T
E
L
KO
M
N
I
KA
T
e
lec
om
m
u
n
icat
ion
,
Com
p
u
t
i
n
g,
E
lec
t
r
on
ics
an
d
Cont
r
ol
Vol.
1
8
,
No.
4
,
Augus
t
2020
,
pp.
2112
~
2117
I
S
S
N:
1693
-
6930,
a
c
c
r
e
dit
e
d
F
ir
s
t
G
r
a
de
by
Ke
me
nr
is
tekdikti
,
De
c
r
e
e
No:
21/E
/KP
T
/2018
DO
I
:
10.
12928/
T
E
L
KO
M
NI
KA
.
v1
8
i
4
.
14032
2112
Jou
r
n
al
h
omepage
:
ht
tp:
//
jour
nal.
uad
.
ac
.
id/
index
.
php/T
E
L
K
OM
N
I
K
A
M
e
d
i
u
m
t
e
r
m
l
oad
d
e
m
a
n
d
f
o
r
e
c
as
t
of
Kan
o z
on
e
u
si
n
g n
e
u
r
al
n
e
t
w
o
r
k
al
gor
it
h
m
s
Hu
z
aim
u
L
awal
I
m
a
m
1
,
M
u
h
am
m
ad
S
an
i
Gaya
2
,
G.
S.
M
.
Gal
ad
an
c
i
3
1
,3
D
e
p
art
me
n
t
o
f
P
h
y
s
i
c
s
,
Bay
ero
U
n
i
v
ers
i
t
y
K
a
n
o
,
N
i
g
e
ri
a
2
D
ep
ar
t
men
t
o
f
E
l
ect
r
i
cal
E
n
g
i
n
eeri
n
g
,
K
a
n
o
U
n
i
v
ers
i
t
y
o
f
Sci
en
ce
a
n
d
T
ec
h
n
o
l
o
g
y
,
N
i
g
eri
a
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
S
e
p
3
,
2019
R
e
vis
e
d
M
a
r
1
,
2020
Ac
c
e
pted
M
a
r
23
,
2020
E
l
ec
t
ri
c
i
t
y
l
o
a
d
fo
recas
t
i
n
g
refers
t
o
p
r
o
j
ec
t
i
o
n
o
f
fu
t
u
re
l
o
ad
req
u
i
reme
n
t
s
o
f
an
area
o
r
reg
i
o
n
o
r
c
o
u
n
t
r
y
t
h
ro
u
g
h
ap
p
ro
p
ri
a
t
e
u
s
e
o
f
h
i
s
t
o
r
i
ca
l
l
o
ad
d
at
a.
O
n
e
o
f
s
ev
era
l
ch
al
l
en
g
es
faced
b
y
t
h
e
N
i
g
eri
a
n
p
o
w
er
d
i
s
t
ri
b
u
t
i
o
n
s
ec
t
o
r
s
i
s
t
h
e
o
v
er
l
o
a
d
ed
p
o
w
er
d
i
s
t
ri
b
u
t
i
o
n
n
et
w
o
r
k
w
h
i
ch
l
e
ad
s
t
o
p
o
o
r
v
o
l
t
ag
e
d
i
s
t
ri
b
u
t
i
o
n
a
n
d
freq
u
en
t
p
o
w
er
o
u
t
ag
e
s
.
A
ccu
ra
t
e
l
o
a
d
d
eman
d
fo
recas
t
i
n
g
i
s
a
k
ey
i
n
ad
d
res
s
i
n
g
t
h
i
s
c
h
al
l
en
g
e.
T
h
i
s
p
a
p
er
p
re
s
en
t
s
a
co
m
p
ari
s
o
n
o
f
g
en
eral
i
ze
d
reg
res
s
i
o
n
n
eu
ra
l
n
et
w
o
r
k
(G
RN
N
),
feed
-
fo
r
w
ar
d
n
eu
ral
n
et
w
o
r
k
(FFN
N
)
an
d
rad
i
al
b
as
i
s
fu
n
ct
i
o
n
n
eu
ra
l
n
et
w
o
rk
fo
r
med
i
u
m
t
erm
l
o
a
d
d
eman
d
es
t
i
mat
i
o
n
.
E
x
p
eri
me
n
t
a
l
d
a
t
a
fro
m
K
a
n
o
e
l
ect
r
i
ci
t
y
d
i
s
t
ri
b
u
t
i
o
n
co
mp
a
n
y
(
K
E
D
CO
)
w
ere
u
s
ed
i
n
v
a
l
i
d
at
i
n
g
t
h
e
mo
d
el
s
.
T
h
e
s
i
mu
l
at
i
o
n
res
u
l
t
s
i
n
d
i
ca
t
ed
t
h
at
t
h
e
n
eu
ra
l
n
et
w
o
r
k
mo
d
e
l
s
y
i
el
d
ed
p
r
o
mi
s
i
n
g
res
u
l
t
s
h
a
v
i
n
g
ach
i
e
v
ed
a
mean
ab
s
o
l
u
t
e
p
ercen
t
ag
e
err
o
r
(MA
PE
)
o
f
l
e
s
s
t
h
an
1
0
%
i
n
al
l
t
h
e
co
n
s
i
d
ered
s
ce
n
ari
o
s
.
T
h
e
g
en
era
l
i
za
t
i
o
n
cap
a
b
i
l
i
t
y
o
f
F
FN
N
i
s
s
l
i
g
h
t
l
y
b
e
t
t
er
t
h
a
n
t
h
a
t
o
f
RBFN
N
an
d
G
RN
N
mo
d
el
.
T
h
e
mo
d
el
s
co
u
l
d
s
er
v
e
as
a
v
a
l
u
a
b
l
e
an
d
p
r
o
mi
s
i
n
g
t
o
o
l
fo
r
t
h
e
fo
reca
s
t
i
n
g
o
f
t
h
e
l
o
ad
d
eman
d
.
K
e
y
w
o
r
d
s
:
C
a
pa
bil
it
y
L
a
ye
r
L
oa
d
Ne
ur
a
l
ne
twor
k
W
e
ight
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e
.
C
or
r
e
s
pon
din
g
A
u
th
or
:
M
uha
mm
a
d
S
a
ni
Ga
ya
,
De
pa
r
tm
e
nt
of
E
lec
tr
ica
l
E
nginee
r
ing,
Ka
no
Unive
r
s
it
y
of
S
c
ienc
e
a
nd
T
e
c
hnology,
02
Ga
ya
R
oa
d,
W
udil
C
it
y
,
713211
,
Nige
r
ia.
E
mail:
muhdgaya
s
a
ni@gm
a
il
.
c
om
1.
I
NT
RODU
C
T
I
ON
E
lec
tr
icity
load
f
or
e
c
a
s
ti
ng
is
a
n
e
s
s
e
nti
a
l
pa
r
t
of
powe
r
s
ys
tem
e
ne
r
gy
mana
ge
ment.
L
oa
d
f
or
e
c
a
s
t
r
e
f
e
r
s
to
e
s
ti
mating
the
f
utur
e
load
thr
ough
the
us
e
of
his
tor
ic
a
va
il
a
ble
load
da
ta.
I
t
is
a
ke
y
in
the
planning,
ope
r
a
ti
on
a
nd
dis
pa
tch
of
e
lec
tr
ica
l
e
ne
r
g
y
.
