Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
5
,
Octo
ber
201
9
, pp.
4130
~
41
37
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
9
i
5
.
pp4130
-
41
37
4130
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
The satu
ration
of popu
lati
on fitnes
s a
s
a sto
pp
ing criterion i
n
ge
n
etic algorith
m
Fo
o
F
ong
Yen
g
1
,
S
oo
Kum
Yoke
2
, Az
ri
na Suh
aimi
3
1,3
Facul
t
y
of
Co
m
pute
r
and
Mat
hemati
c
al Sci
en
c
es,
Unive
rsiti T
e
knologi
MA
RA
Cawa
ngan
Johor
,
Mal
a
y
s
ia
2
Aca
dem
y
of La
nguage
Stud
ie
s,
Univer
siti
Te
kno
logi
MA
RA Ca
wanga
n
Neg
eri
Sem
bil
an
,
Ma
lay
sia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ju
l
1
,
201
8
Re
vised
A
pr
2
4
, 2
01
9
Accepte
d
Ma
y
6
, 2
01
9
Gene
tic
Algor
it
h
m
is
an
a
lgori
th
m
imita
ti
ng
the
nat
ura
l
evol
ut
io
n
proc
ess
in
solving
opti
m
izati
on
proble
m
s.
All
fe
asibl
e
(c
a
ndida
t
e)
solut
io
ns
would
be
enc
oded
int
o
chr
om
oso
m
es
and
under
go
the
exec
uti
on
of
gen
et
i
c
oper
at
ors
in
evol
uti
on
.
Th
e
e
volut
ion
i
tse
lf
is
a
pro
ce
ss
se
arching
for
op
ti
m
um
soluti
on
.
The
se
arc
h
ing
w
ould
stop
wh
en
a
stopping
cr
it
er
ion
is
m
e
t.
The
n
,
th
e
f
it
t
es
t
chr
om
osom
e
of
la
st
gen
erati
o
n
is
dec
l
are
d
as
the
opt
imum
soluti
on.
How
eve
r,
th
is
opti
m
um
soluti
on
m
ight
be
a
local
opt
imum
or
a
g
lo
ba
l
opti
m
um
soluti
o
n.
Hen
ce,
an
app
ropria
t
e
stopp
in
g
criter
ion
is
important
such
tha
t
the
se
arc
h
is
not
end
ed
bef
or
e
a
g
loba
l
opti
m
um
soluti
on
is
fo
und.
In
th
is
pape
r,
saturat
io
n
of
popula
ti
on
f
it
ness
is
propose
d
as
a
stopp
ing
cri
t
eri
on
for
endi
ng
th
e
sea
r
ch.
Th
e
propos
ed
stopping
c
ri
te
ri
a
was
compare
d
wit
h
conve
nt
iona
l
sto
pping
criter
ion,
f
it
te
st
chromos
omes repe
ti
t
ion,
u
nder
var
ious
par
amete
rs
set
ting.
The
resul
ts
show
th
at
the
per
form
ance
o
f
proposed
stopping
cri
t
erion
is
superior
as
compar
ed
to
th
e
conv
en
ti
on
al
stopping
cri
t
eri
on.
Ke
yw
or
d
s
:
Ar
ti
fici
al
i
ntell
igence
Gen
et
ic
a
l
gorithm
Ma
chine
l
ea
rn
i
ng
Op
ti
m
iz
ation
Stoppin
g
c
rite
r
ion
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Foo Fo
ng
Ye
ng
,
Faculty
of Com
pu
te
r
an
d
Ma
them
a
ti
cal
Scie
nces
,
Un
i
ver
sit
i Te
knol
og
i M
ARA
Ca
wanga
n
J
ohor
,
K
am
pu
s Pa
sir Gu
dang,
Jal
an
P
urnam
a, Ban
dar
Seri
A
lam
, 8
1750,
M
asai
, J
oh
or, Ma
la
ysi
a
.
Em
a
il
:
fo
ofo
931@uit
m
.ed
u.
m
y;
fong
ye
ngf@gm
ail.co
m
1.
INTROD
U
CTION
In
19
75,
J
ohn
Ho
ll
an
d,
ins
pir
ed
by
t
he
Darwin
Ev
olu
ti
on
The
or
em
[1,
2]
,
intr
oduce
d
a
n
al
gorithm
[3,
4]
m
i
m
ick
in
g
the
proc
ess
of
ge
netic
inh
erit
a
nce
in
ev
olu
ti
on
[4
-
6]
.
It
wa
s
nam
ed
as
Gen
et
ic
Algorithm
(GA)
.
Du
e
to
it
s
adap
ta
ble
c
ompete
nces
,
t
he
a
lgorit
hm
app
li
cat
ion
s
in
rese
arch
area
s
a
re
gig
a
nti
c
[7]
w
her
e
they
are
no
t
only
found
in
pure
sci
ences
su
c
h
as
e
ng
i
neer
i
ng
[8
-
10]
but
al
s
o
in
so
ci
al
sci
ence
s
su
c
h
as operati
on m
anag
em
ents
[11
-
14]
.
GA
h
as b
ee
n
c
at
egorised under
the f
am
ily
of
m
et
a
-
heu
risti
c
al
gorithm
s
[15,
16]
su
c
h
as Tab
u
Sea
rc
h
and A
rtific
ia
l Neural
N
et
w
or
k.
Me
ta
-
heurist
ic
al
gorithm
s
a
re
al
ways u
se
d
f
or
s
ol
ving
c
om
bin
at
or
ia
l
pr
ob
le
m
s
or
hard
opti
m
iz
at
ion
pro
blem
s
[17]
si
nce
they
can
pro
vid
e
go
od
s
ol
utions
at
reas
on
a
ble
com
puta
ti
on
al
cost
[18
,
19
]
.
Ho
we
ver,
t
hey
m
ay
no
t
be
a
ble
to
gu
a
ra
ntee
the
opti
m
alit
y
of
s
olu
ti
on
due
t
o
t
heir
sto
chasti
c
natu
re
[
20]
.
Sli
gh
tl
y
dif
fere
nt
from
so
m
e
m
et
a
-
heurist
ic
al
gorithm
s
that
i
m
pr
ov
e
a
s
ing
le
s
olu
ti
on,
the
ge
netic
al
gorithm
handles
a
gro
up
of
fe
asi
ble
s
olut
ion
s
sim
ultaneousl
y
[
21
]
.
T
hese
feasible
s
olu
ti
ons
a
re
e
ncode
d
into
c
hrom
os
om
es
[22]
an
d
placed
i
nto
an
en
vir
on
m
ent
analo
gu
e
of
na
tural
e
vo
l
utio
n
wh
e
re
t
hey
need
to
su
r
viv
e
,
a
da
pt,
an
d
pro
pa
ga
te
their
ge
ne
ti
cs
to
the
f
ut
ur
e
ge
ner
at
io
ns
[21]
.
T
he
ev
olu
ti
on
of
these
chrom
os
om
es
is
the
p
ro
ce
ss
of
se
arc
hing
t
he
op
ti
m
u
m
s
olu
ti
on.
T
he
e
vo
l
ution
ta
kes
m
any
gen
erat
ion
s
t
o
conve
rg
e
t
o
pe
rf
ect
ly
ada
pted
ch
ro
m
os
om
e
(g
lo
bal
op
ti
m
um
so
luti
on
)
[
21
]
.
