Get documents about "NEURO-FUZZY INFERENCE SYSTEM APPLICATION FOR CREDIT RATING OF BANK
Document Sample
1 - 21 > / 89 / 92 / @> >
@ >@ - @ >
@
@
@ >@ >
D
mahdiyeh_akhbari@yahoo.com
SI
>< >
farimah@cc.iut.ac.ir
89/3/5 : 87/7/28 : >
of
> @
@ @ .
@ @ . >@
ive
> @ @ >@ @ > > >
@ .> > > > @
> @ <@ @
> @ < >. @ @
>. > @ >< > @
ch
@ @ - @ > @ @
< @> @ .
> <@ >
Ar
@ @ >> .
@ > 69/36 > <1385-1380 @
.
G52 : JEL @
< @ > <@ @ < @ - :
@ >
www.SID.ir
89 / 92 / @> > @ 2
-1
> @ > >
. @ <> @
@ @ @ <
@ @ > <> > < .
<>
> @ . @
@ <@ > @ >
D
. > > @
@ > @ @ >< > >< >
< >
> @
<@
@
@
@ SI @
<
@
>
>
.
of
@ @ @ > >
< > @ >@ @
@ @ @ @ >
ive
@ > @ > > >@ > <@
.[6 5 <4] @
@ @ > > > @
@ <@ > <> @ @
ch
@ - @ @ <@
@ @ > > < >.
@ >> > >
Ar
.> > @ > @ >@ <
.> @ >
@ @ > @ -2
@ @ >@ @ @
> > @ > > @ > <
www.SID.ir
3 …@ @ >@ - @ >
>@ @ @ .[5 <4]>
< @ < > >
> @ @ >@ @ @ <> [6]
. > > >
> @ @ <[7]
@ @ [8] . > @
@ > <> @ @ -
D
@ @ – @ >> .
@ > > < > <
. >
١
@ . >
@
@
>
SI <[9]
@ (ANN)
of
٢
@ - @ > . >
@ > > > @ <(ANFIS)
@ ANN >> > > ( / >)
> >@ @ @ > >@
ive
) @ > ANN >> . >
@ > <> > @ @ @ ( >
. @ @ ( > )
ch
(ANFIS) @ - @ <[10]
> @ > . > (MDA) ٣
@
( @ 250 @ 250) 500 @ > > @
Ar
> > ( ) ( )@ > @ > > < > >
@ @ > > . >
.> MDA ANFIS
1- Artificial neural networks.
2- adaptive network based fuzzy inference system or adaptive neuro-fuzzy inference system.
3- Multiple Discriminate Analysis.
www.SID.ir
89 / 92 / @> > @ 4
<@ @ > @ > >
١
@ @ .> @
< >. > @ @ @ > (FAN)
@ > < @
@ @ @ @ >
@ > . @ >
< >@ . >
D
> > . < @ >
. > > @ > @ >
.> @ @ > >
< @>
> @ > @ < SI @ @ >
.[11]
of
@ <@ @ @ > >
> <> @
@ @ < @
.[12] %80 >
ive
@ - > @
@
ch
.> - @ @
A < B A > @ > - @
>. @ @
Ar
B
> > > >@ @ @
> @ @ < @ >
: > @ .[3 1<2]
. < @ >
1- fuzzy adaptive network.
www.SID.ir
5 …@ @ >@ - @ >
< @ <@ >
> @
.
(1983) @ - @ >
> .> @ @ > < >
< @ -
:>
D
*@ > +β * +γ < @ >
. =α
> @ ( ) > >
@
@ >
@
@
>@ >
@
>
@ >
SI
@
>
-
.
>
of
@ .> >
. > @
> (@ - > > )@ @
: > @
ive
> <( ) > @> @ @ -1
@ @ (@ ) @ @
.(> @ @ < ) >
ch
> ( @ < -T > ) -2
> ( ) @ <( ) >
.
Ar
@ ( @ > ) -3
. @
@ @ < ) > > -4
.(>
www.SID.ir
89 / 92 / @> > @ 6
) - > < >
> .> > > <(>
. > <
@ @ >
> .[13<14] >@ @
< > > >
> >. @ > @ @
D
< @ > > @ @ <
. < > @ > > @
> @
.
<
<
SI ١
>
@
@ >
of
> @ > @ .