Ap
pr
opr
iate
load
pr
e
diction
p
r
ovides
e
lec
tr
icity
c
ompanie
s
or
gove
r
nments
with
ti
mely
a
nd
a
de
qua
te
in
f
or
mation
to
ope
r
a
te
the
s
ys
tem
e
c
onomi
c
a
ll
y
a
nd
r
e
li
a
bly
.
L
oa
d
f
or
e
c
a
s
t
is
c
r
it
ica
l
a
nd
ne
c
e
s
s
a
r
y
be
c
a
us
e
the
a
va
il
a
bil
it
y
of
e
lec
tr
icity
is
one
o
f
t
he
mos
t
i
mpor
ta
nt
f
a
c
tor
s
f
or
indus
tr
ial
de
ve
lopm
e
nt
e
s
pe
c
ially
in
a
de
ve
lopi
ng
c
ountr
y
li
ke
Nige
r
ia.
S
ome
of
the
main
is
s
ue
s
with
the
Nige
r
ian
powe
r
s
e
c
tor
include
h
igh
ope
r
a
ti
ng
c
os
t,
h
igh
e
ne
r
gy
los
s
e
s
a
nd
high
s
uppr
e
s
s
e
d
de
mand
thr
oughout
th
e
c
ountr
y.
T
he
dis
tr
ibut
ion
s
e
c
tor
is
tas
ke
d
with
t
he
ne
e
d
to
e
ns
ur
e
a
de
qua
te
ne
two
r
k
c
ove
r
a
ge
a
nd
p
r
ovis
ion
of
qua
li
ty
powe
r
s
upply
to
the
publi
c
in
a
ddit
ion
to
s
uf
f
icie
nt
mar
ke
ti
ng
a
nd
c
us
tom
e
r
s
e
r
vice
de
li
ve
r
y
.
T
o
r
e
duc
e
the
high
tec
hnica
l
los
s
e
s
a
nd
im
pr
ove
th
e
qua
li
ty
of
volt
a
ge
dis
tr
ibut
ion
a
t
the
e
lec
tr
icity
d
is
tr
ibut
ion
s
e
c
tor
ther
e
is
ve
he
ment
ne
e
d
of
c
ons
tant
ne
twor
k
upgr
a
de
or
ove
r
ha
uli
ng
whic
h
may
not
be
a
c
hieve
d
without
a
c
c
ur
a
te
load
de
mand
f
o
r
e
c
a
s
t.
A
c
c
ur
a
t
e
lo
a
d
f
o
r
e
c
a
s
t
g
r
e
a
tl
y
i
nf
l
ue
n
c
e
s
t
he
p
la
nn
in
g
p
r
o
c
e
s
s
u
nd
e
r
ta
ke
n
in
o
pe
r
a
t
io
n
c
e
n
t
r
e
s
o
f
e
ne
r
g
y
p
r
o
vi
de
r
s
t
ha
t
r
e
la
te
t
o
th
e
a
c
tua
l
e
l
e
c
t
r
ic
i
ty
ge
ne
r
a
t
io
n
,
dis
t
r
i
bu
t
io
n
,
s
ys
t
e
m
m
a
i
nt
e
n
a
nc
e
a
s
we
l
l
a
s
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
M
e
dium
ter
m
load
de
mand
for
e
c
as
t
of
K
ano
z
one
us
ing
ne
tw
or
k
algor
it
hm
(
Huz
aimu
L
aw
al
I
mam
)
2113
e
le
c
t
r
ic
it
y
p
r
i
c
i
ng
a
mo
ng
o
th
e
r
s
.
Ove
r
the
y
e
a
r
s
,
m
a
n
y
r
e
s
e
a
r
c
he
s
w
e
r
e
c
o
nd
uc
te
d
on
t
he
N
ig
e
r
i
a
n
p
ow
e
r
s
e
c
t
or
a
nd
i
ts
r
e
lat
e
d
c
ha
l
len
ge
s
.
Ne
ve
r
t
he
les
s
,
mos
t
of
t
he
s
e
r
e
s
e
a
r
c
he
s
c
e
n
t
r
e
on
ge
ne
r
a
l
p
r
ob
le
ms
o
f
po
we
r
g
e
n
e
r
a
ti
on
o
r
t
r
a
ns
m
is
s
i
on
or
dis
t
r
i
bu
t
io
n
o
r
c
o
mb
in
e
a
n
d
t
he
r
e
s
e
a
r
c
h
e
s
on
loa
d
d
e
ma
nd
f
oc
us
e
s
on
th
e
Ni
ge
r
ia
n
w
ide
e
le
c
t
r
ic
it
y
de
ma
nd
[
1
,
2
]
or
de
ma
nd
of
a
t
ow
n
o
r
c
i
ty
[
3
]
o
r
may
be
s
ho
r
t
t
e
r
m
f
or
e
c
a
s
t
[
2
,
4
]
.
S
e
ve
r
a
l
models
r
e
late
d
to
thi
s
wor
k
we
r
e
de
v
e
loped
s
uc
h
a
s
g
r
e
y
model
[
5]
,
s
uppor
t
ve
c
tor
r
e
gr
e
s
s
ion
[
6]
,
but
the
main
is
s
ue
s
with
the
s
uppor
t
ve
c
tor
mac
hine
a
r
e
the
c
hoice
o
f
the
ke
r
n
e
l
f
unc
ti
on
pa
r
a
mete
r
s
,
e
xtens
ive
memor
y
r
e
quir
e
ment
a
nd
dif
f
iculty
o
f
in
ter
pr
e
tation,
mul
ti
-
model
a
r
ti
f
icia
l
ne
ur
a
l
ne
twor
ks
[
7]
,
f
a
s
t
-
lea
r
ni
ng
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
two
r
k
[
8]
;
s
tabili
ty
is
the
major
dr
a
wba
c
k
of
r
e
c
ur
r
e
n
t
ne
ur
a
l
ne
twor
k.
De
e
p
lea
r
ning
ne
ur
a
l
ne
twor
ks
[
9]
,
lar
ge
a
mount
of
da
ta
r
e
quir
e
ment
a
nd
de
ter
mi
na
ti
on
o
f
s
uit
a
ble
topol
ogy
a
r
e
main
de
mer
i
ts
of
de
e
p
lea
r
ning
met
hod.
Ne
ur
o
-
f
uz
z
y
model
or
f
uz
z
y
-
ne
ur
a
l
ne
twor
k
[
10
,
11
]
uti
li
z
e
s
the
mapping
tec
hniques
of
ne
u
r
a
l
n
e
twor
k
to
obtain
the
F
uz
z
y
pa
r
a
m
e
ter
s
,
ne
v
e
r
thele
s
s
,
whe
n
the
number
of
input
is
lar
ge
,
the
nu
mber
o
f
r
ules
be
c
omes
lar
ge
whic
h
incr
e
a
s
e
s
c
omput
a
ti
ona
l
bur
de
n,
thus
in
tur
n
a
f
f
e
c
ti
ng
the
ge
ne
r
a
li
z
a
ti
on
c
a
pa
bil
it
y
of
the
model.
