Professional Documents
Culture Documents
Kef2 Multicriteria
Kef2 Multicriteria
( )
, .
, :
().
() .
- (
) ( )
.
ai ()
a1
a2
s ()
v(a1,s)
v(a2,s)
am
v(am,s)
, .
2
-1-
!
.
( )
(objective function), 1
"" .
(optimal solution).
,
.
,
.
(,
, , .)
-2-
(ill-formulated)
.
- .
.
3 (Vansnick-1990):
,
"" .
2
-3-
*
1.
2.
3.
4.
( )
( )
.
* Multi-criteria decision analysis (MCDA), sometimes called multi-criteria decision making (MCDM), is a discipline
aimed at supporting decision makers faced with making numerous and sometimes conflicting evaluations.
The International Society on Multi-criteria Decision Making is a professional society of researchers and practitioners in
the field.
2
-4-
1.
, , , .
( )
.
:
(explicitly), (
implicit constraints)
(implicitly), (,
explicit constraints)
(Multi Objective Mathematical Programming MOMP)
( )
: 2
(stable):
.
(evolutive):
.
(globalized):
(fragmented): (
).
-5-
2. / (attributes/criteria)
(attributes)
(criteria) .
(.. , )
:
Top-Down: :
(compound attributes):
(basic attributes):
Bottom-Up: ,
,
.
-6-
( Top-Down)
1
()
- 1.1
-
1.1...1 ()
N
()
- 1.m
-
1.1...1 ()
2
()
-
1.1...1 ()
-
1.1...1 ()
-7-
,
,
.
: .
.
:
.
-8-
3.
( )
. :
,
,
(performance vector) .
..
(, , ) (200, 120, C)
-9-
4.
.
:
-10-
.
, , .
WAS (Weighted Average Sum -
), ELECTRE ( ELECTRE) AHP (Analytic
Hierarchy Process ).
, ,
,
.
.
2
-11-
1/2
MCDA / MCDM. :
-12-
2/2
They all claim that they can accurately solve this type of problem. However,
oftentimes different methods may yield different results for exactly the same (even
simple) problem.
This leads to a decision making paradox
Choosing the best MCDA / MCDM method is itself a multi-criteria decision making
problem, in which the alternatives are the methods themselves and the decision criteria are
the various evaluative ways for comparing them.
The choice of which model is most appropriate depends on the problem at hand and may
be to some extent dependent on which model the decision maker is most comfortable with.
The role of rank reversals in decision making when these methods are used on certain test
problems, plays a crucial role in this regard.
. >>C.
( , , D. . >D).
(. ) ,
.
-13-
, ..
(management information systems).
, :
""
.
"" .
" "
, .
/
(
,
)
, 3 ,
1, 2 3.
-14-
1
1.1
1.2
1.3
2
2.1
2.2
3
3.1
3.2
3.3
3.4
WAS ( ) AHP (
)
(.. ELECTRE-II)
(.. ELECTRE-II)
/
1
1.1
1.2
1.3
2
2.1
2.2
3
3.1
3.2
3.3
3.4
-1
()
-2
()
0-10
0-10
0-10
0-10
0-10
> >
> >
0-10
0-10
0-10
0-10
> >
> >
> >
> >
-15-
1.
-16-
-1
3/7
1.1
3/6
0-10
1.2
2/6
0-10
1.3
1/6
0-10
2/7
2.1
2/5
0-10
2.2
3/5
0-10
2/7
3.1
1/5
0-10
3.2
1/5
0-10
3.3
2/5
0-10
3.4
1/5
0-10
-17-
P1
P2
P3
8
7
8
7
8
6
6
5
6
9
6
4
5
6
5
8
7
9
6
9
4
5
2
7
6
5
5
/
1
(*)
P1
7.66 *
7.20
7.80
7.56 **
P2
7.16
4.60
5.00
5.81
P3
5.66
5.40
5.60
5.56
. :
8*3/6 + 7*2/6 + 8*1/6 = 7.66 ( )
-18-
2. (ELECTRE)
" ", :
,
.
,
' .
, .
""
.
2
-19-
-2
P1
P2
1.1
1.2
> >
1.3
> >
2.1
> >
2.2
> >
3.1
> >
3.2
> >
3.3
> >
3.4
> >
P3
:
.
2
-20-
- ELECTRE II (1/4)
Si(Px, Py) Px Py i.
: 1.1 () 12. S1.1(P1, P2).
S1.1(P1, P2),
S1.2(P1, P3),
S1.3(P1, P2),
S2.1(P1, P2),
S2.2(P1, P2),
S3.1(P1, P2),
S3.2(P1, P2),
S3.3(P1, P2),
S3.4(P1, P2),
S1.1(P1, P3),
S1.2(P2, P1),
S1.3(P1, P3),
S2.1(P1, P3),
S2.2(P1, P3),
S3.1(P1, P3),
S3.2(P1, P3),
S3.3(P1, P3),
S3.4(P1, P3),
S1.1(P2, P1),
S1.2(P2, P3)
S1.3(P2, P3),
S2.1(P2, P3),
S2.2(P2, P3),
S3.1(P2, P1),
S3.2(P3, P1),
S3.3(P2, P3),
S3.4(P2, P3),
S1.1(P2, P3)
S1.3(P3, P2)
S2.1(P3, P2)
S2.2(P3, P2)
S3.1(P2, P3)
S3.2(P3, P2)
S3.3(P3, P2)
S3.4(P3, P2)
" "
, " " "
". .