Hence
,
ti
m
ing
f
or
e
nd
i
ng
t
he
evo
l
ution
/s
ea
rc
hing
pr
ocess
i
s
cru
ci
al
.
Othe
rw
ise
,
prem
atu
re
c
onve
rgen
ce
cou
l
d
happ
en
in
t
he
ev
ol
ution.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Th
e s
at
ur
atio
n of po
pu
l
ation f
it
ness
as
a
sto
ppin
g
cri
te
rio
n
i
n gen
et
ic
algor
it
hm
(
Fo
o Fo
ng Ye
ng)
4131
A
prem
at
ur
e
c
onve
rg
e
nce
shou
l
d
be
av
oide
d
a
s
it
m
ay
le
d
t
o
t
he
a
cq
ui
sit
ion
of
a
l
oc
al
opti
m
u
m
s
olu
ti
on
instea
d
of
glob
al
op
ti
m
u
m
sol
ution
[
20
]
.
Th
e
ch
oice
of
the
sto
pp
i
ng
crit
e
rio
n
would
det
erm
ine
if
the
gl
ob
al
op
ti
m
u
m
so
luti
o
n co
uld be
fo
und be
fore its s
earchi
ng pro
c
e
ss is term
inate
d.
In
this
pap
e
r,
we
i
nvest
igate
d
t
he
sto
ppin
g
crit
erion
of
ge
netic
al
gorit
hm
as
the
ti
m
e
of
searchi
ng
t
he
so
luti
on
sp
ace
is
on
e
of
t
he
ver
y
im
po
rta
nt
facto
rs
to
fin
d
t
he
global
optim
u
m
so
luti
on.
W
e
pro
pos
ed
t
he
densi
ty
or
sat
urat
ion
of
popul
at
ion
fitne
ss
as
the
ne
w
sto
pp
ing
c
rite
rio
n
w
hich
se
rv
e
d
as
a
m
easur
em
ent
key
to
en
d
the
sea
rch
i
ng
proces
s
.
The
pro
pose
d
sto
ppin
g
crit
erio
n
wa
s
co
m
par
ed
to
a
c
onve
ntion
al
st
opping
crit
erion
wh
ic
h
the
sea
rch
i
ng
p
r
ocess
will
be
st
oppe
d
when
the
re
is
no
i
m
pr
ovem
ent
of
fitt
est
ch
rom
os
o
m
e
for
s
om
e
su
cce
ssive
ge
ner
at
io
ns
.
T
he
num
erical
resu
lt
s
sho
w
t
hat
the
pro
po
s
ed
sto
ppin
g
crit
erio
n
has
bette
r
perform
ance as co
m
par
ed
to
t
he
c
onve
ntio
na
l st
opping c
rite
rio
n.
This
pa
pe
r
is
orga
nized
as
f
ollow
.
Sect
io
n
2
giv
es
the
desc
r
ipti
on
of
t
he
ge
netic
al
gorith
m
includin
g
so
luti
on e
nc
od
i
ng / enc
ryptin
g, ev
olu
ti
on a
nd
the stop
ping c
rite
ria that h
as
been us
ed
. S
ec
ti
on
3 detai
ls h
ybrid
al
gorithm
s
fo
r
two
te
ste
d
m
od
el
s,
one
with
conve
nt
ion
al
s
toppin
g
c
rite
rion,
on
e
wit
h
t
he
pro
po
se
d
stoppin
g
crit
erion.
Sect
ion
4
re
ports
on
e
xperim
e
ntal
resu
lt
s
with
di
ff
e
ren
t
par
am
et
ers/
gen
et
ic
dr
i
ft
set
ti
ng.
In sect
ion 5
, the
f
in
dings a
re c
on
cl
ud
e
d
a
nd s
om
e reco
m
m
e
nd
at
io
ns are
gi
ven f
or futu
re st
ud
y.
2.
THE
V
IT
AE
OF GE
NETI
C ALGO
RIT
HM
In
an
arti
fici
al
evo
l
utio
n
syst
em
as
sh
own
i
n
Figure
1
,
ge
net
ic
al
gorithm
se
arch
be
gin
s
by
gen
e
rati
ng
a
popula
ti
on
of
ra
ndom
ly
gen
e
rated
can
did
at
e
s
olu
ti
ons.
Thes
e
ca
ndidate
s
olu
ti
ons
a
re
e
ncode
d
int
o
chrom
os
om
es
[7]
an
d
br
ought
into
the
e
voluti
on
that
is
con
st
ru
ct
e
d
by
gen
et
ic
s
op
erators.
Eac
h
of
the
chrom
os
om
es i
s assig
ne
d wit
h a fit
nes
s fu
nction
[
20]
that ser
ves
as
a
fitness
ind
e
x.
Fig
ure
1.
A
rtific
ia
l evo
l
ution s
yst
e
m
in
gen
et
i
c algori
thm
The
c
hrom
os
om
es
from
the
s
a
m
e
gen
e
rati
on
w
ou
l
d
hav
e
t
o
c
om
pete
with
one
a
no
t
her.
A
sel
ect
ion
pr
ess
ure
t
hat
i
s
biased
to
pr
i
vilege
on
the
f
it
te
st
chrom
os
om
es
is
en
forc
ed
i
n
the
syst
em
[5]
.
T
his
ac
t
is
t
o
ens
ur
e
that
onl
y
those
with
dom
inant
trai
ts
of
o
ptim
isa
t
ion
w
ould
hav
e
hig
he
r
possi
bili
ties
to
be
sel
ect
ed
f
or
passing
the
ge
nes
[
21
]
.
T
hro
ugh
the
exec
ut
ion
of
ge
n
et
ic
op
e
rato
rs
on
these
sel
ect
ed
chrom
os
om
es,
a
ne
w
popula
ti
on
/ ge
ner
at
io
n
of
ch
r
om
os
om
es
is
fo
rm
ed
[
20
]
. The
ev
olu
ti
on/
sea
rc
hi
ng
proce
ss w
il
l
con
ti
nue unti
l
a
stoppin
g
c
rite
r
ion
/
t
hr
es
hold
is
m
e
t
and
t
he
fitt
est
in
the
la
st
gen
e
rati
on
will
be
ide
nt
ifie
d
as
a
n
op
t
i
m
u
m
so
luti
on
[23]
. Ho
we
ver, the
iden
ti
fie
d op
ti
m
um
so
luti
on could
be
a
local
optim
u
m
so
luti
on instea
d of a
gl
ob
al
op
ti
m
u
m
so
luti
on if t
he
sea
rchi
ng
proces
s e
nded
to
o
ea
rly
.
Ba
sic
al
ly
,
the
al
gorithm
sea
rch
in
volves
tw
o
diff
e
re
nt
s
pa
ces,
on
e
is
c
oding
sp
ace
,
a
nd
the
oth
e
r
i
s
so
luti
on
s
pace
[24]
.
T
he
im
plem
entat
ion
of
ge
netic
ope
rators
on
c
ode
d
s
olu
ti
ons,
nam
el
y
chr
om
os
om
es,
works
in
the
c
od
i
ng
s
pace
.
Nonetheless
,
the
eval
uation
and
sel
ect
io
n
of
c
hrom
os
ome
s
are
em
plo
ye
d
in
th
e
so
luti
on s
pace
wh
ic
h
is t
he
s
pa
ce f
or
a
n
a
ct
ua
l solutio
n
[24]
.
2.1.