F
uz
z
y
log
ic
model
[
12
]
,
t
he
main
inconve
nienc
e
s
with
F
uz
z
y
l
ogic
methods
a
r
e
dif
f
iculty
in
r
ules
f
or
mation
,
membe
r
s
hip
f
unc
ti
on
s
e
lec
ti
on
a
nd
inada
ptabili
ty.
T
his
pa
pe
r
f
oc
us
e
s
on
e
s
ti
matin
g
the
medium
-
ter
m
load
de
mand
o
f
Ka
no
z
one
us
in
g
ne
ur
a
l
ne
twor
k
a
lgor
it
hms
.
Ne
ur
a
l
ne
twor
k
ha
s
ga
ined
wide
a
c
c
e
ptabili
ty
o
ve
r
the
las
t
f
e
w
de
c
a
de
s
,
e
s
pe
c
ially
in
the
f
ield
of
s
ys
tem
identif
ica
ti
on,
modell
ing
a
nd
c
ont
r
o
l
a
ppli
c
a
ti
ons
[
13]
.
I
t
pr
e
s
e
nts
a
be
tt
e
r
a
lt
e
r
n
a
ti
ve
in
a
ppr
oxim
a
ti
ng
a
c
ompl
e
x
non
li
ne
a
r
s
ys
tem
a
nd
c
a
pa
ble
to
ha
ndle
we
ll
unc
e
r
tainty
[
14,
15]
.
Ge
ne
r
a
li
z
e
d
r
e
gr
e
s
s
ion
ne
ur
a
l
ne
twor
k
(
GR
NN
)
,
r
a
dial
ba
s
is
f
u
nc
ti
on
ne
ur
a
l
ne
twor
k
(
R
B
F
NN
)
a
nd
f
e
e
d
-
f
or
wa
r
d
ne
twor
k
(
F
F
NN
)
a
r
e
c
las
s
of
ne
ur
a
l
ne
two
r
k
that
a
r
e
mos
tl
y
us
e
d
in
mapping
a
c
ompl
e
x
nonli
ne
a
r
s
ys
tem.
G
R
NN
ha
s
a
gr
e
a
t
a
dva
ntage
of
f
a
s
ter
tr
a
ini
ng
a
nd
c
onve
r
ging
to
a
global
s
olut
ion
[
16]
.
I
n
GR
NN
,
the
output
is
p
r
e
dicte
d
us
ing
we
ight
e
d
a
ve
r
a
ge
of
the
output
s
o
f
tr
a
ini
ng
d
a
ta.
R
a
dial
ba
s
is
f
unc
ti
on
ne
twor
k
s
tr
uc
tur
e
is
a
mu
lt
i
-
laye
r
f
e
e
d
-
f
or
wa
r
d
ne
twor
k
.
I
t
e
nha
nc
e
s
a
c
c
ur
a
c
y
a
nd
r
e
duc
e
s
the
tr
a
ini
ng
ti
me
c
ompl
e
xit
y
.
F
e
e
d
-
f
or
wa
r
d
ne
twor
ks
a
r
e
e
a
s
ier
to
buil
d,
quit
e
s
table
a
nd
ha
ve
unidi
r
e
c
ti
ona
l
f
low
of
inf
or
mation
.
T
he
a
va
il
a
ble
pe
r
f
or
manc
e
mea
s
ur
e
s
s
uc
h
M
APE
,
mea
n
s
qua
r
e
e
r
r
o
r
(
M
S
E
)
,
r
oot
mea
n
s
qua
r
e
e
r
r
o
r
(
R
M
S
E
)
we
r
e
us
e
d
in
e
v
a
luating
the
f
or
e
c
a
s
ti
ng
a
nd
ge
ne
r
a
li
z
a
ti
on
a
bil
it
ies
of
the
p
r
opos
e
d
models
.
T
he
pa
pe
r
is
or
ga
nize
d
a
s
f
oll
ows
:
s
e
c
ti
on
2
de
s
c
r
ibes
r
e
s
e
a
r
c
h
methodology,
s
e
c
ti
on
3
p
r
e
s
e
nts
the
s
im
ulation
r
e
s
ult
s
a
nd
s
e
c
ti
on
4
is
the
c
onc
l
us
ion
.
2.
RE
S
E
AR
CH
M
E
T
HO
D
T
his
s
e
c
ti
on
de
s
c
r
ibes
the
a
ppr
oa
c
he
s
us
e
d
to
buil
d
the
ne
u
r
a
l
ne
twor
k
models
.
S
ince
ne
u
r
a
l
ne
twor
ks
a
r
e
c
las
s
if
ied
ba
s
e
d
on
their
s
tr
uc
tur
e
(
how
the
ne
u
r
ons
a
r
e
or
ga
nize
d
in
a
s
ys
tema
ti
c
manne
r
f
r
om
in
put
laye
r
to
the
output
laye
r
)
a
s
f
e
e
d
f
or
wa
r
d
a
nd
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
,
th
is
pa
pe
r
c
ons
ider
e
d
the
two
c
l
a
s
s
es
of
the
ne
twor
k
.
T
he
typ
ica
l
methods
de
ployed
in
de
ve
lopi
ng
the
ne
ur
a
l
ne
twor
k
models
a
r
e
a
s
f
oll
ows
:
2.
1.
Gener
ali
z
e
d
r
e
gr
e
s
s
ion
n
e
u
r
al
n
e
t
wor
k
(
G
RN
N)
GR
NN
is
quit
e
c
a
pa
ble
to
de
a
l
with
nois
e
,
c
onve
r
ge
to
global
s
olut
ion
a
nd
do
not
t
r
a
ps
in
the
loca
l
mi
nim
a
.
T
he
uti
li
z
a
ti
on
of
Ga
us
s
ian
f
unc
ti
ons
by
GR
NN
ha
s
im
mens
e
ly
a
ided
in
a
c
hieving
h
igh
p
r
e
diction
a
c
c
ur
a
c
y.
T
he
main
pr
inciple
o
f
GR
NN
is
e
xpr
e
s
s
e
d
a
s
:
(
)
=
Σ
=
1
−
2
2
Σ
=
1
−
2
2
(
1)
whe
r
e
(
)
de
picts
the
e
s
ti
mation
va
lue
of
input
,
r
e
pr
e
s
e
nts
the
a
c
ti
va
ti
on
f
unc
ti
on
,
−
2
2
is
the
Ga
us
s
ian
f
unc
ti
on
a
nd
is
the
s
qua
r
e
d
E
uc
l
idea
n
dis
tanc
e
.
T
he
s
tr
uc
tur
e
of
GR
NN
is
il
lus
tr
a
ted
in
F
ig
ur
e
1.
F
r
om
the
F
ig
ur
e
1
,
it
c
a
n
be
s
e
e
n
that
th
e
GR
NN
is
c
ompos
e
d
of
f
our
laye
r
s
.
T
he
input
laye
r
whic
h
r
e
s
pons
ibl
e
f
or
f
e
e
ding
the
ne
xt
laye
r
,
the
pa
tt
e
r
n
laye
r
that
c
omput
e
s
the
E
uc
li
de
a
n
dis
tanc
e
a
nd
a
c
ti
va
ti
on
f
unc
ti
on,
the
s
umm
a
ti
on
laye
r
a
nd
the
output
l
a
ye
r
a
r
e
r
e
s
pons
ibl
e
f
or
no
r
malizing
the
ou
tput
ve
c
tor
.
