I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
A
I
)
V
ol
.
10
, N
o. 1, M
a
r
c
h
2021
, pp.
74
~
83
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
10
.i
1
.pp
74
-
83
74
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
S
t
oc
k
p
r
i
c
e
f
or
e
c
ast
o
f
m
ac
r
o
-
e
c
on
om
i
c
f
ac
t
or
u
si
n
g r
e
c
u
r
r
e
n
t
n
e
u
r
al
n
e
t
w
or
k
M
. R
e
z
a P
ah
la
w
an
, E
d
w
in
R
ik
s
ak
om
ar
a, R
ar
as
T
yas
n
u
r
it
a
, A
h
m
ad
M
u
k
la
s
on
, F
ai
z
al
M
ah
a
n
an
t
o,
R
e
t
n
o A
V
in
ar
t
i
Department of Information
Systems, Ins
titut Teknol
ogi Sepuluh
Nopember
,
Indonesia
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
J
u
l
2
3
, 20
20
R
e
vi
s
e
d
N
ov
11
, 20
20
A
c
c
e
pt
e
d
J
a
n
2
5
, 20
21
The
stock
market
is
one
of
the
investment
choices
that
always
have
t
raction
from
time
to
time.
Aside
from
being
a
means
of
corporate
funding,
in
vesting
in
the
stock
market
can
benefit
investors.
Investing
also
has
a
higher
risk
because the pattern of stoc
k
prices is
volatile, which is
caused by inter
nal and
external
factors.
One
external
factor
that
affects
stock
prices
is
the
macro
-
economic,
where
these
factors
are
events
that
occur
in
a
country
wh
ere
one
of
the
economic
sectors
affected
is
stock
prices.
I
nvestors
often
feel
co
nfused
about
the
right
time
in
decisions
making
related
to
buying
or
selling
stock.
One
way
to
look
at
how
the
prospect
of
stock
prices
is
a
stoc
k
price
forecasting
activity
.
For
this
study,
we
will
be
making
use
of
the
re
current
neur
al
network
(RNN)
to
forecast
stock
prices
for
the
next
period
s.
This
research
involves
two
variables:
the
closing
stock
price
and
the
rupiah
exchange
rate
against
the
dollar
for
the
daily
period.
We
achieve
a
MAPE
value
of
1.546%
for
RNN
model
without
the
variable
foreign
exchan
ge
rate
and
1.558%
for
the
RNN
model
that
uses
the
foreign
exchange
rate
against
the dollar.
K
e
y
w
o
r
d
s
:
E
xc
ha
nge
r
a
te
R
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
S
to
c
k f
or
e
c
a
s
t
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
R
a
r
a
s
T
y
a
s
nur
it
a
D
e
pa
r
tm
e
nt
of
I
nf
or
m
a
ti
on S
ys
te
m
s
I
ns
ti
tu
t
T
e
knol
ogi
S
e
pul
uh N
ope
m
be
r
S
ukol
il
o
S
ur
a
ba
ya
, J
a
w
a
T
im
ur
, I
ndone
s
ia
, 60111
E
m
a
il
:
r
a
r
a
s
@
is
.i
ts
.a
c
.i
d
1.
I
N
T
R
O
D
U
C
T
I
O
N
T
he
s
to
c
k
m
a
r
ke
t
i
s
one
of
th
e
in
v
e
s
tm
e
nt
c
hoi
c
e
s
th
a
t
a
lw
a
ys
ha
ve
tr
a
c
ti
on
f
r
om
ti
m
e
to
ti
m
e
.
O
ne
in
di
c
a
to
r
of
th
e
hi
gh
le
ve
l
o
f
in
te
r
e
s
t
in
in
ve
s
ti
ng
in
th
e
s
to
c
k
m
a
r
ke
t
is
on
th
e
I
ndone
s
ia
s
to
c
k
e
xc
ha
nge
(
I
D
X
)
.
T
he
s
e
in
te
r
e
s
ts
c
a
n
be
r
e
pr
e
s
e
nt
e
d
in
th
e
I
ndone
s
ia
n
c
e
nt
r
a
l
s
e
c
ur
it
ie
s
de
pos
it
or
y
da
ta
,
na
m
e
ly
434,107
ne
w
in
ve
s
to
r
s
in
2015,
894,116
in
2016,
a
nd
1,122,66
8
in
2017
[
1
]
.
