TELKOM
NIKA
, Vol.13, No
.3, Septembe
r 2015, pp. 1
047
~10
5
3
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i3.1735
1047
Re
cei
v
ed Ma
rch 1
3
, 2015;
Re
vised June
16, 2015; Accepte
d
Jul
y
4
,
2015
Expert System Modeling for Land Suitability Based on
Fuzzy Genetic f
o
r Cereal Commodities: Case Study
Wetland Paddy and Corn
Fitri Insani*
1
, Imas Sukaesih Sitangga
ng
2
, Marimin
3
1,2
Department of Computer S
c
ienc
e, F
a
cult
y of Natural Sci
ence a
nd Math
ematics,
Bogor Agr
i
cult
ural U
n
ivers
i
t
y
,
Bogor 16
68
0, Indon
esi
a
3
Departme
n
t of Agroin
dustria
l T
e
chnolog
y,
F
a
cult
y of Agric
u
ltura
l
T
e
chnol
og
y,
Bogor Agr
i
cult
ural U
n
ivers
i
t
y
,
Bogor 16
68
0, Indon
esi
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: fitri_insa
ni@
apps.i
pb.ac.i
d
1
, imas.sitang
ga
ng@i
pb.ac.i
d
2
,
marimin@ipb.ac.id
3
A
b
st
r
a
ct
Now
adays, thr
eat of fo
od
sho
r
tages
is h
a
p
p
en
in
In
do
nesi
a
. Most of cr
op
s that ar
e co
ns
umed
a
s
ma
in foo
d
are
cerea
l
s commoditi
es. Cerea
l
s cultivatio
n often exp
e
rie
n
ce
s some pr
ob
le
ms in d
e
ter
m
in
in
g
w
hether l
and
i
s
suitab
le
or
not for the
cro
p
s. Ex
pert sys
tem c
an
hel
p
researc
her
an
d practiti
on
ers
to
ide
n
tify lan
d
s
u
itab
ility for ce
real cro
p
s. In
this
rese
arch,
an ex
pert system
mod
e
l of l
and s
u
itab
ility
for
cerea
l
s crop w
a
s built. T
he mode
l impl
ement
ed soft comp
uti
ng metho
d
s to deve
l
op i
n
fere
nce en
gin
e
w
h
ich
combi
nes fu
z
zy system
and
genetic
al
gorit
hm
. Ther
e are 16 par
am
eter
s
to define l
a
nd suitability w
h
ich
consists of 12
nu
meric
para
m
eters
an
d 4
categor
ical
par
ameters. T
w
o ty
pes of cere
a
l
crops that w
e
re
used
in
this st
u
d
y n
a
m
ely
w
e
tland
pa
ddy
an
d
corn. T
r
a
p
e
z
o
i
d
me
mbersh
ip
function
w
a
s u
s
ed to
repr
ese
n
t
fu
zz
y
sets for
nu
meric
a
l
par
a
m
eters.
Geneti
c
al
gorith
m
w
a
s use
d
for
tuni
ng th
e
me
mbe
r
ship
functio
n
of
fu
zz
y
setfor la
nd suita
b
il
ity w
h
ich co
nsists o
f
very suit
abl
e (S1), quite su
itabl
e (S2), mar
g
in
al suita
b
l
e
(S3
)
and
not suita
b
l
e
(N). This exp
e
rt system is
a
b
le to
c
hoos
e l
and s
u
itab
ility
classesfor c
e
re
als usi
ng th
e fu
zz
y
gen
etic mod
e
l
w
i
th accuracy of 90% an
d8
5
%
for corn and
w
e
tland p
addy
respectiv
e
ly.