T
he
tr
a
ini
ng
pr
oc
e
dur
e
of
GR
NN
is
e
nti
r
e
ly
di
f
f
e
r
e
nt
f
r
om
other
ne
u
r
a
l
ne
twor
ks
.
T
he
GR
NN
f
ini
s
he
s
the
tr
a
ini
ng
onc
e
e
a
c
h
input
-
output
ve
c
tor
pa
ir
f
r
o
m
the
tr
a
ini
ng
da
tas
e
t
is
f
e
d
int
o
the
input
laye
r
.
T
h
e
number
of
ne
ur
ons
in
the
pa
tt
e
r
n
laye
r
is
mos
tl
y
e
qua
l
to
th
e
number
of
pa
tt
e
r
n
s
in
the
t
r
a
ini
ng
da
tas
e
t
.
2.
2.
Radi
al
b
as
is
f
u
n
c
t
ion
n
e
u
r
al
n
e
t
wor
k
(
RB
F
NN
)
R
a
dial
ba
s
is
f
unc
ti
on
ne
ur
a
l
ne
twor
k
ha
s
pr
ove
n
to
be
unive
r
s
a
l
a
ppr
oxim
a
tor
that
uti
li
z
e
s
r
a
dial
ba
s
is
f
unc
ti
on
a
s
a
c
ti
va
ti
on
f
unc
ti
on.
T
he
F
ig
ur
e
2
de
picts
the
s
tr
uc
tu
r
e
of
th
e
r
a
dial
ba
s
is
f
unc
ti
o
n
ne
ur
a
l
ne
twor
k
[
17
]
.
F
r
om
the
F
ig
ur
e
2
,
i
t
c
a
n
be
s
e
e
n
that
the
ne
twor
k
c
ons
is
ts
of
input
laye
r
,
hidd
e
n
laye
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
1
8
,
No
.
4
,
Augus
t
2020
:
211
2
-
211
7
2114
a
nd
the
output
laye
r
.
T
he
hidden
laye
r
c
ontains
th
e
ne
ur
ons
a
nd
pr
oc
e
s
s
the
given
input
by
a
pplyi
ng
a
r
a
dial
ba
s
is
f
unc
ti
on
.
E
a
c
h
h
idden
unit
c
omput
e
s
it
s
outp
ut
given
by
:
,
(
)
=
(
∥
−
∥
2
)
(
2)
whe
r
e
is
the
c
e
ntr
e
o
f
the
ba
s
is
f
unc
ti
on
a
nd
2
•
de
picts
the
E
uc
li
de
a
n
dis
tanc
e
.
T
he
output
laye
r
c
a
lcula
te
s
the
we
ight
e
d
s
um
thr
ough
im
pleme
ntation
of
li
ne
a
r
a
c
ti
va
ti
on
f
unc
ti
on
a
nd
yields
the
output
given
by
the
e
xpr
e
s
s
ion
:
,
=
∑
+
1
=
1
,
(
3)
I
n
p
u
t
L
a
y
e
r
P
a
t
t
e
r
n
L
a
y
e
r
S
u
m
m
a
t
i
o
n
L
a
y
e
r
O
u
t
p
u
t
L
a
y
e
r
S
S
D
•
•
•
•
•
F
igur
e
1.
Ge
ne
r
a
li
z
e
d
r
e
gr
e
s
s
ion
ne
ur
a
l
ne
twor
k
s
t
r
uc
tur
e
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
z
2
z
1
I
z
+
1
+
11
21
1
I
1
2
I
1
y
2
y
I
y
1
−
2
,
1
I
+
11
w
1
K
w
12
w
2
K
w
1
I
w
1
,
1
I
w
+
,1
KI
w
+
1
o
K
o
.
.
.
F
igur
e
2.
R
a
dial
ba
s
is
f
unc
ti
on
ne
ur
a
l
ne
twor
k
s
tr
u
c
tur
e
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
M
e
dium
ter
m
load
de
mand
for
e
c
as
t
of
K
ano
z
one
us
ing
ne
tw
or
k
algor
it
hm
(
Huz
aimu
L
aw
al
I
mam
)
2115
2.
3.
F
e
e
d
-
f
or
war
d
n
e
u
r
al
n
e
t
wor
k
(
F
F
NN
)
Ne
ur
a
l
ne
twor
ks
a
da
pt
to
the
e
nvir
onmenta
l
c
ha
n
ge
s
.
T
he
a
da
ptabili
ty
e
nha
nc
e
s
their
pe
r
f
or
manc
e
,
e
ve
n
if
ther
e
a
r
e
la
r
ge
va
r
iations
a
nd
unc
e
r
taint
ies
[
18]
.
Ne
ur
a
l
ne
twor
ks
c
ompr
is
e
of
node
s
a
n
d
li
nks
.
T
he
node
s
r
e
c
e
ive
the
incom
ing
s
ignals
,
pr
oc
e
s
s
them
a
nd
y
ield
a
n
output
.
T
he
l
inks
indi
c
a
te
the
dir
e
c
ti
on
of
the
in
f
or
mation
f
low
whic
h
c
a
n
be
in
only
one
di
r
e
c
ti
on
or
bid
ir
e
c
ti
ona
l
[
18
]
.
T
he
c
las
s
if
ica
ti
on
of
t
he
ne
ur
a
l
ne
twor
ks
a
r
e
ba
s
e
d
on
their
a
r
c
hit
e
c
tur
e
a
s
f
e
e
d
-
f
or
wa
r
d
or
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
[
1
4
,
19
]
.
F
e
e
d
-
f
or
wa
r
d
ne
ur
a
l
ne
twor
k
a
s
s
hown
in
F
ig
ur
e
3
is
the
mos
t
c
omm
only
us
e
d
f
or
modelli
ng
a
nd
c
ont
r
ol
be
c
a
us
e
of
it
s
s
table
na
tur
e
a
nd
s
im
pli
c
it
y
[
20,
21]
.
T
his
pa
pe
r
uti
li
z
e
s
f
e
e
d
-
f
or
wa
r
d
ne
ur
a
l
ne
twor
k
f
or
the
f
or
e
c
a
s
ti
ng.
De
tails
r
e
ga
r
ding
c
hoice
of
f
e
e
d
-
f
or
wa
r
d
ne
u
r
a
l
ne
twor
k
c
ould
be
f
ound
in
[
13,
18,
22
-
24
].
I
n
p
u
t
L
a
y
e
r
H
i
d
d
e
n
L
a
y
e
r
O
u
t
p
u
t
L
a
y
e
r
F
igur
e
3.
F
e
e
df
or
wa
r
d
ne
ur
a
l
ne
twor
k
a
r
c
hit
e
c
tur
e
2.
4.
M
od
e
l
b
u
il
d
i
n
g
T
he
his
tor
ic
da
ta
s
e
t
f
r
om
KE
DC
O
wa
s
p
r
e
-
pr
oc
e
s
s
e
d
a
nd
r
a
ndoml
y
divi
de
d
int
o
tr
a
ini
ng
da
ta
s
e
t
a
nd
tes
ti
ng
da
ta
s
e
t.