T
hi
s
in
c
r
e
a
s
e
is
due
to
la
w
no.
11
o
f
2016
c
onc
e
r
ni
ng
ta
x
a
m
ne
s
ty
a
nd
c
a
pi
ta
l
m
a
r
ke
t
e
duc
a
ti
on
a
nd
s
oc
ia
li
z
a
ti
on
c
onduc
te
d
by
I
D
X
[
2]
.
I
nve
s
to
r
s
of
te
n f
e
e
l
c
onf
us
e
d a
bout
t
he
r
ig
ht
ti
m
e
a
na
ly
s
is
a
nd c
om
pa
ny s
e
le
c
ti
on w
he
n i
nve
s
ti
ng i
n t
he
s
to
c
k
m
a
r
ke
t.
T
he
s
to
c
k
m
a
r
ke
t
i
s
known
to
be
vol
a
ti
l
e
a
nd
un
s
ta
bl
e
.
A
s
to
c
k
pr
ic
e
c
a
n
in
c
r
e
a
s
e
on
e
da
y,
but
a
de
c
r
e
a
s
e
in
th
e
n
e
xt
pe
r
io
d
[
3]
.
I
n
th
e
s
to
c
k
m
a
r
ke
t,
th
e
r
e
a
r
e
s
e
ve
r
a
l
c
om
pa
ni
e
s
w
hos
e
s
to
c
k
ha
s
bl
ue
-
c
hi
p
s
ta
tu
s
. B
lu
e
-
c
hi
p
s
to
c
k
c
a
n
be
in
te
r
pr
e
te
d a
s
le
a
di
ng
s
to
c
k
s
b
e
c
a
us
e
it
c
a
n
ge
n
e
r
a
te
la
r
ge
pr
of
it
s
a
nd
r
out
in
e
ly
s
ha
r
e
th
e
s
e
pr
of
it
s
w
it
h
in
v
e
s
to
r
s
.
O
ne
e
x
a
m
pl
e
of
th
is
c
om
pa
ny
is
X
Y
Z
.
X
Y
Z
’
s
s
to
c
k
c
a
n
b
e
a
n
opt
io
n
f
or
in
ve
s
to
r
s
w
ho
w
a
nt
to
in
ve
s
t
in
bl
ue
-
c
hi
p
s
to
c
ks
,
e
s
pe
c
ia
ll
y
th
os
e
w
ho
w
a
nt
to
m
in
im
iz
e
lo
s
s
e
s
du
e
to
in
ve
s
ti
ng i
n a
s
to
c
k. T
ha
t
is
be
c
a
us
e
t
he
s
e
s
to
c
k
s
c
a
n pr
oduc
e
r
e
tu
r
ns
t
ha
t
te
nd t
o be
s
ta
bl
e
. X
Y
Z
c
a
n pr
oduc
e
a
s
ta
bl
e
r
e
tu
r
n
va
lu
e
be
c
a
us
e
of
th
e
la
r
ge
num
be
r
of
s
ubs
id
ia
r
ie
s
s
ha
de
d
by
th
e
m
.
I
f
one
of
th
e
s
e
s
ubs
id
ia
r
ie
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
oc
k
pr
ic
e
f
o
r
e
c
as
t
of
m
a
c
r
o
-
e
c
onomic
f
a
c
to
r
us
in
g
r
e
c
u
r
r
e
nt
ne
ur
al
ne
tw
or
k
(
M
. R
e
z
a P
ahl
aw
an
)
75
s
uf
f
e
r
s
a
lo
s
s
,
it
doe
s
not
s
ig
ni
f
ic
a
nt
ly
a
f
f
e
c
t
X
Y
Z
'
s
s
to
c
k
a
s
th
e
pa
r
e
nt
c
om
pa
ny
be
c
a
us
e
th
e
r
e
a
r
e
s
ti
ll
ot
he
r
s
ubs
id
ia
r
ie
s
t
ha
t
c
a
n c
ove
r
i
t.