.> > @
٢
> @ @
ive
. > (1974) ٣
< [16 <15 <13] <> @ @ >
.>
ch
>
X1
>
Y1
@>
Ar
X2 Y2
@
1- adaptive networks.
2- gradient descent.
3- chain rule.
www.SID.ir
7 …@ @ >@ - @ >
< (1 )
@ . @> @ @ (
١
)
-@> < >@
. <> > @
> < > < @ > > @
>< @ > (٢ ) .> >
< @ @ @ .> @ (
٣
)@ >
D
. @ @ @
@ -1-2-3
@ < -@> >
> >
>.
< >
< nk @ >
SI (
.
<
)@ > @ >>
<>>
@
of
k L
>> O ik ( ) (k,i) k > i
: @ @> @ Oi .
k
Oi = Oi (O1 −1,...Ok −1, a , b, c,...), (1)
ive
k k k
n
@ c <b <a
k −1
k
Oi .
.
< >> P ( ) @ > @ >> @
ch
> @ > @ > > @> (1 ≤ p ≤ P ) p@ (٤@ )
:> >
∑ (Tm, p − OL ,p )2 ,
nL
(2)
Ar
Ep = m
m =1
m < OL ,p
m > p m <Tm,p
1- node function.
2- adaptive node.
3- fixed node.
4- energy function.
www.SID.ir
89 / 92 / @> > @ 8
@ . @> > p@ >
.E = ∑ Ep :
P
p =1
∂E p
@ > @ >> p@ < < > < > @ >
∂O
> (L , i) @ > .> O@ @
:>
D
∂E p
= −2(Ti, p − Oi, p ).
L (3)
∂O i , p
L
:
∂E p
∂Oik p
=
@
∑
n k +1
m =1 ∂O m, p ∂Oi, p
k +1
∂E p ∂O k +p
k
@
m,
1
>
,
SI > <(k,i) @
(4)
of
.1 ≤ k ≤ L − 1
,
@ @ > > @ >
< 1 ≤ i ≤ nk 1 ≤ k ≤ L @ .> @ @
ive
∂E p
.> > (4) (3) > > >
∂O i , p
k
: > < @ @ α
ch
∂E p ∂E p ∂O*
∂α
= ∑ ∂O* ∂α
, (5)
O * ∈S
. α @
Ar
S
: α E @
∂E P ∂E p
=∑ . (6)
∂α p =1 ∂α
: > α @ <
∂E
∆α = − η (7)
∂α
www.SID.ir
9 …@ @ >@ - @ >
:> > @ > > η
η= (8)
k
,
∑α ( ∂α )2
∂E
<@ . @ @ > <١ @ k
. > <k >
@ ٢
> .> > > @ @ >
D
> (6) > α > <٣
٤
> @ > @ >> @
>< >@ .>
@ >
٧
-@ ٦
<
SI < -@>
>
.[15< 14]
٥
@
(5)
of
@ - @ @ -1-2-3
ANFIS > <[20] < @
(y x) @> > @ > (2 )
>< .> (z)
ive
> > <> > @ ANFIS
: > >
if x is A1 and y is B1 , then f1 = p1x + q1y + r1 1@
ch
if x is A 2 and y is B 2 , then f 2 = p2 x + q2 y + r2 2@
( <4) > @
: >> ( <4) >
Ar
ANFIS
1- step size.
2- batch learning.
3- off-line.
4- epoch.
5- sweep.
6- pattern learning.
7- on- line.
www.SID.ir
89 / 92 / @> > @ 10
D
SI
of
@ - -1
ive
(@> @ ) @
: ( ) < @ >i
O1, i = µ A i (x ), i = 1,2, or
(9)
ch
O1, i = µ Bi −2 (x ), i = 3,4.
@ > Ol,i ( ) @ B A
@ > @ > >. l@ > i
. > @
Ar
@ > @> @ > >
> > @ @ (1 0 @> )
@ < < < @ > .
.
@> @ @ > > >
.> > @ @
www.SID.ir
11 …@ @ >@ - @ >
(@ @ @ ) >@
> (@ >) @
: > @
O2, i = w i = µ A i (x ) * µ Bi (x ), i = 1,2. (10)
@ and T-norm > < @ >
@ >. > @ > < >
. > @ ( )
D
(@ @ ) @
@ > @ > > >
O 3,i = w i =
wi
w1 + w 2
:
, i = 1,2.
