Srinivasan, "An
Evaluation of Web Services in the Design of a B2B
Application," Proceedings of the 27th Conf. on Australasian
Computer Science, ACM, Vol. 56, 2004, pp. 331340.
[18] L. Zeng, B. Benatallah, M. Dumas, J. Kalagnanam, and
Q.Z. Sheng, "Quality Driven Web Services Composition,"
Proceedings of 12
th
International Conference of WWW,
Budapest, Hungary, May 2003, pp. 411421.
[19] M. Lin, J. Xie, H. Guo, and H. Wang, "Solving QoS-
driven Web service dynamic composition as fuzzy constraint
satisfaction," Proceedings of the 2005 IEEE International
Conference on e-Technology, e-Commerce and e-Service
(EEE '05), 29
th
Mar~1
st
Apr. 2005. pp. 914.
[20] M. Zeleny, Multiple Criteria Decision Making,
Mcgraw-Hill, 1982, pp. 185198.
[21] R.R. Yager, "Modeling prioritized multi-criteria
decision making," IEEE Trans. Systems, Man and
Cybernetics-Part B, Vol. 34, Dec. 2004, pp. 23962404.
[22] S-H, Chen and C.L. Hwang, Fuzzy Multiple Attribute
Decision Making Methods and Applications, Springer-Verlag,
Berlin, New York,1992, pp. 90 and 292323.
[23] Shuping Ran, "A Model for Web Services Discovery
with QoS," ACM SIGecom Exchanges, March 2003, 4(1), pp.
110.
[24] Tim Berners-Lee, "Weaving the Web," Harper San
Francisco, 1999.
[25] T.L. Satty, The analytic Process, McGraw Hill, 1980.
[26] Y. Liu, H.H. Ngu and L. Zeng, "QoS Computation and
Policing in Dynamic Web Service Selection," Proceedings of
13th Int Conf. World Wide Web 2004, May 2004, pp.6573.
IEEE International Conference on e-Business Engineering (ICEBE'06)
0-7695-2645-4/06 $20.00 2006

Você também pode gostar