2
-21-
- ELECTRE II (2/4)
:
w( g : S ( Px, Py))
c
w( g )
i, j
i, j
i, j
gi,j gi c .
, "" Px Py,
gi, :
gi,j Px Py,
... gi,j
... c.
-22-
- ELECTRE II (3/4)
= 0.6 .
S1(P1, P2).
S1.1, S1.2, S1.3 .
P1 P2 S1.1
S1.3 ( 19).
( 18): 3+1=4
3+2+1=6
4/6=0.66 > 0.6, S1(P1, P2).
:
S1(P1, P2),
S2(P1, P2),
S3(P1, P2),
S1(P1, P2) 1 1 ()
2.
2
-23-
- ELECTRE II (4/4)
,
:
S(P1, P2), S(P1, P3), S(P2, P3)
()
,
.
S(P1, P2), S(P1, P3), S(P2, P3)
P1 2 (). : (P1) = 2-0 =2
: (P1) = 2, (P2) = 0, (P3) = -2
-24-
1
.
, C.
,
.
.
:
1)
2) .
, .
3) .
4)
5) () .
6) .
7) .
8) .
-25-
1)
2)
3)
4)
: , .
, ( ).
, 0..5
, .
/
1
1.1
1.2
2
2.1
2.2
3
3.1
3.2
[0..5]
[0..5]
[0..5]
[0..5]
[0..5]
[0..5]
2
3
1
1
1
1
2
3
1
0.4
0.75
0.25
0.2
0.5
0.5
0.4
0.75
0.25
2
2
2
2.5
2
3
3
3
3
4
4
4
3
3
3
2.5
2
4
4.75*
5
4
3.5
2
5
2.25
2
3
.
, .. ,
. , "",
"" "", 5, :
2/5=0.4, 1/5=0.2 2/5=0.4 . - .
-26-
1 (...)
5)
6)
7)
8)
.
(. , ).
, "" :
0.75*5 + 0.25 * 4 = 4.75
, :
= 0.4 * 2 + 0.2 * 2.5 + 0.4 * 3 = 2.5
= 0.4 * 4 + 0.2 * 3 + 0.4 * 2.5 = 3.2
C= 0.4 * 4.75 + 0.2 * 3.5 + 0.4 * 2.25 = 3.5
0.2
: C > > .
0.5 : C = >
-27-
2
ELECTRE . .
1)
2)
3)
4)
5)
6)
7)
.
. ,
( ,
).
.
.
.
.
.
: ,
. ,
( ) (.. Excel).
2
-28-
1)
2)
.
.
2
1.1
3
, , ,
1.2
1
, ,
2
1
2.1
1
, , ,
2.2
1
, ,
3
2
3.1
3
[0, 106], = 20.000
3.2
1
[0, 105], = 1.000
3)
80.000
15.000
120.000
8.000
110.000
14.000
, .
S1.1(B,A), S1.1(C,A), S1.1(C,B)
S1.1
A
B
C
S1.2
A
B
S1.2(B,A), S1.2(B,C), S1.2(C,A), S1.2(C,B)
S2.1(A,C), S2.1(B,A), S2.1(B,C), S2.1(C,A)
0
0
0
A
A
S2.2(A,B), S2.2(B,A), S2.2(C,A), S2.2(C,B)
3
0
1
B
B
S3.1(A,B), S3.1(A,C), S3.1(B,C), S3.1(C,B)
3
3
1
1
C
C
S3.2(A,C), S3.2(B,A), S3.2(B,C), S3.2(C,A)
-29-
C
0
1
4)
5)
(...)
0.6
(. 1 (S1, S2, S3))
S1(A,B)= 0 /(3+1) = 0 < 0.6,
S1(A,C) = 0 / (3+1) = 0 < 0.6,
S1(B,A) = (3+1) / (3+1) = 1 >0.6 ,
S1(B,C) = 1 / (3+1) = 0.25 < 0.6,
S1(C,A) = (3+1) / (3+1) = 1 >0.6 ,
S1(C,B) = (3+1) / (3+1) = 1 >0.6 ,
S1
A
O :
S1(B,A), S1(C,A), S1(C,B),
S2(B,A), S2(C,A),
S3(A,B), S3(A,C), S3(B,C), S3(C,B)
2
-30-
0
0
6)
2 (...)