So
luti
on e
ncr
yp
ti
on
Each
ge
ner
at
i
on
is
c
on
sti
tut
ed
by
a
popul
at
ion
of
siz
e
M
can
di
date
s
olu
ti
ons.
Th
es
e
can
did
at
e
so
luti
ons
a
re
f
easi
ble
so
l
utio
ns
[
25]
w
hich
will
be
enc
rypt
ed
int
o
ch
ro
m
os
om
es
C
ij
const
ru
ct
ed
by
ge
ne
g
ijk
.
They
c
ou
l
d
be
in
the
f
or
m
of
bin
a
ry
strin
g
[
17
]
,
real
num
ber
stri
ng
[
8]
or
m
at
rix
[26]
de
pendi
ng
on
t
he
t
ypes
of
op
ti
m
iz
a
ti
on
pro
blem
.
Th
e
(
1
)
is
the
so
l
ution
represe
ntati
on
f
or
m
for
j
th
c
hrom
os
om
e
of
k
ge
nes
in
i
th
gen
e
rati
on. Fo
r
the
pur
po
se
of
this r
e
searc
h,
bin
a
ry strin
g
c
on
st
ru
ct
e
d by
gen
e
g
ijk
={
0,1
} w
as
ch
os
e
n.
C
ij
= (
g
ijk
,… g
ij3,
g
ij2,
g
ij1,
g
ij0
)
(1)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
Int
J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
201
9
:
4130
-
4137
4132
2.2.
Sele
ction
,
m
ati
ng
an
d
mu
tat
ion phase
s in
t
he evolu
tion
In
the
arti
fici
al
ev
olu
ti
on
syst
e
m
,
al
l
chrom
os
om
es
of
each
gen
e
rati
on
m
us
t
go
t
hroug
h
t
he
e
voluti
on
[27]
that
c
om
pr
ise
d
of
a
sel
ec
ti
on
,
m
at
ing
a
nd
m
utati
on
pha
ses
as
s
how
n
i
n
Fi
gure
2
.
I
n
the
sel
ect
ion
phase,
the
sel
ect
ion
pressu
re
is
e
nfo
rced
i
n
a
pro
ba
bili
sti
c
m
ann
er
into
eac
h
generati
on
by
us
in
g
sel
ect
ion
ope
rators
su
c
h
as
R
oule
tt
e
W
heel
[
12
]
,
To
urnam
ent
[
28
]
,
an
d
Eli
ti
sm
[9]
.
In
this
pap
e
r,
Ro
ulett
e
Wh
eel
an
d
Eli
ti
s
m
Sele
ct
ion
Oper
at
or
s
we
re s
el
e
ct
ed
to
be
im
po
se
d
in
the st
udy.
Fig
ure
2
.
Ev
ol
ution
As
the
nam
e
of
Ro
ulett
e
Wheel
su
ggest
s,
e
ach
c
hrom
os
om
e
was
gi
ven
a
slot
that
is
pr
oport
ion
al
to
it
s
fitness
i
n
a
n
im
aginar
y
r
oule
tt
e
w
heel.
T
he
siz
e
of
eac
h
slot
determ
ines
the
pro
ba
bili
t
y
of
a
c
hrom
os
om
e
bein
g
sel
ect
ed
for
t
he
ne
xt
ph
ase.
T
his
m
eans
that
the
fitt
es
t
w
ou
l
d
ha
ve
pr
ece
de
nce
i
n
br
ee
din
g
tha
n
tho
se
that
we
re
not
well
-
ada
pted
[
17
]
.
T
he
sel
ec
ti
on
pressu
re
of
Ro
ulett
e
Wh
eel
op
e
rato
r
was
e
nfo
rce
d
i
nto
th
e
chrom
os
om
es p
ool
with s
ome
proba
bili
ty
o
f
cr
os
s
over
(
c
r
os
s
ov
e
r rat
e, C
R).
Tw
o
c
hrom
os
om
es
that
had
bee
n
sel
ec
te
d
by
Ro
ule
tt
e
Wh
eel
we
re
treat
ed
as
the
pa
rent
chrom
os
om
es
(P
C
1,
PC
2
)
f
or
br
ee
ding
in
t
he
m
ating
phas
e.
I
n
this
phas
e,
the
cr
os
s
ove
r
op
e
rato
r
e
xc
hange
d
and
sp
li
ced
t
he
se
gm
entat
i
on
s
of
bo
t
h
par
e
nt
ch
r
omoso
m
es
at
ran
dom
po
int
(
r)
[2
5]
to
f
or
m
new
chrom
os
om
es
cal
le
d
offs
pr
i
ngs
(
O
1
,
O
2
)
(
2
)
.
Thes
e
offs
pri
ng
s
will
car
ry
the
exc
ha
ng
e
d
genet
ic
inf
orm
at
ion
wh
ic
h
i
nh
e
rite
d
from
their
par
e
nt
c
hrom
os
om
es
[7]
.
I
n
this
st
ud
y,
t
he
operati
on
was
re
peated
un
ti
l
a
popula
ti
on
of
M
-
2 offs
pr
i
ng
was fo
rm
ed.
PC
1
= ( g
ijk
,…
g
ijr+1
, g
ijr
,…
g
ij
2,
g
ij1,
g
ij0
)
PC
2
= ( g’
ijk
,…
g’
ijr+1
, g’
ijr,
,…
g’
ij2,
g’
ij1,
g’
ij0
)
(2)
O
1
= (
g
ijk
,… g
i
jr+1
,
g’
ijr
,… g
’
ij2,
g’
ij1,
g’
ij0
)
O
2
= (
g’
ijk
,…
g’
ijr+1
, g
ijr,
,…
g
ij2,
g
ij1,
g
ij0
)
Nex
t,
t
he
offs
pri
ngs
we
re
bro
ught
into
t
he
m
uta
ti
on
phas
e
that
con
sist
e
d
of
m
utati
on
op
e
rato
r
wit
h
so
m
e
pr
oba
bili
ty
of
m
utati
on
(m
utati
on
rate,
MR
).
T
he
pro
bab
il
it
y
of
m
utati
on
is
al
ways
a
lowe
r
pro
ba
bili
t
y
than
the
pro
ba
bili
ty
of
cro
s
s
ov
e
r
[
7]
.
T
he
m
uta
ti
on
oper
at
or
m
igh
t
al
t
er
the
ge
ne
of
the
chrom
os
om
es
at
rand
om
po
sit
ion
(R
)
[
9]
(
3
)
with
the
i
ntent
ion
of
va
ryi
ng
the
ge
netic
[
22]
an
d
hen
ce
f
ur
t
her
e
xpan
di
ng
the
so
luti
on
s
pace
search
[20]
.
The
off
sprin
gs
would
the
n
be
reg
a
rd
e
d
as
new
c
hrom
os
om
es
fo
r
the
c
om
ing
gen
e
rati
on.
Th
e m
utati
on
ope
rators m
igh
t serv
e
as a t
oo
l t
o red
uce t
he
ris
k of p
rem
at
ur
e conve
rg
e
nce.
O= (g
ijk
,
…
g
ijR,
g
ij2,
g
ij1,
g
ij0
)
g
ijR
= 1
if g
ijR
=0
(3)
0
if
g
ijR
=1
Si
m
ultaneo
us
ly
,
Eli
ti
s
m
op
erat
or
was
us
ed
t
o
m
ake
sure
tha
t
the
el
it
e
chrom
os
om
es
(f
it
te
st
chrom
os
om
es
of
each
gen
e
rati
on)
would
not
be
disrupte
d
by
the
exec
utio
n
of
the
c
rosso
ve
r
an
d
m
uta
ti
on
op
e
r
at
or
s
[
28
]
.