>
SI(@ @
@
@ )
(11)
@
of
@ . @ @
> @ > @
: >
ive
O4, i = w i f i = w i ( p i x + q i y + ri ), for i = 1,2 (12)
wi @ @ <ri qi <pi
.> > @ @
( >@ ) @
ch
> @ @ > > @
: > @> @ >> @ @
∑ w i fi
Ar
Z = O 5, i = ∑ w i f i = i (13)
∑ wi
, for i = 1,2.
i
www.SID.ir
89 / 92 / @> > @ 12
(ANFIS) ١ @ –
@ < <ANFIS
: > . (r2 q2 <p2 <r1 <q1 <p1)
w1 w2
Z= f1 + f2
w1 + w 2 w1 + w 2
= w1f1 + w 2f 2 (14)
= ( w1x ) p1 + ( w1y)q1 + ( w1)r1 + ( w 2 x ) p2 + ( w 2 y)q 2 + ( w 2 )r2 .
D
> .> > >@ >
@ @ < @
>
.[20]
>
>
@ > <
>
@
@
>
SI @
(
> <
>.
)
> ANFIS
@
>
of
@ @ > ANFIS > ><
> .> >
:> > @ @
ive
Y=F(X1,X2,…,Xn)
@ @ < Xi @ >
Y
ch
> > @ < >
<" ": :
. > > @
>> > <" "< >
Ar
@ Y > >. >
. @
i@ Pi <Y
<(Yi=1) > >> > > > (xi) >
1- Adaptive network based fuzzy inference system or adaptive neuro-fuzzy inference system.
www.SID.ir
13 …@ @ >@ - @ >
> (Yi=0) > > > i@ 1-Pi
: (15) @ <>
E ( Yi / x i ) = 0 * P ( Yi = 0 / x i ) + 1 * P ( Yi = 1 / x i ) = P ( Yi = 1 / x i ) = Pi
(15)
Yi @ @ >>
> < < Pi i @
.> @ >
> <ANFIS > @> >
D
@ > >
.
> < @
SI >
@ @
> @ @
-1-4
of
. > >@ > >@
85 80 @ < >
. 272 > @ > >
@ > @ 235 > <
ive
> @ 37 > . > @
< 235 @ . > > @ >>
. > > 70 > 165
ch
< <@ @ @ @ >
<> @ > < <@ <
<> @ < > > < >@> < @>
Ar
@ < > @> >
> > @ < > .
< @ > > @> @ @ >
> @
< >@ . < >
< > @ > @ > @
www.SID.ir
89 / 92 / @> > @ 14
@> < > > >>
> @ . @
> @> @
.
> <
< @ > . >< @
. @ @ @ <
D
< > @ > <
> <
. > @
.
>
>
(
@
)>
>
@
@
SI :(x1)
of
> <> >
> .> > >
.> > > @ <
@ > :(x2)
ive
> > @ > @ > > <
. > > @
.@ @ @ <@ @ >
ch
@ < >
. > >
@ : (x3) >
Ar
. @ @ > <> @
> @ < >
. >
@> @ -2-4
:> > @ < >
www.SID.ir
15 …@ @ >@ - @ >
R1: If (x1 is Low and x2 is Low and x3 is Low, then ( Y1 = c1 + c1x1 + c1 x1 + c1 x 3 )
o 1 2 3
2
R2: If (x1 is Low and x2 is Low and x3 is Medium, then ( Y 2 = c 2 + c1 x1 + c 2 x 2 + c2 x 3 )
o 2 3
M
27
R27:If(x 1is High and x2 is High and x3 is High then ( Y 27 = c 27 + c1 x1 + c 27 x 2 + c 27 x 3 )
2 3
) x3 ( ) x 2 <( ) x1 @> @ < >
o
(Medium ) <(Low) > ( >
i@ <Yi . @ @ @ (High)
> >> ><
D
cij
. >
٣
> < ٢
> <
١
@ @ @
<
.
@
@
>
@
>
> @
>
SI
@
@
< 200 <1,1 <0,9 <0,01
>
< > <
< > @ > >
of
. < <@ @ >
<(8) @ > > <k < @ (1-2-3) >
< >
ive
k
@ <> > > >
@> > > > <>
> <> k@ > >@ <>
. > @ > <
ch
@ @ >k <[20] <>
: >
%10 <k < > 4 > -1
Ar
.( 1,1 > )
< 2 > -2
.( 0,9 > ) %10 <k
1- initial step size.