Ke
y
w
ords
: cer
eal, exp
e
rt system, fu
z
z
y,
gen
etic
alg
o
rith
m, l
and su
itab
ility eval
uatio
n
Co
p
y
rig
h
t
©
2015 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Indone
sia is f
a
cin
g
food sh
ortage p
r
obl
e
m
. Cult
ivation area
s decre
ase ea
ch yea
r
as an
impact of u
n
controlla
ble la
nd conversio
n
to nona
gr
i
c
ultural a
r
ea
s
[1]. Because
of that, farme
r
s
must optimi
z
e their existi
ng land to p
r
odu
ce crop
s
effectively. Most of
Indon
esia
n crop
s
are
cereal
s. Cult
ivation of cereal
s often
experie
n
c
e
s
many probl
ems such a
s
difficultie
s to
determi
ne
whetherth
e lan
d
is
suitabl
e
or n
o
t for
several
spe
c
i
e
s of
ce
real
s whe
r
ea
s
crop
prod
uctivity d
epen
ds
on it
s la
nd q
ualit
y. Meanw
hile
, farmers l
a
ck of
kno
w
le
d
ge a
bout la
n
d
cha
r
a
c
teri
stics an
d suitabili
ty for their crops. In
a
dditi
on, it also n
e
ed lon
g
time to determine l
and
suitability. Therefore an
expert sy
stem is
needed to simplify a pr
ocess to evaluate land
suitability for cereal
s cro
p
s
. Expert sy
stems wh
ich i
n
clu
de
kno
w
l
edge from e
x
perts
can
h
e
lp
farmers an
d agri
c
ultu
ral e
x
ecutive to determin
e
suit
ability of land [2, 3].
In this research the
proble
m
to be
covered is
ho
w to
make
a mo
de
l of expert
system fo
r
land suitabilit
y
evaluation based
on sof
t
computin
g
for cereal co
mmoditie
s
by
combi
n
ing fu
zzy
system
and
g
enetic algo
rit
h
m. The
com
b
ination
of
th
ese
two
meth
ods ha
s
been
implem
ented
for
solving an el
ectro
m
ag
neti
c
field pro
b
le
m [4], formedical data
cl
assificatio
n
[5], and for crew
grou
ping
[6].
Some oth
e
r specifi
c
re
sea
r
che
s
abo
ut fu
zzy
and
ge
ne
tic alg
o
rithm
have al
so
do
ne
by previou
s
rese
arch [7-8].
The p
u
rp
ose
of this
re
sea
r
ch i
s
to
creat
e an
optimi
z
a
t
ion mod
e
l fo
r fuzzy me
m
bership
function
s i
n
f
u
zzy sy
stem
s usi
n
g
ge
netic
algo
ri
thm a
nd build an e
x
pert system
for cereal
s la
nd
suitability evaluation ba
sed on soft computing. The system
based on the genet
ic algorithmis
able to impro
v
e itself whe
n
actual in
pu
t data ar
e av
ailable. The
benefits of th
is re
se
arch a
r
e
gene
rating
a new alternati
v
e
expert system
usi
ng
sof
t
comp
uting
method
s that
can
be
used
for
learni
ng,
de
ci
sion
-ma
k
in
g sup
port and
l
and
d
e
velop
m
ent
for re
se
arche
r
s and pra
c
titione
rs, in
a
particula
r co
mmodity.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 1047 – 10
53
1048
This
re
sea
r
ch
wa
s limited f
o
r two comm
odities
ce
real
s na
med
wetl
and p
addy a
nd corn
with the ca
se study in Bogor.
Ho
wever the ex
pe
rt system mo
d
e
l can b
e
im
proved fo
r othe
r
spe
c
ie
s of crops by m
odif
y
ing its para
m
eters.
2. Res
earc
h
Method
This system contai
ns
t
w
o comp
one
nts namel
y fuzzy
system
and
geneti
c
alg
o
ri
thm. The
system frame
w
ork
ca
n be
see
n
in
Fi
gure 1. Kno
w
led
ge resource
of the sy
stem
is obtai
ned f
r
om
soil an
d land
experts f
r
om
Dep
a
rtme
nt of Soil
Science and L
and
Re
sou
r
ces, B
ogor Ag
ri
cult
ural
University an
d Indon
esi
an
Cente
r
for Ag
ricult
u
r
al
Lan
d Re
so
urce
s
Re
sea
r
ch an
d Devel
opme
n
t,
Ministry of A
g
riculture. Kn
owle
dge
re
so
urce i
s
al
so
obtaine
d fro
m
text books reg
a
rdi
ng L
and
Suitability Evaluation. T
h
e knowl
edge i
s
used as
i
n
put values i
n
inference engine
and trai
ning
data for tunin
g
fuzzy mem
bership fun
c
ti
on.