I
nve
s
ti
ng
in
th
e
s
to
c
k
m
a
r
ke
t
ha
s
a
hi
ghe
r
r
is
k
of
s
a
vi
ng
a
t
th
e
ba
nk
e
ve
n
th
ough
th
e
oppor
tu
ni
ty
f
or
hi
ghe
r
pr
of
it
s
.
T
he
r
e
a
s
on
i
s
due
to
f
lu
c
tu
a
ti
ng
s
to
c
k
pr
ic
e
s
w
hi
c
h
i
s
c
a
us
e
d
by
in
te
r
na
l
a
nd
e
xt
e
r
na
l
f
a
c
to
r
s
[
4]
.
O
ne
e
xa
m
pl
e
of
e
xt
e
r
na
l
f
a
c
to
r
s
th
a
t
a
f
f
e
c
t
s
to
c
k
pr
ic
e
s
i
s
m
a
c
r
o
-
e
c
onomi
c
.
T
he
s
e
f
a
c
to
r
s
a
r
e
e
v
e
nt
s
th
a
t
oc
c
ur
a
s
a
w
hol
e
r
e
ga
r
di
ng
th
e
e
c
onomy
in
on
e
c
ount
r
y
th
a
t
c
a
n
a
f
f
e
c
t
s
to
c
k
pr
ic
e
s
.
S
to
c
k
m
a
r
ke
t
vol
a
ti
li
ty
de
pe
nds
not
onl
y
on
th
e
nonl
in
e
a
r
na
tu
r
e
of
da
ta
,
but
a
ls
o
o
n
th
e
c
om
m
uni
ty
’
s
s
e
nt
im
e
nt
,
e
c
onomi
c
a
nd
pol
it
ic
a
l
c
ondi
ti
ons
[
5]
.
T
hi
s
f
a
c
to
r
oc
c
ur
s
out
s
id
e
th
e
c
om
p
a
ny,
s
o
th
a
t
it
is
di
f
f
ic
ul
t
to
c
ont
r
ol
it
.
O
ne
e
xa
m
pl
e
of
m
a
c
r
o
-
e
c
onomi
c
f
a
c
to
r
s
is
th
e
e
xc
ha
nge
r
a
te
.
E
x
c
ha
nge
r
a
te
s
c
a
n
a
f
f
e
c
t
dom
e
s
ti
c
in
ve
s
tm
e
nt
s
s
uc
h
a
s
s
to
c
ks
. T
hi
s
s
it
ua
ti
on
w
il
l
c
a
us
e
a
de
c
r
e
a
s
e
in
de
m
a
nd f
or
s
ha
r
e
s
s
o
th
a
t
s
to
c
k
pr
ic
e
s
de
c
li
ne
[
6]
.
I
t
is
a
ls
o
s
uppor
te
d
th
r
ough
pr
e
vi
ous
r
e
s
e
a
r
c
h,
s
to
c
k
pr
ic
e
s
a
n
d
c
ur
r
e
nc
y
e
x
c
ha
nge
r
a
t
e
s
h
a
ve
a
po
s
it
iv
e
c
or
r
e
la
ti
on [
7]
.
A
ppr
opr
ia
te
de
c
is
io
n
m
a
ki
ng
is
ne
e
d
e
d
w
he
n
in
ve
s
ti
ng
in
s
ha
r
e
s
w
he
r
e
th
e
de
c
i
s
io
n
c
a
n
be
in
th
e
f
or
m
of
s
e
ll
in
g
s
ha
r
e
s
w
he
n
s
to
c
k
pr
ic
e
s
go
up
or
buy
w
he
n
it
is
dow
n.
O
ne
w
a
y
is
to
lo
ok
a
t
how
th
e
pr
os
pe
c
t
of
s
to
c
k
pr
ic
e
s
in
th
e
f
ut
ur
e
w
he
r
e
to
do
s
o
w
e
ne
e
d
a
s
to
c
k
pr
ic
e
f
or
e
c
a
s
ti
ng
a
c
ti
vi
ty
.