E
a
c
h
of
the
models
wa
s
de
ve
l
ope
d
us
ing
tr
a
ini
ng
da
ta
c
ontaining
80
%
of
the
whole
da
ta
s
e
t
while
th
e
ge
ne
r
a
li
z
a
ti
on
c
a
pa
bil
it
ies
of
the
de
ve
loped
models
we
r
e
e
va
luate
d
us
in
g
the
tes
t
da
ta
s
e
t
whic
h
c
ontaine
d
20%
of
the
da
ta
s
e
t
.
T
he
r
e
maining
pa
r
t
of
thi
s
s
e
c
ti
on
be
low
s
how
s
how
models
we
r
e
r
e
a
li
z
e
d.
2.
4.
1.
Gener
ali
z
e
d
r
e
gr
e
s
s
ion
n
e
u
r
al
n
e
t
wor
k
m
od
e
l
T
he
s
tr
uc
tur
e
of
the
GR
NN
is
s
e
lec
ted
a
s
de
picte
d
in
F
ig
ur
e
1
.
T
he
pa
tt
e
r
n
laye
r
(
s
e
c
ond
laye
r
)
ha
s
r
a
dba
s
ne
ur
ons
a
nd
bias
e
s
.
T
he
we
ight
s
of
pa
tt
e
r
n
laye
r
a
r
e
s
e
t
to
1
.
T
he
b
ias
is
s
e
t
to
c
olum
n
ve
c
tor
of
0.
8328/s
pr
e
a
d.
T
he
s
umm
a
ti
on
laye
r
(
thi
r
d
laye
r
)
ha
s
pur
e
li
n
ne
ur
ons
.
High
va
lue
of
s
pr
e
a
d
e
nha
nc
e
s
the
ne
twor
k
ge
ne
r
a
li
z
a
ti
on
c
a
pa
bil
it
y,
mi
nim
ize
s
f
or
e
c
a
s
ti
ng
e
r
r
or
a
nd
the
r
e
s
ult
s
of
the
ne
twor
k
be
c
omes
s
moot
he
r
.
T
he
s
pr
e
a
d
is
c
hos
e
n
to
be
1
.
0.
2.
4.
2.
RB
F
NN
m
od
e
l
T
he
s
tr
uc
tur
e
of
the
R
B
F
NN
wa
s
c
hos
e
n
e
xa
c
tl
y
the
s
a
me
a
s
that
of
the
GR
NN
with
the
only
dif
f
e
r
e
nc
e
that
the
th
ir
d
laye
r
of
the
R
B
F
NN
is
a
ls
o
c
ompos
e
d
of
bias
e
s
.
S
ince
ther
e
is
no
e
s
tablis
he
d
s
ys
tema
ti
c
a
ppr
oa
c
h
of
s
e
lec
ti
ng
the
s
tr
uc
tu
r
e
.
I
t
wa
s
c
hoos
e
n
thr
ough
tr
ial
a
nd
e
r
r
o
r
method
a
nd
r
e
a
li
z
e
d
s
tr
uc
tur
e
is
s
im
il
a
r
to
that
s
hown
in
F
ig
ur
e
2
.
2.
4.
3.
F
F
NN
m
od
e
l
T
h
e
a
r
c
h
it
e
c
t
u
r
e
o
f
F
F
NN
is
s
i
mi
la
r
t
o
tha
t
il
l
us
tr
a
te
d
i
n
F
ig
ur
e
3.
C
ho
ice
o
f
a
p
pr
op
r
ia
te
ne
tw
o
r
k
p
a
r
a
met
e
r
s
a
r
e
k
e
y
f
o
r
e
f
f
e
c
t
i
ve
l
e
a
r
n
in
g
a
nd
b
e
t
te
r
pe
r
f
or
ma
nc
e
.
T
he
h
i
dde
n
l
a
ye
r
is
m
a
de
up
o
f
te
n
(
1
0)
n
e
u
r
ons
.
T
h
e
t
a
g
-
s
ig
a
nd
pu
r
e
l
in
we
r
e
us
e
d
a
s
the
t
r
a
ns
f
e
r
f
un
c
t
i
ons
f
o
r
th
e
h
id
de
n
a
n
d
o
u
tp
ut
la
ye
r
r
e
s
p
e
c
ti
ve
ly
.
3.
RE
S
UL
T
S
A
ND
AN
AL
YSI
S
T
hr
ough
s
im
ulation
the
pe
r
f
or
manc
e
c
a
pa
bil
it
ies
a
nd
a
c
c
ur
a
c
ies
of
dif
f
e
r
e
nt
models
c
ould
be
tes
ted.
T
he
one
-
mont
h
p
r
e
diction
pe
r
f
o
r
manc
e
s
of
the
m
ode
ls
dur
ing
tr
a
ini
ng
a
nd
tes
ti
ng
pha
s
e
we
r
e
il
lus
tr
a
ted
in
F
ig
ur
e
s
4
a
nd
5
r
e
s
pe
c
ti
ve
ly.
T
he
a
c
c
ur
a
c
y
of
t
he
models
we
r
e
e
va
luate
d
us
ing
c
omm
only
pe
r
f
or
manc
e
mea
s
ur
e
s
a
nd
the
r
e
s
ult
s
a
r
e
pr
e
s
e
nted
in
the
T
a
ble
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
1
8
,
No
.
4
,
Augus
t
2020
:
211
2
-
211
7
2116
F
ig
ur
e
4.
M
ode
ls
pr
e
diction
pe
r
f
or
manc
e
s
f
or
one
-
mont
h
tr
a
ini
ng
pha
s
e
F
ig
ur
e
5.
M
ode
ls
pr
e
diction
pe
r
f
or
manc
e
f
o
r
one
-
mont
h
tes
ti
ng
pha
s
e
T
a
ble
1.
One
-
mont
h
models
pe
r
f
or
manc
e
M
ode
l
T
r
a
in
in
g P
ha
s
e
T
e
s
ti
ng P
ha
s
e
M
S
E
R
M
S
E
M
A
P
E
(
%
)
M
S
E
R
M
S
E
M
A
P
E
(
%
)
F
F
N
N
0.0041
0.0642
0.0016
0.0531
0.0729
0.0017
R
B
F
N
N
0.0041
0.0642
8.5954E
-
15
531.9490
23.064
0.0404
G
R
N
N
0.0389
0.1971
0.0055
1.8499
1.3601
0.0307
S
im
il
a
r
ly,
the
F
ig
u
r
e
6
a
nd
F
ig
u
r
e
7
il
lus
tr
a
ted
the
models
pr
e
diction
pe
r
f
or
manc
e
s
f
or
the
two
mont
hs
dur
ing
the
t
r
a
ini
ng
a
nd
tes
ti
ng
pha
s
e
r
e
s
pe
c
ti
ve
ly.
T
he
a
c
c
ur
a
c
y
of
the
models
w
as
e
va
luate
d
a
nd
th
e
r
e
s
ult
s
a
r
e
il
lus
tr
a
ted
in
the
T
a
ble
2.