Us
e
r
W
o
rk
st
a
t
i
o
n
Us
e
r
I
n
t
e
r
f
a
c
e
Kno
w
l
e
dg
e
A
c
quti
t
i
o
n
W
o
rk
st
a
t
i
o
n
L
a
nd
S
u
i
t
a
b
ili
t
y
E
x
p
e
r
t
S
y
s
t
e
m
M
o
de
lin
g
fo
r
C
e
r
e
a
l
s
C
o
m
m
o
d
iti
e
s
Ex
p
e
rt
K
n
o
w
le
d
g
e
A
c
q
u
it
iti
o
n
a
n
d
R
e
p
r
e
s
e
n
t
a
t
i
o
n
Kn
o
w
l
e
dg
e
B
a
s
e
In
f
e
r
e
n
c
e
E
n
g
i
n
e
G
e
ne
tic
A
l
g
o
r
i
t
h
m
fo
r
Tu
n
i
n
g
Me
m
b
e
r
s
h
ip
Fu
nc
tio
n
K
n
o
w
l
e
dg
e
R
e
pre
s
e
n
ta
ti
o
n
AN
D
AN
D
Cl
a
s
s
1
Cl
a
s
s
2
P
a
ra
m
e
t
e
r 1
P
a
ra
m
e
t
e
r 2
P
a
ra
m
e
t
e
r 3
P
a
ra
m
e
t
e
r 4
P
a
ra
m
e
t
e
r 5
AN
D
Cl
a
s
s
3
P
a
ra
m
e
t
e
r x
P
a
ra
m
e
t
e
r y
.
.
.
.
Non
F
u
z
z
y
P
a
ra
met
ers
Fu
z
z
y
P
a
ra
m
e
t
ers
Ru
le
Ba
se
P
a
ra
met
er Gro
u
p
i
n
g
Figure 1. System Frame
w
o
r
k
2.1. Fuzz
y
Fuzzy is u
s
e
d
in the
inference
system
to rep
r
e
s
ent
human
kno
w
l
edge
whi
c
h
n
o
t alway
s
exactly true
o
r
false. Fu
zzy
can
re
pre
s
e
n
t val
ues
of
variabl
e
u
s
ing
membe
r
ship degree su
ch as
very bad,
ba
d, mode
rate,
goo
d, an
d v
e
ry go
od.
In ca
se of
the
l
and suita
b
ility
system, no
t
all
variable
s
a
r
e
rep
r
e
s
ente
d
in fuzzy
set, only some
of t
hem can be
repre
s
e
n
ted a
s
fuzzy variab
les
su
ch a
s
tem
p
eratu
r
e, hu
mi
dity, and rai
n
fall. Some oth
e
r vari
able
s
a
r
e n
o
t re
pre
s
ented a
s
fu
zzy
variable
s
be
cause input value
s
from e
x
perts
an
d textbook a
r
e not numeri
c
a
l
values but in
ordin
a
l value
s
su
ch a
s
lo
w, mode
rate,
and hi
gh wi
thout kno
w
in
g its values.
The non
-fuzzy
variables
are drai
nage, texture
and erosion ri
sk.
The two kind
v
a
riabl
es (fuzzy and non-fuzzy
)
are
sep
a
rate
d becau
se th
e fuzzy varia
b
les
will be t
uned u
s
in
g G
enetic Alg
o
rit
h
m (GA
)
. But the
non-fu
zzy variable
s
are no
t proce
s
sed
usin
g GA
. The two kind v
a
riabl
es a
r
e
combi
ned in
the
next step after GA tuning.
The fu
zzy a
p
p
roa
c
h
u
s
ed
in this
syste
m
is Su
geno.
We
ado
pts t
h
is a
p
p
r
oa
ch
becau
se
the purp
o
se of this syste
m
is to prod
u
c
e cl
asse
s of land suitability which
con
s
i
s
ts of S1, S2, S3,
and N.S1
me
ans “very suit
able”,
S2
m
e
ans “quite
sui
t
able”, S3 me
ans
“ma
r
gin
a
l
l
y suitable
”
, a
n
d
N me
an
s “no
t
suitable
”
. T
hose value
s
f
o
rm
con
s
e
q
u
ence of fu
zzy
rule. F
o
r
sim
p
licity, the la
nd
suitability cla
s
ses
rep
r
e
s
e
n
ted as 1 fo
r S1, 2 for S2,
3 for S3, and 4 for N.