2- step size decrease rate.
3- step size increase rate.
www.SID.ir
89 / 92 / @> > @ 16
< >> 2 4 <%10 >
< @ > .
. > @ k@
< @ > < @ >
> @ > > RMSE
200 > RMSE
> > >
D
@ <200 > < > > >@ >>
.>
> @ > < @ < @ >
>
. (
. >
:>
SI
@ِ
@ ) 108 < ci <
>
@
of
27
∑ w l Yl
Y = l =1
ˆ
27
(16)
∑
ive
wl
l =1
:> > wl
wl = ∏ µ (x i ) = µ (x i ) * µ (x i ) * µ (x i ) (17)
ch
Fil Fil Fil Fil
Fil . > @ > Yl < >
>. > xi @ @ (@ ) @
>@ @ < @ @
Ar
@ > (17) @ > > > .
< <@>
. @ @ wl . >
@ > (@> @ ) @ @
> 135 > > 27 >< <
.> >
www.SID.ir
17 …@ @ >@ - @ >
< > MATLAB ANFIS @ @
.> > 27 > <1,6GHz > Pentium IV > @
@ - 3- 4
:> > @ @ @
<(Y=1) @ > ," @ >"(1
. > > @
D
@ > =" @ >" (2
. > > @ <(Y=0)
>
@ >
SI >
<
>
>
@ -4-4
>
of
> > [0<1] @ >@ .> @
@ @ > @ ><
.
ive
.> @ @ @
>. @ > > @ <[22]
@ > > @ @ <
.> @ >
ch
> < @>
> @ < > @ >
> >@ @ <(3) > >.
Ar
> > @ > @ > >
@ . >> <
@ @ > <
. @ > <
ANFIS > @ <> (3) >
. 0,32
www.SID.ir
89 / 92 / @> > @ 18
D
@ @ SI -2
of
ANFIS > -5-4
<0,32 <(C) @ > <> @ <(1) >
Y@ @ @ >
@ >> > @ > ><
ive
. @ <
@ > > > ANFIS @ > @ >
. 0,70 0,67
ch
ANFIS > -1
( ) Y=1 ( ) Y =o > >
Ar
138 23 116 P(Y)≤C
97 47 49 P(Y)>C
235 70 165
163 47 116 >
69,36 67,14 70,30 (%) >
30,64 32,86 29,70 (%) >
www.SID.ir
19 …@ @ >@ - @ >
@ < @ <(1) >
> )@ < > @ ( @ >@ > )
. 0,30 0,33 > <( @ >@
@ >> > -6-4
> @ >> @ > <
@ > > 272 > > 37 > .> > @ >>
D
> > >
> > @ < @ <
> > >> >.
>
@
@
ANFIS
<>
SI (@
.
@
) >
>
of
( Y =o ) >@ < (@ )
>@ < >
(2) > .> @ ( Y =1 )
.
ive
ANFIS @ > <>
. 0,69 0,625 >
ch
ANFIS -2
( ) Y =1 ( ) Y =o > >
23 3 20 P(Y)≤C
14 5 9
Ar
P(Y) >C
37 8 29
25 5 20 >
67,57 62,5 68,97 (%) >
32,43 37,5 31,03 (%) >
www.SID.ir
89 / 92 / @> > @ 20
@
@ .> @ < > @
> > > @ > @ > <> @
> > .> > >
@ @ . @
>. @ @ < @
@ < @ -
D
<> > < . @
@ @ >
> < > . @ <>
<@
. SI
> @
>
<>
@
@
@
of
<@ @ @ ” (1381) < . . . .>. -1
> <@ >: <“ >
.
ive
: <" @ > @ @ @ "(1383) < -2
> >> < > > >
.
ch
> > @ " (1381) < < > > < > -3
.< C < > < @ <"
@ @ (1382) < < < / < -4
Ar
. >> <
5- Wilson, R. and Sharda, R. (1997) “Business Failure Prediction Using
Neural Networks.”, Encyclopedia of Computer Science and
Technology, Vol.37, No.22, pp.193-204.
6- Generation Approach for Managing Credit Scoring Problems. (2001)”
In Fuzzy Sets in Management, Economics and Marketing.(ed.), pp.223-
228.