I
t
is
a
ppa
r
e
nt
that
du
r
ing
the
t
r
a
ini
ng
pha
s
e
de
picte
d
in
F
ig
ur
e
4,
the
p
r
e
dictions
of
the
models
we
r
e
a
ble
to
f
oll
ow
e
xa
c
tl
y
the
tr
a
jec
tor
y
of
the
obs
e
r
ve
d
da
ta
a
nd
the
a
gr
e
e
ment
t
a
ll
y
with
the
e
va
luate
d
r
e
s
ult
s
il
lus
tr
a
ted
in
the
T
a
ble
1
a
n
d
the
pr
e
dictions
a
r
e
highl
y
a
c
c
ur
a
te
[
2
5
]
ha
ving
a
c
hieve
d
the
M
APE
of
les
s
than
10
%
[
2
5
]
by
e
a
c
h
mod
e
l.
Du
r
ing
the
tes
ti
ng
pha
s
e
a
s
s
hown
in
F
ig
ur
e
.
5
,
a
ls
o
the
pr
e
dictions
of
the
models
a
r
e
quit
e
a
c
c
ur
a
te
h
a
ving
a
c
hieve
d
the
M
APE
of
les
s
than
10%
.
F
o
r
the
two
mont
hs
,
the
p
r
e
dictions
of
the
models
dur
ing
tr
a
ini
ng
pha
s
e
is
quit
e
p
r
omi
s
ing
a
s
s
hown
in
F
ig
u
r
e
6
a
nd
e
a
c
h
of
the
model
wa
s
a
ble
to
a
c
hieve
d
the
M
APE
of
les
s
than
10%
a
s
pr
e
s
e
nt
e
d
in
the
T
a
ble
2
indi
c
a
ti
ng
highl
y
a
c
c
ur
a
te
pr
e
diction.
Dur
ing
the
tes
ti
ng
pha
s
e
a
s
il
lus
tr
a
ted
in
F
ig
u
r
e
7,
the
models
de
mons
t
r
a
t
e
d
their
c
a
pa
bil
it
ies
of
tr
a
c
king
we
ll
the
pa
th
of
the
obs
e
r
v
e
d
da
ta
a
nd
the
a
c
hieve
d
M
APE
s
a
r
e
quit
e
a
tt
r
a
c
ti
ve
.
F
igur
e
6.
M
ode
ls
pr
e
diction
pe
r
f
or
manc
e
s
f
or
two
-
mont
h
tr
a
ini
ng
pha
s
e
F
igur
e
7.
M
ode
ls
pr
e
diction
pe
r
f
or
manc
e
s
f
or
two
-
mont
h
tes
ti
ng
pha
s
e
T
a
ble
2.
T
wo
-
mont
h
models
’
pe
r
f
or
manc
e
M
ode
l
T
r
a
in
in
g P
ha
s
e
T
e
s
ti
ng P
ha
s
e
M
S
E
R
M
S
E
M
A
P
E
(
%
)
M
S
E
R
M
S
E
M
A
P
E
(
%
)
F
F
N
N
0.0045
0.0668
0.0005
0.01
0.0873
0.00051
R
B
F
N
N
7.101E
-
06
0.0027
2.444E
-
05
109.11
10.4455
0.0254
G
R
N
N
0.0003
0.0185
7.152E
-
05
15.22
3.9011
0.0150
4.
CONC
L
USI
ON
T
he
pa
pe
r
h
a
s
pr
e
s
e
nted
the
ne
ur
a
l
ne
twor
k
a
lgor
it
hms
f
or
medium
ter
m
load
f
or
e
c
a
s
ti
ng
of
Ka
no
z
one
.
Dur
ing
the
t
r
a
ini
ng
pha
s
e
in
both
the
two
s
c
e
na
r
ios
the
obtaine
d
r
e
s
ult
s
de
mons
tr
a
ted
that
the
m
ode
ls
a
r
e
quit
e
e
f
f
e
c
ti
ve
a
nd
r
e
li
a
ble
in
f
or
e
c
a
s
ti
ng
the
load.
Although,
the
models
we
r
e
a
ble
to
a
c
h
ieve
d
the
M
APE
of
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
M
e
dium
ter
m
load
de
mand
for
e
c
as
t
of
K
ano
z
one
us
ing
ne
tw
or
k
algor
it
hm
(
Huz
aimu
L
aw
al
I
mam
)
2117
les
s
than
10%
dur
ing
the
tes
ti
ng
pha
s
e
s
,
the
pe
r
f
o
r
manc
e
s
of
the
F
F
NN
is
s
li
ghtl
y
be
tt
e
r
than
the
R
B
F
NN
a
nd
GR
NN
model.
T
he
pr
e
diction
pe
r
f
o
r
manc
e
s
of
the
models
a
r
e
quit
e
p
r
omi
s
ing
a
nd
r
e
li
a
ble.
T
he
models
c
ould
s
e
r
ve
a
s
the
us
e
f
ul
a
nd
e
f
f
icie
nt
tool
s
f
o
r
the
load
f
or
e
c
a
s
ti
ng
of
the
z
one
.
AC
KNOWL
E
DGE
M
E
NT
S
T
he
a
uthor
s
wis
h
to
thank
KE
DC
O,
Ka
no
Unive
r
s
i
ty
of
S
c
ienc
e
a
nd
T
e
c
hnology,
W
udil
a
nd
B
a
ye
r
o
Unive
r
s
it
y
Ka
no
f
or
their
s
uppor
t.
RE
F
E
RE
NC
E
S
[1
]
A.
A
.
Mat
i
,
et
al
.
,
“E
l
ect
r
i
ci
t
y
D
eman
d
Fo
reca
s
t
i
n
g
i
n
N
i
g
eri
a
u
s
i
n
g
T
i
me
Seri
es
Mo
d
el
,
”
Th
e
P
a
c
i
f
i
c
Jo
u
r
n
a
l
o
f
S
ci
e
n
ce
a
n
d
Tech
n
o
l
o
g
y
,
v
o
l
.
1
0
,
n
o
.
2
,
p
p
.
4
7
9
-
4
8
5
,
J
an
u
ary
2
0
0
9
.
[2
]
M.
Bu
h
ari
,
S.
S.
A
d
am
u
,
“Sh
o
r
t
-
T
erm
L
o
a
d
Fo
reca
s
t
i
n
g
U
s
i
n
g
A
r
t
i
f
i
ci
a
l
N
e
u
ral
N
et
w
o
r
k
,
”
P
r
o
ceed
i
n
g
s
o
f
t
h
e
In
t
er
n
a
t
i
o
n
a
l
M
u
l
t
i
-
co
n
f
e
r
en
ce
o
f
E
n
g
i
n
ee
r
s
a
n
d
C
o
m
p
u
t
e
r
S
ci
e
n
t
i
s
t
s
,
2
0
1
2
.
[3
]
A.
E
.
J
ame
s
,
et
a
l
.
,
“
A
rt
i
fi
c
i
al
N
eu
ral
N
et
w
o
r
k
fo
r
E
n
er
g
y
D
ema
n
d
F
o
recas
t
,
”
In
t
er
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
E
l
ec
t
r
i
c
a
l
a
n
d
E
l
ec
t
r
o
n
i
c
S
ci
e
n
ce
,
v
o
l
.
5
,
n
o
.
1
,
p
p
.
8
-
1
3
,
2
0
1
8
.
[4
]
H.
L
.
Imam
,
et
al
.