As the a
n
teceden
ce, vari
able
s
of lan
d
pr
o
p
e
r
ties
mentione
d a
bove are u
s
ed. Th
e
membe
r
ship functio
n
(MF)
of the variables fo
rm
s the trape
zoid
al shape. The
r
e
are 15 vari
ab
les
of land p
r
ope
rties. Ea
ch variabl
e contai
ns
some
M
F
. Numb
er of li
ngui
stic term
s in MF fo
r e
a
ch
variable
ra
ng
es from 2 to
7 term
s. If all 15 varia
b
le
s are
use
d
to
form a
singl
e
rule
set, the
n
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Expe
rt Syste
m
Modeling for Lan
d Suita
b
ilityba
s
e
d
on
Fuzzy G
enet
ic for Cere
al
… (Fitri Insan
i
)
1049
there
are
ab
o
u
t 4
15
(about
one billi
on) rules
gen
erated. It re
sults
high
com
puta
t
ion co
st fo
r the
system espe
cially
in apply
i
ng
the gen
etic
al
gorit
hm.T
o solve this p
r
oble
m
, the 1
5
varia
b
le
s a
r
e
c
a
te
go
r
i
z
e
d in
to
s
i
x gr
ou
ps
: te
mp
er
a
t
ure
,
r
o
o
t
ing
me
dia, nutri
ent retention, e
r
o
s
ion, floo
d p
o
o
l,
and la
nd p
r
e
paratio
n [9]. If each
group
has
2 vari
abl
es a
nd e
a
ch
grou
p ha
s
4
MF, there
are
4
2
=
16 rul
e
s i
n
a
grou
p which
need lo
w
co
mputation
co
st
for the
syst
em to pro
c
e
s
s such num
b
e
r of
rule
s.
2.2. Gene
tic Alg
o
rithm
These wo
rkt
une
s
fu
zzy
membe
r
ship
function
s usi
ng
n
a
tural selectio
n
a
n
d
gen
etic
mec
h
anis
m
called the genetic
algor
i
thm [6], [
10-
12]. It is
us
ed to
optimiz
e fuzzy members
h
ip
function
s. Fig
u
re
2 sho
w
s
a usual g
ene
tic algo
rith
m
(GA)
stag
e which
co
nsi
s
ts of initializati
o
n
,
evaluation, selectio
n, cro
s
sover, an
d mutation [
13].
The GA stag
e of this system wa
s done
for
each gro
up of
land suita
b
ility paramete
r
s
to
incre
a
se the system p
e
rf
orma
nce.
Figure 2. Gen
e
tic algo
rithm
process
2.2.1. Indiv
i
dual Repres
entatio
n
The m
a
in
problem
of G
A
is h
o
w to
rep
r
e
s
e
n
t a
pro
b
lem
int
o
a
ch
rom
o
some
o
r
individual. Th
e individual
repre
s
e
n
tation
can b
e
se
e
n
in Figu
re
3. The figure
sho
w
s
sin
g
le
individual d
e
sign. The indiv
i
dual
s forme
d
by severa
l g
ene
s in whi
c
h
the pro
c
e
ss i
s
de
scrib
ed a
s
follows: An individual Cr
rep
r
e
s
ent
s a
fuzzy rule
set. In other word, each individual con
t
ains
several
rule
s. Lets define
each rule
as Cr
i
chromo
some, whe
r
e
the num
ber
of Cr
i
sh
ows t
he
numbe
r of rul
e
s. Cr
i
is calle
d as a chro
m
o
som
e
be
cau
s
e this i
s
a pi
ece of individ
ual.
A rule
(singl
e Cr
i
) contai
ns va
riabl
es whi
c
h
are
sep
a
rate
d to
ante
c
ed
en
ces
and
a
con
s
e
que
nce
.
As descri
b
e
d
in se
ction 2
.
1, antec
ed
en
ce is trape
zoi
dal form
s me
aning that ea
ch
variable
ne
ed
s fou
r
gen
es (four
point
s).
The
gen
e n
u
m
ber ne
ede
d
in
a
ch
romo
some i
s
ba
sed
on
the num
ber o
f
its variabl
e.