:
S(A,B)= 2 / (2+1+2) =0.4,
S(A,C)= 2 / (2+1+2) =0.4,
S
A
B
S(B,A)= (2+1) / (2+1+2) =0.6,
S(B,C)= 2 / (2+1+2) =0.4,
0
0
A
S(C,A)= (2+1) / (2+1+2) =0.6,
1
0
B
S(C,B)= (2+2) / (2+1+2) =0.8,
:
S(B,A), S(C,A), S(C,B)
7)
+
0
1
2
, ,
:
A = 0 2 = 2
B=11=0
C=20=2
: C > B > A
:
.
2
-31-
3.
Analytic Hierarchy Process AHP
4 :
(),
(),
()
().
wi
,
WAS.
2
-32-
WAS
WAS, .
AHP :
AHP :
, marketing, , ,
AHP
'
( ).
;
2
-33-
(Pairwise Comparison Matrix A)
n , nxn,
aij i j.
1-9,
:
aij
1
3
5
7
9
2, 4, 6, 8
1
1 / 5
1 / 2
1 / 4
5 2
4
1 1 / 2 1 / 2
2 1
2
2 1/ 2 1
i j .
i j.
i j.
i j.
i j.
.. 4
(3) (5) (. )
: () (5)
(), a12=5
i=j aij=1 ( ).
2
-34-
() a32=2, 1 / 4 2 1 / 2
1
() 2
().
() 2x2=4
(), a12=4 a12=5.
( ) .
, a13=2
a32=8 a12=16 ().
, .
, CI (Consistency Index),
.
( CI )
2
-35-
AHP Saaty
: AwT=wT
(1) :
n x n .
n
w1 w1
T
w (: transposed).
w1 w2
wi i.
w2 w2
A=
w1
wn
w1
w
w
w
w
1
(1)
""
wn wn
:
w2
wn
(1)
w=[w1, w2, ..., wn].
'
(1) w'.
Saaty ,
' n w' w.
-36-
AHP w
'
, Saaty
, w'
:
1:
sum
1
1/5
1/2
1/4
1.95
5
1
2
2
10.00
2
1/2
1
1/2
4.00
A norm
4
1/2
2
1
7.50
0.513
0.103
0.256
0.128
0.500
0.100
0.200
0.200
0.500
0.125
0.250
0.125
0.533
0.067
0.267
0.133
.. 0.128=(1/4) /1.95
2: w' (
w) :
wi i Anorm.
= 0.5115
.. w1 =
4
: w' = [w1, w2, w3, w4] = [0.5115, 0.0986, 0.2433, 0.1466]
, w1
()
norm.
2
-37-
AHP -
(. )
CI (Consistency
Index) :
wT:
1
1 / 5
=
1 / 2
1 / 4
4 0.5115 2.0775
1 1 / 2 1 / 2 0.0986 0.3959
=
2 1
2 0.2433 0.9894
2 1 / 2 1 0.1466 0.5933
5
:
1 n AwT i
1 2.0775 0.3959 0.9894 0.5933
T n
+
+
+
4
n
w
4 0.5115 0.0986 0.2433 0.1466
i
4.05 4
i =1
=
CI =
=
= 0.017
4 1
3
n 1
CI RI (Random Index) n.
n
RI
2
0
3
4
5
6
7
8
9
10
0.58 0.9 1.12 1.24 1.32 1.41 1.45 1.51
RI CI aii=1.
CI=0.
CI/RI<0.1 , AHP .
: n=4, CI=0.017 CI/RI=0.017/0.9=0.019 < 0.1 . .
-38-
AHP- 1
.
"" si (score)
.
() .
.
,
, 1 (
) 3, ...
E1 E2
E1 1
2
E2 1 / 2 1
E3 1 / 4 1 / 2
E3
4
2
1
wmax
:
s1 =
norm
() ().
-39-
AHP- 2
, .
:
E1
E2
E3
E1 E2 E3
1 1/ 2 1/ 3
2 1 1/ 3
3 3
1
E1 E2
E3
E1 1 1 / 7 1 / 3
E2 7 1
3
E3 3 1 / 3 1
E1
E2
E3
E1 E2 E3
1 1/ 4 1/ 7
4 1
2
7 2
1
, s,
s, sAE sEO. :
s=[0.159, 0.252, 0.589]
s=[0.088, 0.669, 0.243]
s=[0.069, 0.426, 0.506]
WAS.
-40-
AHP -
WAS
: (. 33) :
(w') w (. 35)
(. 36)
(score) (. 37, 38)
WAS ( ):
0.571
0.159
ScoreE = w A = [0.5115 0.0986 0.2433 0.1466]
0.088
0.069
0.286
0.252
0.669
0.426
0.143
0.589
= [0.339 0.396 0.265]
0.243
0.506
2 :
(0.5115 0.286) + (0.0986 0.252) + (0.2433 0.669) + (0.1466 0.426) = 0.396
: E2 score (0.396).
Wayen L. Winston, "Operations Research Applications and Algorithms", Duxbury Press, 1994.
2
-41-
, ,
.
" ".
:
()
`
( )
2
-42-