,
“Sh
o
r
t
T
erm
L
o
ad
F
o
recas
t
o
f
K
an
o
zo
n
e
u
s
i
n
g
A
rt
i
fi
c
i
al
I
n
t
el
l
i
g
en
t
T
ec
h
n
i
q
u
es
,
”
I
n
d
o
n
e
s
i
a
n
Jo
u
r
n
a
l
o
f
E
l
ect
r
i
c
a
l
E
n
g
i
n
ee
r
i
n
g
a
n
d
Co
m
p
u
t
er
S
ci
e
n
c
e
,
v
o
l
.
1
6
,
n
o
.
2
,
p
p
.
5
6
2
-
5
6
8
,
2
0
1
9
.
[5
]
X
.
Ch
e,
“A
p
p
l
i
cat
i
o
n
o
f
Imp
ro
v
ed
G
rey
Mo
d
e
l
i
n
Med
i
u
m
an
d
L
o
n
g
T
erm
L
o
ad
Fo
recas
t
i
n
g
,
”
IO
P
Co
n
f
.
S
er
i
es
:
E
a
r
t
h
a
n
d
E
n
vi
r
o
n
m
e
n
t
a
l
S
c
i
en
ce,
v
o
l
.
1
2
8
,
p
p
.
1
-
5
,
2
0
1
8
.
[6
]
A
.
Z
are
-
N
o
g
h
ab
i
,
et
al
.
,
“Med
i
u
m
-
T
erm
L
o
a
d
Fo
recas
t
i
n
g
U
s
i
n
g
Su
p
p
o
r
t
V
ect
o
r
Reg
re
s
s
i
o
n
,
Feat
u
re
Sel
e
ct
i
o
n
,
an
d
Sy
mb
i
o
t
i
c
O
r
g
an
i
s
m
Searc
h
O
p
t
i
m
i
zat
i
o
n
,
”
Co
n
f
er
e
n
ce:
2
0
1
9
IE
E
E
P
E
S
G
en
e
r
a
l
M
eet
i
n
g
,
2
0
1
9
.
[7
]
R.
M.
N
ezzar,
et
al
.
,
“Mi
d
-
L
o
n
g
-
T
erm
L
o
ad
Fo
rec
as
t
i
n
g
u
s
i
n
g
Mu
l
t
i
-
M
o
d
e
l
A
rt
i
fi
c
i
al
N
eu
ra
l
N
et
w
o
r
k
s
,
”
In
t
e
r
n
a
t
i
o
n
a
l
Jo
u
r
n
a
l
o
n
E
l
ect
r
i
c
a
l
E
n
g
i
n
eer
i
n
g
a
n
d
In
f
o
r
m
a
t
i
on
,
v
o
l
.
8
,
n
o
.
2
,
p
p
.
3
8
9
-
4
0
1
,
2
0
1
6
.
[8
]
G
.
M.
K
h
an
,
et
al
.
,
“
E
l
ect
r
i
cal
L
o
ad
Fo
reca
s
t
i
n
g
u
s
i
n
g
Fas
t
L
earn
i
n
g
Recu
rre
n
t
N
e
u
ral
N
e
t
w
o
rk
s
,
”
P
r
o
cee
d
i
n
g
s
o
f
In
t
e
r
n
a
t
i
o
n
a
l
Jo
i
n
t
C
o
n
f
er
e
n
ce
o
n
Ne
u
r
a
l
Net
w
o
r
k
s
(IJC
NN),
p
p
.
1
-
6
,
2
0
1
3
.
[9
]
D
.
L
.
Mari
n
o
,
K
.
A
maras
i
n
g
h
e
a
n
d
M.
Man
i
c,
"
Bu
i
l
d
i
n
g
en
er
g
y
l
o
a
d
fo
reca
s
t
i
n
g
u
s
i
n
g
D
ee
p
N
e
u
ral
N
e
t
w
o
r
k
s
,
"
IE
CO
N
2
0
1
6
-
4
2
n
d
A
n
n
u
a
l
Co
n
f
er
e
n
ce
o
f
t
h
e
I
E
E
E
In
d
u
s
t
r
i
a
l
E
l
ect
r
o
n
i
c
s
S
o
ci
e
t
y
,
p
p
.
7
0
4
6
-
7
0
5
1
,
2
0
1
6
.
[1
0
]
O
.
E
.
D
rag
o
mi
r,
F.
D
ra
g
o
m
i
r,
et
al
.
,
"
Med
i
u
m
t
erm
l
o
a
d
fo
recas
t
i
n
g
u
s
i
n
g
A
N
FIS
p
re
d
i
c
t
o
r,
"
1
8
t
h
M
ed
i
t
e
r
r
a
n
ea
n
Co
n
f
er
e
n
ce
o
n
C
o
n
t
r
o
l
a
n
d
A
u
t
o
m
a
t
i
o
n
,
p
p
.
5
5
1
-
5
5
6
,
2
0
1
0
.
[1
1
]
A
.
J
ar
n
d
a
l
,
"
L
o
ad
fo
re
cas
t
i
n
g
f
o
r
p
o
w
er
s
y
s
t
em
p
l
a
n
n
i
n
g
u
s
i
n
g
a
g
en
e
t
i
c
-
f
u
zzy
-
n
eu
ra
l
n
et
w
o
r
k
s
a
p
p
ro
ac
h
,
"
2
0
1
3
7
t
h
IE
E
E
G
CC
Co
n
f
e
r
en
ce
a
n
d
E
xh
i
b
i
t
i
o
n
(G
CC)
,
p
p
.
4
4
-
4
8
,
2
0
1
3
.
[1
2
]
N
.
A
mmar,
et
al
.
,
“A
n
al
y
s
i
s
L
o
ad
Fo
reca
s
t
i
n
g
o
f
Po
w
er
Sy
s
t
em
U
s
i
n
g
Fu
zzy
L
o
g
i
c
an
d
A
rt
i
fi
c
i
al
N
e
u
ral
N
e
t
w
o
r
k
,
”
Jo
u
r
n
a
l
o
f
Tel
ec
o
m
m
u
n
i
c
a
t
i
o
n
,
E
l
ec
t
r
o
n
i
c
a
n
d
Co
m
p
u
t
e
r
E
n
g
i
n
eer
i
n
g
,
v
o
l
.
9
,
n
o
.
3
,
p
p
.
1
8
1
-
1
9
2
,
2
0
1
7
.
[1
3
]
M.
S.
G
ay
a,
et
al
.
,
“
E
s
t
i
ma
t
i
o
n
o
f
T
u
r
b
i
d
i
t
y
i
n
W
a
t
er
T
reat
me
n
t
Pl
an
t
u
s
i
n
g
H
ammer
s
t
e
i
n
-
W
i
e
n
er
an
d
N
eu
ral
N
et
w
o
r
k
T
ec
h
n
i
q
u
e,
”
In
d
o
n
es
i
a
n
J
o
u
r
n
a
l
o
f
E
l
e
ct
r
i
ca
l
E
n
g
i
n
eer
i
n
g
a
n
d
Co
m
p
u
t
e
r
S
c
i
en
ce
,
v
o
l
.
5
,
n
o
.
3
,
p
p
.
6
6
6
-
6
7
2
,
2
0
1
7
.