So the nu
mb
er of g
ene
s i
n
a
chromo
so
me (singl
e rul
e
) i
s
(4 × n)
+1,
whe
r
e
n i
s
nu
mber of va
ria
b
les. In
thi
s
case
the
r
e
are
four poi
nts o
f
trape
zoi
d
, a
nd o
n
e
varia
b
l
e
of the con
s
eq
uen
ce. The t
o
tal gene fo
r an individu
al can b
e
cal
c
ul
ated usi
ng thi
s
formul
a ((4
×
n)
+1)
×
r, wh
ere
r is the n
u
m
ber
of rul
e
s.
For ex
a
m
ple,
in Figu
re 3
we have fu
zzy
set of 16
rul
e
s
and 2 vari
abl
es, the total gene
s ne
ede
d to rep
r
e
s
en
t the rule set is ((
4 × 2
)
+ 1) × 1
6
)
= 1
4
4
gene
s.
Figure 3. Individual and
ch
romo
som
e
re
pre
s
entatio
n
2.2.2. Initializ
ation
The
popul
atio
n si
ze
nam
el
y the nu
mbe
r
of individ
ual
s in GA
can
b
e
defin
ed
as
need
ed
.
Let’s
define
P as th
e po
p
u
lation
size. In this
ca
se,
a set P indivi
dual
s was ge
nerate
d
a
s
i
n
itial
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 1047 – 10
53
1050
popul
ation. The first individ
ual wa
s creat
ed ba
sed
on i
n
formatio
n from referen
c
e
s
and/o
r
expe
rts
and the P- 1 i
ndividual
s we
re gen
erated
rand
omly.
2.2.3. Ev
aluation
Each in
dividu
al is eval
uate
d
usi
ng a fitn
ess
fun
c
tion t
o
get their fit
ness valu
e. Fitness
function u
s
e
d
in this ca
se
is mea
n
sq
u
a
re e
r
ror (M
SE). First, ea
ch individ
ual
gene
rated in
the
initialization stage
or re
co
mbination re
sult
(cross
ove
r
and mutatio
n
) is
conve
r
te
d to a fuzzy rule
set. Then t
r
ai
ning dat
a are
tested to e
a
c
h in
dividual
so that the
n
u
mbe
r
of false cla
s
s for e
a
ch
individuali
s
o
b
tained. The
numbe
r of false
cla
ss i
s
used to cal
c
ulate the M
SE. Because
the
evaluation val
ue is an e
r
ror value, smalle
r evaluation v
a
lue indi
cate
s better individ
ual re
sult.
2.2.4. Selection
Selection
is
a metho
d
to
cho
o
se pa
re
nts to b
e
cro
s
sed ove
r
o
r
mutated. To
maintain
several b
e
st
i
ndividual
s, eli
t
ism
wa
s d
o
n
e
by
cho
o
si
n
g
at le
ast
10
% of po
pulati
on from the
b
e
st
individual
s of
the initial p
o
p
u
lation o
r
re
combine
d
p
o
p
u
lation. 90%
remai
n
ing
ne
eded i
ndividu
als
are sele
cted randomly u
s
in
g the roulette
whe
e
l app
roa
c
h.
2.2.5. Cros
sov
e
r
Cro
s
sove
r which i
s
used
in this syst
em is
max-min-a
r
ithmeti
c
al cro
s
sover [14]. The
numbe
r of in
dividual
s sele
cted is b
a
sed
on the pr
ob
a
b
ility of crossover (P
c). Pc along with P
m
(proba
bility of mutation) was determine
d at the
beginning of GA.
If
and
are crosse
d, we
gene
rate fou
r
new individu
als:
1
1
,
′
,
′
is eithe
r
a consta
nt, or a
variable
wh
ose valu
e de
pend
s on th
e
age of the p
opulatio
n. Th
e
resulting offspring a
r
e the
two be
st
of the four individ
uals a
bove.
All children a
s
cro
s
sover result
s are co
mb
ined with the
parents. The
combin
a
t
ion
of
pare
n
ts of
childre
n is
cal
l
ed intermedi
ate pop
ulatio
n. The inte
rmediate p
o
p
u
lation i
s
set as
pare
n
t for the
next stage i.e mutation.
2.2.6. Muta
tion
Several g
ene
s of inte
rmed
iate pop
ulatio
n we
re
sel
e
ct
ed rando
mly as m
u
tation
obje
c
ts.