I
n
or
de
r
to
pr
oduc
e
m
e
a
ni
ngf
ul
pr
e
di
c
ti
ons
,
th
e
e
r
r
or
r
a
te
ge
ne
r
a
te
d
by
th
e
f
or
e
c
a
s
ti
ng
m
ode
l
m
u
s
t
be
m
in
im
iz
e
d.
S
e
ve
r
a
l
w
a
ys
in
w
hi
c
h
in
v
e
s
to
r
s
pr
e
di
c
t
s
to
c
k
pr
ic
e
s
s
uc
h
a
s
te
c
hni
c
a
l
a
n
a
ly
s
is
,
f
unda
m
e
nt
a
l
a
na
ly
s
is
,
a
nd
m
a
th
e
m
a
ti
c
a
l
m
ode
ls
[
8]
.
P
a
s
t
d
a
ta
in
f
or
m
a
ti
on
is
ne
e
d
e
d
to
f
or
e
c
a
s
t
s
to
c
k
pr
ic
e
s
.
T
hi
s
in
f
or
m
a
ti
on
c
a
n
be
c
ol
le
c
te
d
to
ge
t
a
n
id
e
a
of
how
th
e
s
to
c
k
pr
ic
e
m
ove
m
e
nt
tr
e
nd
s
.
F
r
om
th
is
tr
e
nd,
pr
e
di
c
ti
ons
c
a
n
be
m
a
d
e
on
how
t
he
s
to
c
k pr
ic
e
goi
ng f
or
w
a
r
d.
T
hi
s
r
e
s
e
a
r
c
h
pr
ovi
de
s
a
s
ol
ut
io
n
to
m
in
im
iz
e
r
is
k
in
th
e
f
or
m
of
a
s
to
c
k
pr
ic
e
f
o
r
e
c
a
s
ti
ng
m
ode
l
w
hi
c
h
c
a
n
m
in
im
iz
e
th
e
r
is
k
of
in
ve
s
ti
ng
in
s
to
c
ks
.
T
he
r
e
f
or
e
,
it
is
ne
c
e
s
s
a
r
y
to
c
hoos
e
th
e
r
ig
ht
f
or
e
c
a
s
ti
ng
m
e
th
od.
T
oda
y,
s
e
ve
r
a
l
m
e
th
ods
c
a
n
b
e
us
e
d
to
pr
e
di
c
t
a
da
t
a
s
e
t,
one
is
th
e
im
pl
e
m
e
nt
a
ti
on
of
m
a
c
hi
ne
le
a
r
ni
ng.
M
a
c
hi
ne
le
a
r
ni
ng
is
c
la
s
s
if
ie
d
a
s
a
n
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
m
e
th
od
a
nd
ha
s
b
e
e
n
w
id
e
ly
im
pl
e
m
e
nt
e
d
in
th
e
ta
s
ks
of
c
la
s
s
if
ic
a
ti
on,
s
pa
m
f
il
te
r
in
g,
a
nd
f
or
e
c
a
s
ti
ng
[
9]
.
O
ne
m
a
c
hi
ne
le
a
r
ni
ng
m
e
th
od
th
a
t
c
a
n
be
us
e
d
is
a
n
a
r
ti
f
ic
ia
l
ne
ur
a
l
ne
twor
k
(
A
N
N
)
,
one
of
th
e
m
os
t
a
c
c
ur
a
te
a
nd
w
id
e
ly
us
e
d
f
or
e
c
a
s
ti
ng
m
e
th
ods
[
10]
.
A
N
N
is
a
m
a
th
e
m
a
ti
c
a
l
m
ode
l
in
s
pi
r
e
d
by
bi
ol
ogi
c
a
l
ne
ur
a
l
ne
twor
ks
[
8]
.
T
he
a
dv
a
nt
a
ge
of
A
N
N
is
th
a
t
it
c
a
n
pr
ov
id
e
a
n
e
a
s
y
w
a
y
to
m
ode
l
non
-
li
ne
a
r
pa
tt
e
r
n
da
ta
[
11]
.
O
ne
de
ve
lo
pm
e
nt
of
th
e
A
N
N
m
ode
l,
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
(
R
N
N
)
,
is
a
ls
o
one
of
th
e
m
o
de
ls
th
a
t
c
a
n
be
us
e
d
to
pr
e
di
c
t
a
ti
m
e
-
s
e
r
ie
s
da
ta
pa
tt
e
r
n.