T
wo
el
it
es
w
ou
ld
be
rep
li
cat
ed
directl
y
as
ne
w
ch
ro
m
osom
es
fo
r
the
c
om
ing
gen
e
rati
on.
T
hi
s
was
a
pr
es
erv
at
io
n
act
f
or
reduci
ng
t
he
pro
bab
il
it
y
of
l
os
in
g
the
fitt
est
gen
e
from
the
chrom
os
om
e p
oo
l a
nd
decr
ea
sing t
he
ti
m
e o
f
c
onverge
nce
to an o
ptim
u
m
so
luti
on.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Th
e s
at
ur
atio
n of po
pu
l
ation f
it
ness
as
a
sto
ppin
g
cri
te
rio
n
i
n gen
et
ic
algor
it
hm
(
Fo
o Fo
ng Ye
ng)
4133
2.3.
St
oppi
n
g
c
ri
te
ri
a
Last
ly
,
a
j
ud
gem
ent
to
stop
the
ev
olu
ti
on
is
i
m
p
or
ta
nt.
The
decisi
on
of
ha
ving
a
n
appr
opriat
e
thres
ho
l
d/sto
ppin
g
c
rite
rio
n
gr
eat
ly
af
fects
the
ca
pa
bili
ty
of
t
he
al
go
rithm
[26]
.
Ge
ne
rall
y,
there
a
re
tw
o
stoppin
g
c
rite
ria that ha
ve
b
ee
n widely
used:
a)
Ma
xim
u
m
g
en
erati
on
s
or m
axim
u
m
CPU
ti
m
e all
ow
ed
[
26]
.
b)
N
o
im
pr
ovem
ent of
fitt
est
chrom
os
om
e fo
r
su
ccessi
ve gen
erati
on
s
(
fi
tt
est
ch
r
omos
ome r
epeti
ti
on
)
[
14]
.
The
first
crit
er
ion
(a
)
w
ou
l
d
end
the
e
volut
ion
if
the
al
go
rithm
had
m
et
the
pr
e
-
de
fine
d
m
axim
u
m
nu
m
ber
of
ge
ne
rati
on
s
or
CP
U
ti
m
e.
Howe
ver,
to
determ
i
ne
t
he
a
ppr
opr
ia
te
tim
e
or
m
axim
u
m
nu
m
ber
of
gen
e
rati
ons
is
a
pe
rp
le
xing
puzzl
e.
Def
i
ning
a
ve
ry
huge
num
ber
of
m
axim
u
m
gen
er
at
ion
s
or
CP
U
tim
e
would
le
ad
to
unpract
ic
al
c
om
pu
ta
ti
on
al
t
i
m
e
wh
il
e
a
s
m
al
l
nu
m
ber
m
igh
t
cause
the
al
go
rithm
has
not
enou
gh
ti
m
e
to
reach
a
global
op
ti
m
u
m
so
luti
on
.
T
he
siz
e
of
t
he
po
pu
la
ti
on
w
ou
l
d
so
m
et
i
m
es
influ
en
ce
t
he
durati
on r
e
quir
ed fo
r
c
onve
rgence
[28]
.
In
this
as
pect,
the
sec
ond
c
rite
rio
n
(
b)
m
igh
t
seem
a
bette
r
ch
oice
since
t
he
al
go
rithm
woul
d
sto
p
i
f
there
is
no
i
m
pr
ov
em
ent
of
fitt
est
chro
m
os
o
m
e
fo
r
f
ew
ge
ner
at
io
ns
s
uccessively
(
fi
tt
est
chr
omos
ome
repeti
ti
on
)
.
A
ga
in,
it
is
ano
t
he
r
pa
rado
x
w
he
re
the
al
gorit
hm
us
er
nee
ds
t
o
dete
rm
ine
the
appr
opriat
e
num
ber
of
su
cce
ssive
r
epeti
ti
on
for
di
sm
issi
n
g
the
s
earch
.
T
he
s
uc
cessi
ve
rep
et
it
ion
of
fitt
est
ch
ro
m
os
om
e
is
great
l
y
influ
e
nce
d by the size
a
nd c
om
plexit
y of
t
he
r
esea
rch p
robl
e
m
[26]
.
In
this
pa
pe
r,
a
st
opping
cr
it
erion
that
m
easur
e
s
t
he
sa
turati
on
of
popu
la
ti
on
fitnes
s
(
F)
of
M
chrom
os
om
es
was
propose
d.
The
pro
po
se
d
stoppin
g
c
rite
rion
ai
m
ed
to
w
ork
as
a
t
hr
es
hold
t
o
sto
p
sea
rchi
ng
wh
e
n
t
he
fitne
ss
de
viati
o
n
of
the
popula
ti
on
is
sm
all
(
4
)
(
wh
e
n
δ→
0)
an
d
furthe
r
e
nh
a
nce
the
com
petency
of
al
gorithm
in
find
i
ng
gl
ob
al
optim
u
m
so
luti
on
.
[(1/M
)
∑
(
F
ij
–
F)
2
]
< δ ,
δ
→
0
(4)
3.
E
X
PERI
MEN
TAL M
ODEL
A
LG
ORI
T
H
M
To
te
st
the
pro
po
s
ed
sto
ppin
g
crit
eri
on,
t
wo
m
od
el
s
with
di
ff
e
ren
t
sto
ppin
g
c
rite
ria
we
re
dev
el
op
e
d.
The
fi
rst
m
od
el
,
nam
ed
as
Norm
al
s
top
pi
ng
crit
erion
(
Nsc)
m
od
el
,
was
de
sign
e
d
with
a
n
al
go
rit
hm
that
ends
the
sea
rch
i
ng
process
w
he
n
fi
tt
est
chro
m
osome
re
petit
ion
s
ha
ve
reac
he
d
t
he
plate
au,
pr
e
fix
up
per
boun
d.
The
sec
ond
m
od
el
,
nam
ed
a
s
the
Sat
ur
at
io
n
sto
ppin
g
c
rite
rio
n
(
Ssc)
m
od
el
,
was
c
reated
with
the
pro
po
s
e
d
stoppin
g
c
rite
r
ion
.
The
searc
hing
was
sto
pped
w
he
n
the
gen
e
rati
on
sat
ur
at
e
d
wit
h
th
e
fitt
est
chrom
os
om
es.
The
al
gorit
hm
s
of the
tw
o
m
od
el
s ar
e
il
lustra
te
d
in
Fig
ur
e
3.
(a)
(b)
Fig
ure
3
.
Al
gorithm
s
fo
r (a)
Nsc
M
od
el
;
(b)
Ssc
M
od
el
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
Int
J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
201
9
:
4130
-
4137
4134
These
tw
o
m
od
el
s
a
re
hybri
d
with
a
f
or
ecas
ti
ng
e
quat
ion.
This
forecast
in
g
e
quat
ion
c
onsist
s
of
on
e
adjuste
d
pa
ra
m
et
er
α
that
r
ang
e
betwee
n
zero
a
nd
1.
T
he
ch
oice
of
t
he
pa
ram
et
er
α
w
ould
in
flu
ence
the
accuracy
of
f
oreca
sti
ng
sim
ulati
on
.