The num
be
r of gene
s sel
e
cted i
s
ba
se
d on Pm. Pm is mutation
prob
ability of entire ge
ne
s in
popul
ation. M
u
tation wa
s d
one by shifting the sel
e
cte
d
gene to th
e
left or right.
The left and
right
shifting ba
rri
e
r
is dete
r
mine
d by following
formula
s
:
c
kl
= c
k
(c
k
c
k-1
)/2;
c
kr
= c
k
+
(c
k+1
c
k
)/2;
W
h
er
e c
kl
: lef
t
s
h
ifting limit; c
kr
: right shifting limit; c
k
: mutated gen
e
value; c
k-1
: the left g
ene
of
c
k
; c
k+1
: the right gene of c
k
.
For
each g
e
ne sele
cted,
the directio
n
of
the mutati
on is dete
r
mi
ned u
s
in
g a
rand
om
value. If the random val
u
e produces
0 then the gene
shifting di
rection is to the left, but if th
e
rand
om valu
e
pro
d
u
c
e
s
1
then th
e ge
ne
shifting
di
rection is to the
right. The
shift
i
ng di
stan
ce
of
the gene i
s
d
e
termin
ed by followin
g
formula:
′
∆
,
;
∆
,
;
Whe
r
e t is the curre
n
t gen
eration
seq
u
e
n
ce a
nd
∆
( t, y) is a function that return
a value in range
[0,y] so that the probability
of
∆
( t,
y) is
c
l
os
e to 0 increas
e
s
as
t inc
r
eases
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Expe
rt Syste
m
Modeling for Lan
d Suita
b
ilityba
s
e
d
on
Fuzzy G
enet
ic for Cere
al
… (Fitri Insan
i
)
1051
2.3. Rule Based
As sh
own in
Figure 1, all non-fu
zzy pa
ramete
rs
and
fuzzy pa
ram
e
ters th
at ha
ve been
tuned in p
r
ev
ious
stage
s
were combin
ed usi
ng if
-th
en rule
ba
se.
The dete
r
mi
nation facto
r
o
f
land
suitabilit
y is gai
ned
b
y
lookin
g at t
he
worst val
ue of the
pa
rameter. F
o
r
example, if t
h
e
cla
s
ses of 5
variable
s
(bo
t
h fuzzy an
d
non-fu
zz
y va
riable
s
) a
r
e
S1, S1, S2, S3, S1, then
we
kno
w
that the
worst value i
s
S3. So it ca
n be det
e
r
mi
ned that the final cla
s
s of the 5 varia
b
le
s is
S3.
2.4. Resul
t
s and
Analy
s
is
We carry out experim
ents f
o
r wetla
nd pa
ddy and corn
with the follo
wing p
a
ra
met
e
rs:
a) Population
si
ze:
20
b)
Probability of
crossover: P
c
= 0.6,
= 0..35
c)
Probability of
mutation: P
m
=
0.
1
Figure 4
an
d Fig
u
re
5
provide
the
fitness value
s
fo
r
wetlan
d pa
ddy a
n
d
corn
respe
c
tively.
The
fig
u
re
s show
that
the
best and
ave
r
age
fitne
s
s were getting better until
1
5
th
gene
ration
a
nd the
n
it
rea
c
h it
s
co
nvergen
ce.
Ho
we
ver, the
wo
rst fitness
ke
ep
s flu
c
tuating
t
hat
may be cau
s
ed by the mutation pro
c
e
ss that does not
always return better re
sult
s.
Figure 4. Fitness value
s
for wetland p
a
d
d
y
Figure 5. Fitness value
s
for co
rn
The variatio
n
s
of Pc and Pm also affect
to
the fitness
results. Tabl
e
1 and Table
2 sho
w
the fitness of
Pm variations of wetland p
addy and
corn respe
c
tively.