R
N
N
ha
s
a
ls
o
be
e
n
us
e
d
in
pr
e
vi
ou
s
ti
m
e
s
e
r
ie
s
da
ta
f
or
e
c
a
s
ti
ng
r
e
s
e
a
r
c
h
w
he
r
e
th
e
r
e
s
ul
ti
ng
pe
r
f
or
m
a
nc
e
c
a
n
out
pe
r
f
or
m
s
uppor
t
v
e
c
to
r
m
a
c
hi
ne
(
S
V
M
)
r
e
gr
e
s
s
io
n,
w
hi
c
h
is
1.687%
c
om
pa
r
e
d
to
1.86%
[
12]
.
I
n
a
not
he
r
r
e
s
e
a
r
c
h,
th
e
R
N
N
th
a
t
pr
oduc
e
d a
M
A
P
E
va
lu
e
of
1.5617%
ha
d
be
tt
e
r
p
e
r
f
or
m
a
nc
e
th
a
n
th
e
ba
c
kpr
opa
ga
ti
on ne
ur
a
l
ne
twor
k (
B
P
N
N
)
, w
hi
c
h only pr
oduc
e
d a
2.9561%
[
13]
.
R
N
N
is
a
s
ubt
ype
of
ne
ur
a
l
n
e
twor
ks
th
a
t
u
s
e
s
f
e
e
dba
c
k
c
on
ne
c
ti
ons
[
14]
.
T
he
b
a
s
ic
c
onc
e
pt
of
R
N
N
is
s
e
que
nt
ia
l
da
ta
pr
oc
e
s
s
in
g
[
15]
.
I
n
A
N
N
,
da
ta
a
r
e
pr
oc
e
s
s
e
d s
im
ul
ta
ne
ous
ly
to
pr
oduc
e
out
put
w
hi
c
h
is
di
f
f
e
r
e
nt
f
r
om
R
N
N
.
T
h
e
R
N
N
pr
oc
e
s
s
e
s
da
ta
s
e
que
nt
ia
ll
y
f
or
w
a
r
d
s
o
th
a
t
th
e
out
put
va
lu
e
d
e
pe
nds
on
th
e
va
lu
e
of
th
e
p
r
e
vi
ous
in
put
a
nd
ha
s
a
m
e
m
or
y
o
f
pa
s
t
in
f
o
r
m
a
ti
on.
T
hi
s
s
tu
dy
us
e
s
R
N
N
be
c
a
us
e
it
ha
s
a
dva
nt
a
ge
s
in
m
a
ki
ng
m
ode
ls
f
or
ti
m
e
s
e
r
ie
s
da
t
a
a
nd
non
-
li
ne
a
r
pr
e
di
c
ti
ve
c
a
p
a
bi
li
ti
e
s
,
f
a
s
te
r
c
onve
r
ge
n
c
e
,
a
nd
m
or
e
a
c
c
ur
a
te
m
a
ppi
ng
c
a
pa
bi
li
ti
e
s
[
16]
.
R
N
N
ha
s
good
ti
m
e
-
s
e
r
ie
s
pr
oc
e
s
s
in
g
a
bi
li
ty
be
c
a
us
e
it
c
a
n
pr
oc
e
s
s
m
ul
ti
s
t
e
p
in
put
of
ti
m
e
-
s
e
r
ie
s
a
nd
it
s
hi
dde
n
la
ye
r
i
s
c
onne
c
te
d
[
17]
.
R
N
N
us
e
s
it
s
c
onne
c
te
d
hi
dde
n
la
ye
r
to
it
s
r
e
c
ur
r
e
nt
la
ye
r
to
obt
a
in
te
m
por
a
r
y
in
f
or
m
a
ti
on
f
r
o
m
th
e
pr
e
vi
ous
s
te
p
a
s
in
put
da
ta
f
or
th
e
ne
x
t
s
te
p,
m
a
ki
ng
th
is
m
e
th
od
s
ui
ta
bl
e
f
or
ti
m
e
-
s
e
r
ie
s
da
ta
[
18]
.