[1
4
]
M.
S.
G
ay
a,
et
al
.
,
“Feed
-
Fo
rw
ard
N
e
u
ral
N
e
t
w
o
rk
A
p
p
ro
x
i
ma
t
i
o
n
A
p
p
l
i
e
d
t
o
A
ct
i
v
a
t
ed
S
l
u
d
g
e
Sy
s
t
em,
”
CC
IS
,
v
o
l
.
4
0
2
,
p
p
.
5
8
7
–
5
9
8
,
2
0
1
3
.
[1
5
]
N.
S.
A
.
Y
as
mi
n
,
et
al
.
,
“E
s
t
i
mat
i
o
n
o
f
p
H
an
d
ML
SS
u
s
i
n
g
N
e
u
ral
N
et
w
o
r
k
,
”
TE
LKO
M
NIKA
Te
l
eco
m
m
u
n
i
ca
t
i
o
n
,
Co
m
p
u
t
i
n
g
,
E
l
ect
r
o
n
i
c
s
a
n
d
Co
n
t
r
o
l
,
v
o
l
.
1
5
,
n
o
.
2
,
p
p
.
9
1
2
-
9
1
8
,
2
0
1
7
.
[1
6
]
A.
J
.
A
l
-
Mah
a
s
n
e
h
,
et
al
.
,
“Rev
i
e
w
o
f
A
p
p
l
i
ca
t
i
o
n
s
o
f
G
en
eral
i
zed
Reg
re
s
s
i
o
n
N
e
u
ral
N
e
t
w
o
rk
s
i
n
I
d
en
t
i
f
i
cat
i
o
n
an
d
Co
n
t
r
o
l
o
f
D
y
n
am
i
c
Sy
s
t
ems
,
”
Neu
r
a
l
a
n
d
E
v
o
l
u
t
i
o
n
a
r
y
Co
m
p
u
t
i
n
g
,
2
0
1
8
.
[1
7
]
A.
P.
E
n
g
el
b
rech
t
,
“
Co
m
p
u
t
at
i
o
n
al
i
n
t
e
l
l
i
g
e
n
ce:
A
n
i
n
t
r
o
d
u
c
t
i
o
n
,
”
Seco
n
d
E
d
i
t
i
o
n
.
C
h
i
c
h
es
t
er,
W
e
s
t
S
u
s
s
ex
E
n
g
l
a
n
d
:
Jo
h
n
W
i
l
ey
&
S
o
n
s
,
2
0
0
7
.
[1
8
]
M.
S.
G
ay
a,
“N
eu
ro
-
F
u
zzy
Mo
d
el
l
i
n
g
a
n
d
N
eu
ra
l
N
et
w
o
rk
In
t
ern
a
l
Mo
d
el
Co
n
t
r
o
l
o
f
an
A
c
t
i
v
at
e
d
Sl
u
d
g
e
Sy
s
t
e
m,
”
Ph
.
D
.
T
h
e
s
i
s
.
Sk
u
d
a
i
,
De
p
t
.
o
f
Co
n
t
r
o
l
a
n
d
Mech
a
t
ro
n
i
cs
,
U
n
i
v
er
s
i
t
i
T
e
k
n
o
l
o
g
i
Mal
a
y
s
i
a
,
2
0
1
4
.
[1
9
]
Z
.
Y
u
s
u
f,
et
al
.
,
“N
eu
ral
N
e
t
w
o
rk
Mo
d
el
D
e
v
el
o
p
me
n
t
w
i
t
h
So
ft
Co
m
p
u
t
i
n
g
T
ec
h
n
i
q
u
es
fo
r
Memb
ra
n
e
Fi
l
t
rat
i
o
n
Pro
ces
s
,
”
In
t
e
r
n
a
t
i
o
n
a
l
Jo
u
r
n
a
l
o
f
E
l
ect
r
i
c
a
l
a
n
d
Co
m
p
u
t
e
r
E
n
g
i
n
eer
i
n
g
,
v
o
l
.
8
,
n
o
.
4
,
p
p
.
2
6
1
4
-
2
6
2
3
,
2
0
1
8
.
[2
0
]
M.
S.
G
ay
a,
et
al
.
,
“A
N
FIS
Mo
d
e
l
l
i
n
g
o
f
Car
b
o
n
Re
mo
v
a
l
i
n
D
o
me
s
t
i
c
W
a
s
t
e
w
at
er
T
rea
t
men
t
Pl
an
t
,
”
A
p
p
l
i
e
d
M
ech
a
n
i
cs
a
n
d
M
a
t
er
i
a
l
,
v
o
l
.
3
7
2
,
p
p
.
5
9
7
-
6
0
1
,
2
0
1
3
[2
1
]
M.
S.
G
ay
a,
et
a
l
.
,
Co
mp
ari
s
o
n
o
f
Co
n
t
r
o
l
St
ra
t
eg
i
es
A
p
p
l
i
e
d
t
o
N
o
n
l
i
n
ear
Q
u
art
er
l
y
Car
Pas
s
i
v
e
Su
s
p
e
n
s
i
o
n
Sy
s
t
em.
In
t
e
r
n
a
t
i
o
n
a
l
R
evi
ew
o
f
A
u
t
o
m
a
t
i
c
Co
n
t
r
o
l
,
v
o
l
.
8
,
n
o
.
3
,
p
p
.
2
0
3
-
2
0
8
,
2
0
1
5
.
[2
2
]
S.
Su
mat
j
h
i
,
S.
Pan
eers
el
v
am,
“
Co
m
p
u
t
at
i
o
n
al
In
t
el
l
i
g
e
n
ce
Parad
i
g
ms
:
T
h
e
o
ry
an
d
A
p
p
l
i
cat
i
o
n
s
u
s
i
n
g
MA
T
L
A
B,
”
CR
C
P
r
e
s
s
,
L
o
n
d
o
n
,
2
0
1
0
.
[2
3
]
T
.
Mas
t
er
s
,
“Pract
i
ca
l
N
eu
ra
l
N
e
t
w
o
rk
Rec
i
p
e
s
i
n
C+
+
,
”
A
ca
d
e
m
i
c
P
r
e
s
s
,
N
e
w
Y
o
r
k
,
1
9
9
3
.
[
2
4
]
A
.
J
a
i
n
,
e
t
a
l
.
,
“
A
r
t
i
f
i
c
i
a
l
N
e
u
r
a
l
N
e
t
w
o
r
k
:
A
t
u
t
o
r
i
a
l
,
I
E
E
E
-
C
o
m
p
u
t
e
r
M
a
g
a
z
i
n
e
,
”
C
o
m
p
u
t
e
r
,
v
o
l
.
2
9
,
n
o
.
3
,
p
p
.
3
1
-
4
4
,
M
a
r
c
h
1
9
9
6
.
[2
5
]
L
aw
ren
ce
K
.
D
,
K
l
i
mb
er
g
R.
K
,
L
aw
ren
ce
S.
M
,
“
Fu
n
d
amen
t
a
l
o
f
Fo
reca
s
t
i
n
g
u
s
i
n
g
E
x
cel
,”
New
Yo
r
k:
I
n
d
u
s
t
r
i
a
l
P
r
e
s
s
In
c
.
,
2
0
0
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.