He
nce,
the
two
m
odel
s
wer
e
assig
ned
t
o
ide
ntify
a
glo
bal
opt
i
m
u
m
so
luti
on
α
s
uc
h
th
at
the
accu
rate
f
or
eca
sti
ng
sim
ulati
on
c
ou
l
d
be
pro
du
c
ed
.
N
um
erical
te
sts
wer
e
use
d
to
te
st
the ef
fici
ency
of the t
wo m
od
el
s.
The
par
am
et
er
α
would
be
encode
d
int
o
chrom
os
om
e
i
n
the
f
or
m
of
(
1
)
a
nd
the
c
hrom
os
om
e
decip
her
of
α
is
as
(
5
)
.
Eac
h
e
nc
oded
pa
ram
et
er
α
ij
wa
s
as
sign
e
d
with
a
fitnes
s
fun
ct
io
n
(F
ij
(
α
ij
)),
Me
a
n
A
bsolute
E
r
ror
(
6
)
.
T
he
F
ij
(
α
ij
)
e
xpresse
d
the
fitness
f
un
ct
io
n
of
j
th
chrom
os
om
es
in
i
th
gen
e
rati
on.
T
he
ch
ro
m
os
om
e
that
is
able
to
m
ini
m
ize
the
errors
of
forecast
w
as
viewe
d
as
fitt
est
.
The
f
oreca
sti
ng
eq
uatio
n
was
m
od
ifie
d
a
nd
hyb
rid
int
o
al
g
or
it
hm
s
sta
te
d
as
(
7
)
.
T
he
h
ijt
was
t
he
foreca
st
da
t
a
si
m
ulate
d
by
(
7
)
an
d
y
t
repres
ented
t
he real
dat
a.
α
ij
=
(∑
g
ijk
×
2
K
)/
100 ,
K=0
,1,2…k
(5)
Mi
ni
m
iz
e F
ij
(
α
ij
)
=
(
1/k)∑|
h
ij
t
-
y
t
|
sub
j
ec
t t
o
0<
α
ij
<
1
(6)
h
ijt
=
α
ij
(
f
ij1t
–
f
ij2t
)/ (
1
–
α
ij
)+
[
2
f
ij1t
–
f
ij2t
]
f
ij1t
=
α
ij
y
t
+
(1
–
α
ij
)
f
ij1t
–
1
(7)
f
ij2t
=
α
ij
f
ij1t
+
(1
–
α
ij
)
f
ij2t
–
1
Gen
e
ra
ti
on,
i
=1,
2…
up
per bou
nd o
f
ge
ner
ation
Chrom
os
ome,
j
=1, 2
...
M
c
hrom
osome
Da
t
a
Ti
me,
t
=
1,
2 ...
m
axim
um
num
ber of ti
me
In
t
he
sel
ect
io
n
phase
,
the
se
le
ct
ion
press
ure
was
a
ff
ect
e
d
by
sel
ect
ion
prob
a
bili
ty
(RP
ij
(C
ij
))
(
8
)
of
each
c
hrom
os
om
e.
At
t
he
sam
e
tim
e,
two
el
it
e
ch
r
om
os
om
e
s
(EC
w
)
woul
d
be
re
plica
te
d
di
rec
tl
y
into
t
he
nex
t
gen
e
rati
on
(
9
)
.
Two
el
it
es
(E
C
w
)
of
g
ene
rati
on
i
th w
ere
a
ss
ign
e
d
as
the f
ir
st
two
ne
w
c
hrom
os
om
es
(C
i+
1,w
)
of
the
c
om
ing
ge
ner
at
io
n
i
+1
t
h.
I
n
the
cr
ossove
r
an
d
m
utati
on
phases,
the
possibil
it
ie
s
of
ha
pp
e
ning
we
re
con
t
ro
ll
ed
b
y c
ro
ss
over
r
at
e
(
CR
)
an
d
m
utati
on
rate (MR
).
P
ij
(C
ij
)=F
ij
/ ∑
F
ij
CP
ij
(C
ij
)=
1
-
P(C
ij
)
(8)
RP
ij
(C
ij
)= CP
ij
/
∑ CP
ij
If
EC
w
= C
i,j
,
the
n
C
i+1,w
=
EC
w
, w
=
1,2 ,
j
=
tw
o
ide
ntif
ie
d
el
it
e chr
om
os
omes
(9)
4.
RESU
LT
S
A
ND
DI
SCUS
S
ION
Both Nsc an
d Ssc
m
od
el
s w
e
re teste
d
f
or
t
he
ir capa
bili
ti
es
in f
oreca
sti
ng s
i
m
ulati
on
b
y u
sing
a set of
tim
e
series
data
range
[13
00,
1600]
.
T
he
c
om
petence
of
m
od
el
s
in
av
oi
din
g
prem
at
ur
e
co
nv
e
rg
e
nce
and
fin
ding
the
global
optim
um
so
luti
on
wa
s
use
d
as
a
n
in
dex
for
ga
ugin
g
a
good
m
od
el
.
A
good
m
od
el
sh
al
l
be
able
to
pro
du
ce
a
sim
ulated
r
esult
that
i
s
ve
ry
cl
os
e
t
o
act
ual
val
ue
[28]
under
va
rio
us
c
irc
um
s
ta
nces
(of
diff
e
re
nt cross
ov
e
r rat
e an
d
m
utati
on
r
at
e
).
In
this
researc
h,
bo
t
h
m
od
el
s
carried
out
sim
ula
ti
on
ex
perim
ent
with
1000
tria
ls
to
acc
um
ulate
the
sta
ti
sti
cal
reco
r
ds
of
sto
ppin
g
crit
eria
eff
ic
ie
ncy.
The
e
ff
ect
iveness
of
st
op
ping
crit
erio
n
cou
l
d
be
rev
ea
le
d
by
the
pro
ba
bili
ty
of
ha
ving
a
good
forecast
(c
on
sist
of
globa
l
op
ti
m
u
m
so
luti
on
α
)
w
he
n
the
searc
hing
end
e
d.
The ge
netic
drift/
par
am
et
ers
of
bo
t
h
e
xp
e
rim
ent m
od
el
s ar
e
sta
nd
a
rd
iz
e
d
as
in
Ta
ble
1.
Table
1.
Ge
netic
d
rift
/
pa
ram
e
te
rs
f
or ex
pe
rim
ent
m
od
el
s
Gen
etic drif
t/p
ara
m
e
te
rs
Po
p
u
latio
n
size,
M
=
2
0
chro
m
o
so
m
e
s
Cro
ss
o
v
er
rate,
CR
=
[
0
.5, 0.9
]
with
r
ate interval 0
.1
Mutatio
n
r
ate,
M
R
=
[
0
.01
,0.1
]
with
r
ate interval 0
.01
Fig
ure
4
de
picts
the
pro
ba
bili
t
ie
s
of
Nsc
m
odel
of
ac
hieving
gl
ob
al
optim
um
al
ph
a
α
unde
r
dif
f
e
r
e
nt
cro
ss
over
rates
(CR)
an
d
m
ut
at
ion
rates
(C
R).
The
perf
orm
ances
of
Nsc
m
od
el
wer
e
a
bove
a
ver
a
ge
wh
e
n
cro
ss
over
r
at
e w
as
set
in
t
he
r
ang
e
o
f
[
0.5
, 0
.