A
N
N
doe
s
not
ha
ve
th
a
t
la
ye
r
,
s
o
a
ll
in
put
da
ta
be
in
g pr
oc
e
s
s
e
d
s
im
ul
ta
ne
ous
ly
.
I
n
pr
e
vi
ous
s
to
c
k
f
or
e
c
a
s
ti
ng
r
e
s
e
a
r
c
h,
s
to
c
k
pr
ic
e
f
or
e
c
a
s
ti
ng
is
done
ba
s
e
d
on
th
e
in
f
lu
e
nc
e
of
s
oc
ia
l
m
e
di
a
.
T
he
R
N
N
m
ode
l’
s
pe
r
f
or
m
a
nc
e
,
w
hi
c
h
ha
s
a
M
A
P
E
va
lu
e
of
22.31
%
,
c
a
n
out
pe
r
f
or
m
A
N
N
a
nd
s
uppor
t
ve
c
to
r
r
e
gr
e
s
s
io
n
(
S
V
R
)
,
e
a
c
h
of
w
hi
c
h
is
a
bl
e
to
pr
oduc
e
M
A
P
E
va
lu
e
s
of
78.37%
a
nd
133.9
%
[
19]
.
I
n
ot
he
r
s
tu
di
e
s
,
R
N
N
c
a
n
out
pe
r
f
or
m
A
N
N
w
he
r
e
one
of
th
e
M
A
P
E
va
lu
e
s
ge
ne
r
a
te
d
is
1.61%
f
or
R
N
N
a
nd
27.76%
f
or
A
N
N
[
16]
.
F
r
om
th
e
r
e
s
e
a
r
c
h
th
a
t
ha
s
b
e
e
n
m
e
nt
io
ne
d
pr
e
vi
ous
ly
it
pr
ove
s
th
a
t
R
N
N
ha
s
a
be
tt
e
r
pe
r
f
or
m
a
nc
e
in
s
to
c
k
pr
ic
e
f
or
e
c
a
s
ti
ng.
T
hi
s
s
tu
d
y'
s
e
xpe
c
te
d
r
e
s
ul
ts
a
r
e
to
obt
a
in
X
Y
Z
'
s
s
to
c
k
pr
ic
e
pr
e
di
c
ti
ons
ba
s
e
d
on
m
a
c
r
o
-
e
c
onomi
c
f
a
c
to
r
s
a
nd
pr
ovi
d
e
ne
w
in
s
ig
ht
in
to
th
e
e
f
f
e
c
t
of
th
e
s
e
f
a
c
to
r
s
in
f
or
e
c
a
s
ti
ng
s
to
c
k
pr
ic
e
s
be
c
a
u
s
e
f
r
om
th
e
r
e
s
e
a
r
c
h
th
a
t
ha
s
be
e
n
m
e
nt
io
ne
d
be
f
or
e
no
one
m
e
nt
io
ne
d
a
bout
th
e
R
N
N
pe
r
f
or
m
a
nc
e
on
s
to
c
k
pr
ic
e
f
or
e
c
a
s
ti
ng
ba
s
e
d
m
a
c
r
o
-
e
c
onomi
c
f
a
c
to
r
.
M
or
e
ove
r
,
it
c
a
n
pr
ovi
de
in
f
or
m
a
ti
on
f
o
r
in
ve
s
to
r
s
in
m
a
ki
ng
in
ve
s
tm
e
nt
de
c
is
io
ns
,
e
s
pe
c
ia
ll
y
w
he
n
th
e
e
xc
ha
nge
r
a
te
i
s
r
e
a
s
on
a
bl
e
or
not
.
2.
R
E
S
E
A
R
C
H
M
E
T
H
O
D
F
ig
ur
e
1
is
th
e
w
or
kf
lo
w
in
th
is
r
e
s
e
a
r
c
h.
D
a
t
a
w
a
s
c
ol
le
c
te
d
on
hi
s
to
r
ic
a
l
da
ta
of
s
to
c
k
pr
ic
e
s
a
nd
th
e
dol
la
r
e
xc
ha
nge
r
a
te
a
ga
in
s
t
th
e
r
upi
a
h.