Table 1. Effect of Pm variation to
the tuning re
sult of wetlan
d pad
d
y
Experime
nt
Pc
Pm
Last
B
est Fit
n
e
ss
Last
A
v
e
r
age Fi
tness
1
0.6
0.001
0.010000
0.010000
2
0.6
0.010
0.010000
0.010063
3
0.6
0.100
0.010000
0.010563
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 1047 – 10
53
1052
Table 2. Effect of Pm variation to the tuning re
sult of corn
Experime
nt
Pc
Pm
LastB
est Fitn
es
s
Last
A
v
e
r
age Fi
tness
1
0.6
0.001
0.017
500
0.020375
2
0.6
0.010
0.010
000
0.012125
3
0.6
0.100
0.010
000
0.010500
The be
st Pm and Pc were used to cre
a
te a land su
itability system. The main
page of
Land S
u
itabili
ty system
ca
n be
se
en in
Figure 6. A
u
s
er ne
ed
s to
fill land p
a
ra
meter valu
es in
the form the
n
the system
d
i
splay
s
output
in the rig
h
t b
o
ttom side. T
he re
sult i
s
n
o
t only for si
n
g
le
comm
odity but it may show multiple land su
itability both for corn and wetland ri
ce.
Figure 6. The
main page of
the Land Sui
t
ability Syste
m
The
system
h
a
s
been
teste
d
by comp
ari
ng
its output and
la
nd suit
ability
judge
m
ent
from
experts. 2
0
l
and
suitabilit
y data of wet
l
and ri
ce
and
corn from
e
x
pert sh
ows
that 85% of the
system outp
u
t
is appro
p
ria
t
e to the expert esti
matio
n
for wetlan
d
rice commo
dity, meanwh
ile
90% is
appropriate for
c
o
rn c
o
mmodity.
3. Conclu
sion
This
work de
veloped
a m
odel of
soft comp
uting
b
a
se
d exp
e
rt
sy
stem
that
combi
ne
fuzzy
system
and g
eneti
c
a
l
gorithm to
de
termine
l
and
suitability
for cereal
s com
m
odity.
Gene
tic
algorith
m
wa
s u
s
ed to tu
ne mem
bership fun
c
tion
s by addin
g
n
e
w a
c
tual
da
ta. By using
the
actual
data
a
s
training
d
a
ta, the
system
ca
n im
pr
ove
its
infe
ren
c
e
engin
e
to get better re
sult. By
experim
entin
g variation
s
of Pm, we ge
t different
be
st re
sult b
e
tween
corn an
d
wetlan
d pa
d
d
y,
best result for wetland
p
addy were
p
r
odu
ce
d by
Pm 0.001 a
nd be
st re
su
lt for co
rn
were
produced
by Pm 0.1. T
h
i
s
expert
syst
em is abl
e t
o
determine
the land
suit
abilityclasses of
cereal
s
usin
g
the fu
zzy g
enetic mo
del
with
a
c
cura
cy of
90% fo
r
corn
and
8
5
% for wetla
nd
paddy.
In this
res
e
arc
h
, we only implement two k
i
nd of cereals, but a
c
t
ually the syst
em wa
s
desi
gne
d for land
suita
b
il
ity of any ki
nd of pl
ant
s.
We
also
su
gge
st to u
s
e
the real
dat
a to
improve its a
c
cura
cy.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Expe
rt Syste
m
Modeling for Lan
d Suita
b
ilityba
s
e
d
on
Fuzzy G
enet
ic for Cere
al
… (Fitri Insan
i
)
1053
Referen
ces
[1]
Direktorat
Ris
e
t
dan
Ka
jia
n S
t
rategis Instit
ut Pertan
ia
n B
o
gor. Pers
pektif
Baru
Pem
ban
gun
an
unt
uk
Mena
ngg
ul
ang
i Krisis Pan
gan
dan Ener
gi. B
ogor
. Institut Pertani
an Bo
gor
. 2008: 1-1
03.
[2]
Sevani N, Marimin, Suk
o
co H. Sistem P
a
kar Pe
ne
ntua
n Kes
e
sua
i
a
n
Lah
an
Berd
a
s
arkan F
a
kt
or
Peng
ham
bat
T
e
rbesar (Maximum Lim
i
tatio
n
F
a
ctor) Unt
u
k T
anaman
Pang
an.
Jurnal Informatika
.
200
9; 10(1): 23
-31.
[3]
Hartati S, Sitangga
ng IS. A Fu
zz
y
Base
d Dec
i
sion
Su
pp
ort S
y
stem for Evalu
a
ting L
a
n
d
Suitab
ilit
y
an
d
Selecti
ng Cro
p
s
.