7],
t
he
pro
ba
bili
ti
es
of
getti
ng
go
od f
oreca
st wer
e
re
porte
d
abov
e
0.6
but
fa
r
fro
m
1.
U
nd
e
r
t
he
se
CR
s,
t
he
a
bi
li
ti
es
of
t
he
m
od
el
wer
e
im
pr
oved
wh
e
n
t
he
MR
was
raise
d
to
t
he
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Th
e s
at
ur
atio
n of po
pu
l
ation f
it
ness
as
a
sto
ppin
g
cri
te
rio
n
i
n gen
et
ic
algor
it
hm
(
Fo
o Fo
ng Ye
ng)
4135
range
[
0.08,
0.1],
ye
t
the
pro
bab
il
it
ie
s
we
re
sti
ll
sli
gh
tl
y
f
urt
her
f
r
om
1.
T
he
e
ff
ect
ive
nes
s
of
Nsc
in
obt
ai
nin
g
accurate
f
or
eca
sti
ng
res
ults
w
as
re
ported
go
od
w
he
n
CR
w
as
set
hi
gh
(
CR
=[0
.
8,0.9]
),
the
pro
bab
il
it
ie
s
we
r
e
aver
a
gely
abo
ve
0
.
7 re
gardles
s of the
MR
.
Fig
ure
4
.
The
pro
bab
il
it
ie
s o
f
N
sc m
od
el
of
ob
ta
ini
ng g
l
ob
al
o
ptim
u
m
α
unde
r
CR
=[0
.
5,0.9]
with inte
rv
al
ra
te
0
.
1
a
nd MR
=[0.0
1,0.1]
w
it
h
inte
rv
al
rate
0.01
The
ra
da
r
cha
r
ts
of
Fig
ure
5
rev
eal
the
ca
pa
bili
ty
of
the
Ssc
m
od
el
under
the
sam
e
gen
et
ic
dri
ft.
The
Ssc
m
od
el
see
m
ed
able
to
pro
vide
bett
er
res
ults
than
the
Nsc
m
od
el
wh
e
n
CR
an
d
MR
rates
were
no
t
high
(CR=
[
0.5
,0
.
7],
MR
=[
0.0
1,0.06])
.
U
nde
r
the
sam
e
CR
i
nterv
al
[
0.5,
0.7],
the
com
pe
te
ncy
of
t
he
m
od
el
becam
e
app
are
nt
w
he
n
MR
w
as
set
ab
ove
0.07.
T
he
perfor
m
ances
of
the
Ssc
m
od
el
we
r
e
superi
or
w
he
n
bo
t
h
CR
an
d
MR
wer
e
high
(C
R=
[0
.
8,
0.9]
a
nd
MR
=[
0.0
7,
0.1])
,
t
he
pro
bab
il
it
ie
s
of
ge
tt
ing
global
optim
u
m
so
luti
on
wer
e
a
ppr
oach
i
ng 1.
Fig
ure
5
.
The
pro
bab
il
it
ie
s o
f
Ssc
m
od
el
of
ob
ta
ini
ng g
l
ob
al
o
ptim
u
m
α
unde
r
CR
=[0
.
5,0.9]
with inte
rv
al
ra
te
0
.
1and
MR
=
[0.01,
0.1] wit
h i
nter
val r
at
e
0.01
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
Int
J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
201
9
:
4130
-
4137
4136
Figure
6
s
how
the
com
par
iso
n
of
S
sc
a
nd
Nsc
m
od
el
s
pe
rfor
m
ances
under
(a
)
CR
=
[0.
5,
0.9]
wit
h
interval
rate
0.1
an
d
(
b)
unde
r
MR
=[0.0
1,
0.1]
with
inter
val
rat
e
0.01.
From
bo
t
h
the
graph
ic
a
l
represe
ntati
on
s
,
the
Ssc
m
od
e
l
has
dem
on
st
r
at
ed
trac
king
a
bili
ty
bette
r
th
an
t
he
Nsc
m
od
el
.
The
prob
a
bili
ti
es
of
t
he
S
sc
m
od
el
in
ge
ner
at
i
ng
go
od
sim
ul
at
ion
m
od
el
w
ere
cl
os
e
to
1
especial
ly
wh
e
n
CR
s
wer
e
hi
gh
as
sh
ow
n
in
Fig
ure
6(a)
.
In
Fig
ure
6
(
b),
t
he
N
sc
m
od
el
was
bette
r
at
the
l
ow
m
utati
on
rat
e
(MR=
[
0.01,
0.02
]
)
bu
t
the
prevail
ing
c
har
act
erist
ic
of
the
Ssc
m
od
el
em
erg
ed
wh
e
n
MR
was
gr
eat
er
tha
n
0.03.
T
he
possib
il
ities
of the
Ssc m
odel
h
avi
ng a
good
forecast
wer
e
v
e
ry close t
o 1
when M
R was
abov
e
0.07.
(a)
(b)
Fig
ure
6
.
Com
par
is
on of
Ssc
and N
s
c m
od
el
s p
e
rfor
m
ances
(prob
a
bili
ti
es)
(a)
unde
r
CR
=[
0.5,0.
9] w
it
h
i
nter
val r
at
e
0.1
;
(b) unde
r
MR
=[0.0
1,0.1]
w
it
h
inte
rv
al
rate
0.01
5.
CONCL
US
I
O
N
AND REC
OM
MEN
D
A
TION
Tw
o
resea
rch
m
od
el
s,
Nsc
a
nd
Ssc
m
od
el
s,
with
diff
e
rent
app
r
oac
hes
i
n
sto
pp
i
ng
c
rite
ria
that
ha
d
been
dev
el
ope
d
a
nd
we
re
nu
m
erical
ly
te
sted
with
f
or
eca
sti
ng
sim
ulati
on
.
T
he
Nsc
m
od
el
ad
op
te
d
t
he
idea
of
set
ti
ng
the
uppe
r
bo
und
for
th
e
fitt
est
chrom
os
om
e
rep
et
it
ion.
I
f
a
c
hrom
os
om
e
was
de
c
la
red
as
t
he
fitt
est
of
it
s
gen
e
rati
on
and
as
bein
g
s
uccessively
sel
ect
ed
f
or
M
ge
ner
at
io
ns
,
the
n
the
searc
hing
process
wou
ld
be
te
rm
inate
d
an
d t
his c
hrom
os
om
e w
as
viewe
d
as
the
best s
ol
ution
.
Nonetheless
,
t
he
Ssc
m
od
el
was
desi
gn
e
d
t
o
a
pply
the
c
oncept
o
f
fitt
est
dom
inancy.
It
is
a
known
fact
that
the
fit
te
st
of
each
ge
ner
at
io
n
w
ou
l
d
ha
ve
a
hi
gh
e
r
pro
bab
il
it
y
in
transm
it
t
ing
t
he
ge
netic
inf
orm
at
ion
.
Gr
a
dual
ly
,
the
popula
ti
on
w
ould
be
dom
inate
d
by
the
fitt
est
’s
trai
t.
If
the
de
ns
it
y
of
sat
ur
at
io
n
or
degr
ee
of
do
m
inati
on
of
the
fitt
est
ha
d
reac
he
d
the
pr
e
-
def
in
ed
boun
dar
y
(
4
)
,
t
hen
the
e
volu
ti
on
was
c
on
s
ider
e
d
com
plete
an
d
t
he fit
te
st of
t
he
last
g
e
ner
at
io
n was
com
m
end
ed
as
the
optim
um
so
luti
on.