S
to
c
k
da
ta
c
ol
le
c
ti
on
is
c
a
r
r
ie
d
out
on
th
e
I
nve
s
ti
ng
s
it
e
,
w
he
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
1, M
a
r
c
h
20
21
:
74
–
83
76
th
e
c
om
pa
ny
s
to
c
k
us
e
d
i
s
X
Y
Z
.
T
he
da
ta
u
s
e
d
is
th
e
da
il
y
p
e
r
io
d
of
w
or
ki
ng
da
ys
f
r
om
O
c
to
be
r
2014
to
S
e
pt
e
m
be
r
2019.
F
e
a
tu
r
e
e
ngi
ne
e
r
in
g
is
done
by
pe
r
f
or
m
in
g
f
e
a
tu
r
e
s
e
le
c
ti
on
a
nd
da
ta
m
e
r
gi
ng
.
F
e
a
tu
r
e
s
e
le
c
ti
o
n
is
th
e
pr
oc
e
s
s
of
s
e
le
c
ti
ng
f
e
a
tu
r
e
s
or
va
r
ia
bl
e
s
th
a
t
w
il
l
be
in
put
va
r
ia
bl
e
s
in
th
e
R
N
N
m
ode
l
f
or
m
a
ti
on. Da
ta
m
e
r
gi
ng i
s
ba
s
e
d on the
s
a
m
e
da
te
on both data
.
T
he
a
ppr
oa
c
h
us
e
d
to
ove
r
c
om
e
th
e
m
is
s
in
g
va
lu
e
is
th
e
la
s
t
obs
e
r
va
ti
on
c
a
r
r
ie
d
f
or
w
a
r
d
(
L
O
C
F
)
w
he
r
e
if
th
e
r
e
i
s
a
m
is
s
in
g
va
lu
e
da
ta
,
th
e
n
th
e
da
ta
f
il
li
ng
is
ba
s
e
d
on
pr
e
vi
ous
va
lu
e
s
[
20]
.
A
f
te
r
c
he
c
ki
ng
th
e
m
is
s
in
g
va
lu
e
,
da
ta
nor
m
a
li
z
a
ti
on
is
p
e
r
f
or
m
e
d
on
th
e
in
put
va
r
ia
bl
e
.
N
or
m
a
li
z
in
g
da
ta
c
a
n
s
pe
e
d
up
m
ode
l
tr
a
in
in
g
[
21]
.
O
ne
a
ppr
oa
c
h
th
a
t
c
a
n
b
e
us
e
d
is
th
e
m
in
-
m
a
x
m
e
th
od
w
hi
c
h
c
ha
ng
e
s
th
e
va
lu
e
to
be
twe
e
n
0 a
nd
1.
B
ui
ld
in
g
th
e
R
N
N
m
ode
l,
th
e
d
a
ta
is
di
vi
de
d i
nt
o
tr
a
in
in
g
da
ta
a
nd
te
s
ti
ng
da
ta
, w
hi
c
h
ha
ve
a
pr
opor
ti
on
of
70:
30
a
nd
th
e
s
e
pr
opor
ti
ons
a
r
e
c
om
m
onl
y
us
e
d
in
s
e
ve
r
a
l
r
e
s
e
a
r
c
h. T
he
tr
a
in
in
g
da
ta
is
us
e
d
a
s
th
e
R
N
N
m
ode
l’
s
c
ons
tr
uc
ti
on,
w
hi
le
th
e
te
s
ti
ng
da
ta
is
us
e
d
a
s
th
e
pe
r
f
or
m
a
nc
e
m
e
a
s
ur
e
m
e
nt
of
th
e
R
N
N
m
ode
l.
T
o
a
voi
d
ov
e
r
f
it
ti
ng
in
th
e
tr
a
in
in
g
m
ode
l,
th
is
r
e
s
e
a
r
c
h
a
ls
o
u
s
e
s
a
va
li
da
ti
on
s
e
t.
T
he
va
li
da
ti
on
pr
oc
e
s
s
w
il
l
he
lp
r
e
duc
e
th
e
c
ha
nc
e
of
ove
r
f
it
ti
ng
th
e
te
s
t
da
t
a
[
22]
.