7- Syau, Y., Hsieh, H. and Lee, E. S., (2001) “Fuzzy Numbers in the
Credit Rating of Enterprise Financial Condition.” Review of
Qunatitative Finance and Accounting, Vol.17, pp. 351-360.
www.SID.ir
21 …@ @ >@ - @ >
8- Cheng, C. B. and Lee, E. S., (1999) “Applying Adaptive Network to
Fuzzy Regression Analysis.”, Computers and Mathematics with
Applications, Vol.38, pp.123-140.
9- Boussabaine, A. H. and Wanoous, M.” (2000) A Neurofuzzy Model for
Predicting Business Bankruptcy.” In Business Applications of Neural
Networks: The State-of-the-Art of Real-World Applications(ed.), pp.55-
69.
10- Castillo, O. and Melin, P., (2002) ”Hybrid Intelligent Systems for Time
Series Prediction Using Neural Networks, Fuzzy Logic and Fractal
Theory.” IEEE Transactions on Neural Networks, Vol.13, No. 16.
D
11- Malhotra, R and Malhotra, D. K., (2002) ”Differentiating between
Good Credits and Bad Credits Using Neural-fuzzy Systems.” European
Journal of Operational Research, Vol. 136, pp.190-211.
12-
13-
Vol.45, pp.717–731.
SI
Jiao,Y., Syau,Y. and Lee, E. S., (2007) ”Modelling credit rating by
fuzzy adaptive network”, Mathematical and Computer Modelling
Hoffmann, F. Baesens, C. and Mues, T., (2007) "Inferring descriptive
and approximate fuzzy rules for credit scoring using evolutionary
algorithms" in European Journal of Operational Research, Vol.177, pp.
of
540–555.
14- Jang, J.S.R, (1993) “ANFIS: Adaptive – network based fuzzy inference
systems.”, IEEE transactions on systems Man. and Cybernetics, Vol.23,
No.3, pp.665-685.
15- Douligeris, C., and Palazzo, S., (1999) “Fuzzy Expert Systems in ATM
ive
Networks”,Fusion of Neural Networka, Fuzzy Sets, and Genetic
Algorithms: Industrial Application (ed.).
16- Jang, J.S.R, (1992) “Neuro fuzzy modeling: Architectures, Analysis and
Applications”, Department of Electrical Engineering and Computer
Science, University of California, Berkeley.
ch
17- Jang, J.S.R, (1995) “Neuro fuzzy modeling and Control.”, Proceedings
of IEEE, Vol.83, No.3, pp.378-406.
18- Rumelhart, D.E, Hinton, G.E. and Williams, R.J. (1986) "Learning
internal representations by error propagation," D. Rumelhart and J.
McClelland, editors. Parallel Data Processing, Vol.1, pp. 318-362.
Ar
19- Hagan, M.T., Demuth, H.B. and Beale, M.H. “Neural Network
Design”, PWS Publishing Company, Boston, MA 1996.
20- Jang, J.S.R, (1993) ANFIS: Adaptive – network based fuzzy inference
systems, Department of Electrical Engineering and Computer Science,
University of California, Berkeley.
21- Korsholm, L. (2004) ”Analysis of Diagnostic Studies, Sensitivity and
specificity positive predicted values ROC curves tests based on logistic
regression”, Departmant of statistics and demography, University of
Southern Denmark.
www.SID.ir
Related docs
Other docs by ps94506
Selberg Trace Formulae and Equidistribution Theorems for Closed Geodesics and Laplace Eigenfunctions
Views: 44 | Downloads: 0
Static Headspace-Gas Chromatography Theory and Practice (B Kolb & L S Ettre)
Views: 54 | Downloads: 0
Kocherlakota, N - Statistical Approach To Reporting Uncertainty on Certified Values of Chemical Reference Materials for Trace Metal Analysis (2002)
Views: 79 | Downloads: 0
(COINS)(BMC - GREEK 03) Poole-Catalogue of the Greek Coins in the British Museum The Tauric Chersonese Sarmatia Dacia Moesia Trace 1877
Views: 21 | Downloads: 0
Guitar World 2001-08 ACDC, Alien Ant Farm, Zeppelin, Linkin Park, Static-X, Beatles, Weezer
Views: 48 | Downloads: 0