In
the
num
erical
te
st,
it
wa
s
f
ound
t
hat
t
he
Ssc
m
od
el
was
m
or
e
ca
pab
le
,
the
pro
bab
il
it
ie
s
of
ob
ta
ini
ng
glob
al
op
ti
m
u
m
par
am
et
er
α
we
re
f
ound
hi
gher
tha
n
Nsc
m
od
el
.
The
N
sc
m
od
el
sho
wed
a
m
od
erate
perf
or
m
ance
with
l
ow
value
s
of
MR
and
CR
.
I
ts
com
petency
was
sli
ghtl
y
i
m
pr
ov
e
d
after
the
MR
and
CR
wer
e
r
ai
sed.
Ge
ne
rall
y,
the
Ssc
m
o
del
dem
on
strat
ed
a
bette
r
s
kill
in
find
i
ng
a
good
so
l
utio
n
w
he
n
MR
and
CR
w
ere
not
hi
gh.
The
ca
pab
il
it
y
of
t
he
Ssc
m
od
el
boost
ed
w
hen
CR
a
nd
MR
had
bee
n
set
high.
Hen
ce
,
it
can
be
con
cl
uded
t
ha
t
the
pro
pose
d
sto
ppin
g
crit
erio
n
ha
s
sho
w
n
a
great
im
pr
ovem
ent
in
enhancin
g
the
al
gorithm
abili
ty
in
so
lvi
ng
the
opti
m
izati
on
pro
blem
and
re
duci
ng
t
he
risk
of
pr
e
m
at
ur
e
co
nver
gen
c
e
.
Fo
r
f
uture
stu
dy
,
one
m
ay
research
on
the
in
flue
nce
of
dif
f
eren
t
gen
et
ic
operat
or
s
i
n
e
voluti
on
.
Th
e
pr
opose
d
stoppin
g
c
rite
rion m
ay
also b
e
test
ed
in
o
t
her fiel
ds
of
op
ti
m
iz
at
ion
s
uch as
eng
i
neer
i
ng.
REFERE
NCE
S
[1]
H.
M.
Pand
e
y
,
e
t
a
l.
,
“
A
compar
at
iv
e
re
vie
w
o
f
a
pproa
che
s
to
pr
e
vent
pr
ematur
e
c
onver
gence
in
G
A,”
Ap
plied
Sof
t
Computing
,
vo
l.
24,
pp
.
1047
-
10
77,
Nov
2014
.
[2]
J.
Zh
ang,
e
t
al
.
,
“
Optimization
of
Ac
tuators
i
n
Sm
art
Truss
Based
on
Gene
ti
c
Algorit
hm
s,”
TEL
KOMNIK
A
Indone
sian J
our
nal
of
Elec
tric
al
Engi
ne
ering
,
vol
/i
ss
ue:
10
(
7
)
,
pp
.
1615
-
1620,
201
2.
[3]
M.
Bolha
san
i
a
nd
S.
Aza
d
i,
“
Para
m
et
er
est
imat
ion
of
ve
hicle
h
andl
ing
m
ode
l
u
sing
gene
t
ic
al
g
orit
hm
,
”
S
ci
en
tia
Iranica
,
vo
l
/i
ss
u
e:
11
(
1
&2
)
,
pp.
121
-
127,
2004
.
[4]
K.
Deb,
“
New
Optimiza
ti
o
n
Techni
ques
in
Eng
ine
er
ing,
”
in
Stu
die
s
in
Fuzzines
s
and
Soft
Com
puti
ng
,
Spring
er
,
Berl
in
,
He
ide
lb
e
rg,
pp
.
13
-
14
,
20
04.
[5]
J.
Andre,
et
al.
,
“
An
improvem
e
nt
of
th
e
stand
ar
d
gene
t
ic
al
gor
ithm
fight
ing
pre
m
at
ure
conv
erg
e
nce
in
continuou
s
opti
m
iz
ation,
”
A
dvanc
es
in Engi
nee
ring S
o
ft
war
e
,
vo
l
/i
ss
ue:
32
(
1
)
,
pp
.
49
-
60
,
20
01.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Th
e s
at
ur
atio
n of po
pu
l
ation f
it
ness
as
a
sto
ppin
g
cri
te
rio
n
i
n gen
et
ic
algor
it
hm
(
Fo
o Fo
ng Ye
ng)
4137
[6]
B.
Benlahbi
b
,
e
t
al.
,
“
W
ind
Far
m
Mana
gement
using
Artifi
c
ia
l
Inte
lligen
t
T
ec
h
nique
s,”
Inte
rna
ti
onal
Journal
o
f
El
e
ct
rica
l
and
C
omputer
Engi
n
e
ering
(
IJE
CE)
,
v
ol
/i
ss
ue:
7
(
3
)
,
p
p
.
1133
,
2017
.
[7]
A.
F.
J.
Al
-
gbur
i,
et
al
.
,
“
An
he
te
roge
n
eous
pop
ula
ti
on
-
b
ase
d
g
e
net
i
c
a
lgori
thm
for
data
c
luste
ri
ng,
”
Indon
esian
Journal
of
Elec
t
rical
Engi
ne
erin
g
and
Informati
c
s
,
vol
/
issue:
5
(
3
)
,
pp
.
275
-
284
,
2
017.
[8]
S.
Jangjit
and
P.
La
oh
ac
h
ai
,
“
Para
m
et
er
Esti
m
at
i
on
of
Thr
ee
-
Pha
se
Induc
t
ion
Mo
tor
b
y
Us
ing
Ge
net
i
c
Algori
thm,
”
Journal
of
Elec
t
rical
Engi
ne
erin
g
&
Technol
ogy
,
vol. 4, no. 3, pp. 360
-
364,
2009.
[9]
H.
Sait
o
and
N.
Tsunashima,
“
Superqua
dri
cs
par
amete
r
esti
m
atio
n
from
shading
i
m
age
using
gen
e
ti
c
al
gor
it
hm
,
”
i
n
Proce
ed
ings
of
I
ECON’94
-
20th
A
nnual
Con
fe
re
nce
of
IE
EE
Ind
ustrial
Elec
troni
cs
Control
and
I
nstrum
ent
ati
on
,
vol.
2
,
pp
.
978
-
9
83
,
1994
.
[10]
A.
P.
Unnikris
hnan,
et
al
.
,
“
Li
c
ense
Pla
te
Loc
a
li
z
at
ion
Us
ing
Gene
t
ic
Al
gorit
hm
inc
lud
i
ng
Color
Fea
tu
re
Ext
ra
ct
ion
,
”
Pro
ce
dia
Technol
og
y
,
vo
l. 24, pp. 1
4
45
-
1451,
2016
.
[11]
F
.
Jola
i,
e
t
al.
,
“
Gene
ti
c
a
lgori
t
hm
for
bi
-
criter
i
a
sing
le
m
ac
hin
e
sch
edul
ing
pr
oble
m
of
m
ini
m
iz
ing
m
axi
m
um
ea
rl
ine
ss
and
nu
m
ber
of
t
ard
y
jo
bs,”
App
lied
Mat
hemati
cs
And
C
omputati
on
,
vol
.
194,
pp.
552
-
56
0,
2007
.
[12]
S.
H.
R.
Pasand
ide
h,
et
al
.