Journal of C
o
mputer Sci
e
n
c
e
. 2010; 6(
4): 417-
424.
[4]
Damousis IG, Satsios KJ, Labridis
DP. Comb
ined Fuzz
y
Logic And Genet
ic Algorithm
T
e
chniques-
Appl
icatio
n to An Electroma
g
netic F
i
el
d Pro
b
lem.
F
u
zz
y
S
e
ts and Syste
m
s
. 20
02; 12
9(
3): 371-3
86.
[5]
Den
n
is B, Mut
hukris
hna
n S. AGF
S
: Adaptive G
enetic F
u
z
z
y
S
y
stem for
Medic
a
l D
a
ta
Classific
a
tio
n
.
Appl
ied S
o
ft Computi
n
g
. 20
1
4
; 25: 242-
252.
[6]
Cord
on O, Herrera F, Villar P. Generatin
g the Kn
o
w
l
e
d
ge B
a
se of a Fuzz
y Rule-Bas
ed S
y
stem b
y
th
e
Genetic Le
arn
i
ng of the Data
Base.
IEEE Tr
ansactions on Fu
z
z
y
System
s
. 2001; 9(4): 66
7-67
4.
[7]
Casil
l
as J, C
o
r
d
´on O, de
l Je
sus M, Herrer
a
F
.
G
enetic tun
i
ng
of fuzz
y
r
u
l
e
de
ep structur
es pres
ervin
g
interpr
e
tabi
lit
y
and
its i
n
teract
io
n
w
i
th fuzzy
rule set reduction.
IEEE Transactions
on Fuzz
y
System
s
.
200
5; 13(1): 13
-29.
[8]
W
ang Y,
Ch
en
L, C
hen
Z
.
A
F
u
zz
y Se
lf-tunin
g
C
ontro
lle
d F
e
e
d
in
g S
e
r
v
o S
y
stem
of
Machi
ne T
ool.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2013; 1
1
(5): 2
351-
235
8.
[9]
Djae
n
u
ddi
n D, Mar
w
an H, Su
bag
yo
H. Hida
y
at A.
Petunj
u
k
T
e
knis Evalu
a
si La
han U
n
tuk Komod
i
tas
Pertani
an. Bo
gor: Bala
i Be
sar Litba
ng S
u
mber
da
ya L
a
han Perta
n
i
an.
Badan L
i
tba
n
g
Pertani
an.
Dep
a
rtemen P
e
rtani
an. 20
11:
1-154.
[10]
Herrera F
,
Lo
zano M, Verd
ega
y JL. T
uning
F
u
zz
y Lo
gic Co
ntroll
er
s b
y
Gen
e
tic
Algorithms.
Internatio
na
l Journ
a
l of Appr
oxi
m
ate R
easo
n
in
g
. 199
5; 12: 299-3
15.
[11]
Casil
l
as J, C
o
r
d
´on O, de
l Je
sus M, Herrer
a
F
.
Genetic tun
i
ng
of fuzz
y
r
u
l
e
de
ep structur
es pres
ervin
g
interpr
e
tabi
lit
y
and
its i
n
teract
io
n
w
i
th fuzzy
rule set reduction.
IEEE Transactions
on Fuzz
y
System
s
.
200
5; 13(1): 13
-29.
[12]
Qin G, Ma H.
An Intelli
ge
nt Cours
e
Sche
d
u
lin
g Mo
del B
a
sed o
n
Gen
e
t
ic Algorithm.
TELKOMNIKA
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2014; 1
2
(4): 2
985-
299
4.
[13]
Eibe
n AE, Sm
i
t
h JE. Introd
uc
tion to
Evo
l
uti
onar
y
Com
puti
ng. N
a
tura
l C
o
mputin
g S
e
ries
. Ne
w
York:
Sprin
ger. 20
03
: 1-301.
[14]
Jain
a R, Sivak
u
mara
n N, Ra
dhakr
i
shn
an T
K
. De
sign
of self tunin
g
fuz
z
y
co
ntroll
ers for nonl
in
ear
s
y
stems.
Expe
rt Systems w
i
th Applic
ations
. 2
011; 38: 4
466-
447
6.
Evaluation Warning : The document was created with Spire.PDF for Python.