,
“
A
gene
t
ic
a
lgorith
m
for
vendor
m
ana
ged
inve
nto
r
y
cont
rol
s
y
ste
m
of
m
ult
i
-
produc
t
m
ult
i
-
constraint
ec
onom
ic
orde
r
quant
ity
m
od
el,
”
E
xpe
rt
Syst
ems
wit
h
Appl
i
cat
i
ons
,
vol
/
issue:
38
(
3
)
,
pp
.
2708
-
2716,
2011
.
[13]
R.
Venk
atesan
a
nd
V.
Kum
ar,
“
A
geneti
c
al
gor
ithm
s
appr
oac
h
to
growth
phase
fo
rec
ast
ing
of
wi
re
le
ss
subs
cri
ber
s
,
”
Inte
rnational
Jo
urnal
of Forec
as
ti
ng
,
vol
/
issue:
18
(
4
)
,
pp
.
625
-
64
6,
2002
.
[14]
B.
C.
Ervur
a
l,
et
al.
,
“
Model
Estim
at
ion
of
ARM
A
Us
ing
Gene
tic
Algor
it
hm
s:
A Ca
se
Stud
y
of
F
ore
ca
st
ing
Natur
al
Gas Consum
pti
on,
”
Proce
d
ia
-
S
oci
al
and
B
ehavi
oral
Scienc
es
,
v
ol.
235
,
pp
.
537
-
545,
Oct
2016.
[15]
C.
Dj
el
l
al
i
and
M.
Adda,
“
A
ne
w
pre
di
ct
iv
e
app
roa
ch
to
var
i
able
s
select
ion
throu
gh
Gene
ti
c
Algo
rit
hm
and
Fuzz
y
Adapti
ve
Reson
anc
e
Th
eor
y
Us
i
ng
m
edi
c
al
di
ag
nosis
as
a
ca
se
,
”
Proc
edi
a
Computer
Sc
ie
n
ce
,
vo
l.
109,
pp.
448
-
457,
2017
.
[16]
S.
Kal
ant
a
ri
an
d
M.
S.
Abade
h,
“
GA
SA
-
JOS
H :
A
H
y
brid
Evol
uti
on
ar
y
-
An
nea
l
ing
Approa
ch
for
Jo
b
-
Shop
Schedul
ing
Problem,”
vol
/i
ss
ue
:
2
(
2
)
,
pp
.
132
-
14
0,
2013
.
[17]
N.
Dugan
and
Ş.
Erkoç
,
“
Gen
et
i
c
Algorit
hm
s
in
Applicati
on
to
the
Geom
et
r
y
Optimiz
ation
of
Nanopa
rticle
s
,
”
Al
gorithms
,
vo
l
/is
sue:
2
(
1
)
,
pp.
4
10
-
428,
2009
.
[18]
C.
R
.
R.
and
J
.
E
.
Bea
sl
e
y
,
“
In
troduc
ti
on
,
”
in
I
ntroduct
ion
in
Mode
rn
Heurist
ic
Techni
qu
es
f
o
r
Combinatori
al
Proble
ms
,
C
.
R.
Ree
ves,
Ed
.
Ne
w York:
Halst
ed
Press
(an
imprin
t
of
John
W
il
e
y
&
Sons
,
In
c),
pp
.
1
-
19
,
1992
.
[19]
N.
Hass
an
,
et
al.
,
“
Num
ber
of
itera
t
ion
an
aly
sis
for
complex
fss
shape
using
GA
for
eff
ic
i
ent
ESG,”
Bu
ll
e
ti
n
o
f
El
e
ct
rica
l
Eng
in
ee
ring a
nd
Infor
matic
s
,
vo
l
/i
ss
ue
:
7
(
4
)
,
pp.
505
-
5
13,
2018
.
[20]
V.
N.
W
i
jay
an
ingrum
and
W
.
F.
Mahm
ud
y
,
“
Fodder
Com
positi
on
Optimi
za
t
ion
using
Modifie
d
Gene
tic
Algorit
hm
,
”
Inter
nati
onal Journal
of
Intelli
g
ent E
ngine
ering
and
Syste
ms
,
vol
/i
ss
ue:
9
(
3
)
,
pp
.
1
-
8
,
2016.
[21]
X.
Yan,
et
al
.
,
“
An
Im
prove
d
Gene
t
ic
Algor
it
h
m
and
Its
Applicati
on
in
Cla
ss
ifica
t
ion,
”
TEL
KOMNIKA
Indone
s
ian
Journal
of
Elec
t
rical
Engi
ne
erin
g
,
vol
/i
ss
ue:
10
(
1
)
,
pp
.
337
-
346
,
2013.
[22]
S.
Kal
ant
ar
i
and
M.
S
.
Abad
eh,
“
An
Eff
e
ct
iv
e
Multi
Population
Base
d
H
y
br
id
Gene
tic
Algor
ithm
for
Job
Sho
p
Schedul
ing
Problem,”
Bulletin
of
Elec
tri
cal E
ngi
nee
ring a
nd
Info
rm
ati
cs
,
vol
/i
ss
ue:
2
(
1
)
,
pp
.
59
-
6
4,
2013
.
[23]
M.
S.
Hos
sain
,
et
a
l.
,
“
An
im
prove
d
fi
tne
ss
f
unct
ion
for
autom
at
ed
c
r
y
pta
n
aly
s
is
using
g
e
net
i
c
a
lgorithm
,
”
Indone
sian J
our
nal
of
Elec
tric
al
Engi
ne
ering
and
Computer
Sc
ie
n
ce
,
vol
/i
ss
ue:
13
(
2
)
, p
p
.
643
,
201
9.
[24]
G.
Renne
r
and
A.
Eká
rt
,
“
Gene
t
ic
a
lgori
thms
in
computer
ai
d
ed
design,
”
Comput
er
-
Ai
ded
Design
,
vol
/
issue:
35
(
8
)
,
pp.
709
-
726
,
Jul
2003.
[25]
R.
B.
Bah
awe
res
,
e
t a
l.
,
“
Software
T
esti
ng
W
it
h
Gene
tic
Al
gor
it
h
m
,
”
no.
Sept
ember,
pp
.
19
-
21,
20
17.
[26]
C.
Oĝuz
and
M.
F.
Ercan,
“
A
Gene
t
ic
Algorit
h
m
for
H
y
brid
Fl
ow
-
shop
Schedu
li
ng
wi
th
Mul
ti
p
roc
essor
T
asks,”
Journal
of
Sch
ed
uli
ng
,
vol
/
issue:
8
(
4
)
,
pp
.
323
-
35
1,
Jul 20
05
.
[27]
F.
F
.
Ye
ng
,
et
a
l.
,
“
The
Conseq
uenc
e
of
E
li
t
ism
and
R
abbl
e
Se
l
ec
t
ion
Ope
rat
or
in
Gen
et
i
c
Algo
rit
hm
,
”
in
2014
IEE
E
Coll
oqu
iu
m on
Hum
anit
ies
,
Scienc
e
and
E
ngine
ering
,
pp
.
469
-
474
,
2014
.
[28]
A.
R.
Melo
,
et
a
l
.
,
“
Blood
Perfusi
on
Para
m
eter
Esti
m
at
ion
in
Tumors
b
y
m
ea
ns
o
f
a
Gen
et
i
c
Algor
i
thm,”
Proce
di
a
Computer
Scien
ce
,
vol
.
108
,
pp
.
1384
-
1393,
201
7.
Evaluation Warning : The document was created with Spire.PDF for Python.