S
o,
th
e
pr
opor
ti
on
of
da
ta
us
e
d
in
th
is
r
e
s
e
a
r
c
h i
s
52%
t
r
a
in
in
g s
e
t,
18%
va
li
da
ti
on s
e
t,
a
nd 30%
t
e
s
ti
ng s
e
t.
R
N
N
is
a
m
e
th
od
of
le
a
r
ni
ng
s
e
que
nt
ia
l
da
ta
pa
tt
e
r
ns
th
a
t
a
r
e
ve
r
y
pow
e
r
f
ul
.
T
he
m
e
th
od
is
c
a
p
a
bl
e
of
r
e
c
e
iv
in
g
va
r
io
us
in
f
or
m
a
ti
on
w
it
h
a
f
le
xi
bl
e
a
nd
r
e
s
il
ie
nt
to
ti
m
e
di
s
to
r
ti
on.
T
ha
t
m
a
ke
s
R
N
N
good
a
t
s
e
que
nt
ia
l
la
b
e
li
ng
w
he
r
e
th
e
in
put
s
e
qu
e
nc
e
is
tr
a
n
s
c
r
ib
e
d
i
n
a
la
b
e
l
s
tr
e
a
m
[
23]
.
I
n
bui
ld
in
g
f
or
e
c
a
s
ti
ng
m
ode
ls
us
in
g
th
e
R
N
N
m
e
th
od,
two
a
s
pe
c
ts
a
r
e
ne
e
de
d.
F
ir
s
t
is
de
f
in
in
g
in
put
a
nd
out
put
va
r
ia
bl
e
s
.
A
t
th
is
s
ta
ge
,
hi
s
to
r
ic
a
l
d
a
ta
of
s
h
a
r
e
s
a
nd
e
xc
ha
ng
e
r
a
te
s
be
c
om
e
in
put
va
r
ia
bl
e
s
in
th
e
in
put
l
a
ye
r
,
in
f
or
m
in
g
th
e
R
N
N
m
ode
l
a
s
s
how
n
in
T
a
bl
e
1. M
e
a
n
w
hi
le
,
s
to
c
k
pr
ic
e
s
f
or
th
e
ne
xt
pe
r
io
d
be
c
om
e
out
put
va
r
ia
bl
e
s
in
th
e
out
put
la
ye
r
f
or
e
c
a
s
ti
ng
m
ode
l.
T
o
de
te
r
m
in
e
how
th
e
in
f
lu
e
nc
e
of
a
m
a
c
r
o
-
e
c
onomi
c
f
a
c
to
r
in
f
or
e
c
a
s
ti
ng
s
to
c
k
pr
ic
e
s
,
a
c
om
pa
r
is
on
is
m
a
d
e
be
twe
e
n
two
ty
pe
s
of
m
o
de
ls
.
T
h
e
f
ir
s
t
m
ode
l
onl
y
u
s
e
s
va
r
ia
bl
e
s
to
c
k
pr
ic
e
s
(
uni
va
r
ia
t
e
)
.
T
he
s
e
c
ond
us
e
s
va
r
ia
bl
e
s
to
c
k
pr
ic
e
s
a
lo
n
g
w
it
h
a
m
a
c
r
o
-
e
c
onomi
c
f
a
c
to
r
(
m
ul
ti
va
r
ia
te
)
a
s
a
n i
nput
m
ode
l.
F
ig
ur
e
1
.
R
e
s
e
a
r
c
h
w
or
kf
lo
w
T
a
bl
e
1
. M
od
e
l
in
put
c
onf
ig
ur
a
ti
on
T
i
m
e
s
t
e
p
N
um
be
r
of
N
ode
i
n
I
nput
L
a
ye
r
U
ni
va
r
i
a
t
e
M
ul
t
i
va
r
i
a
t
e
1
1
2
2
2
4
…
…
…
9
9
18
10
10
20
S
e
c
ond
is
bui
ld
in
g
a
f
or
e
c
a
s
ti
ng
m
ode
l
us
in
g
E
lm
a
n
-
R
N
N
a
r
c
hi
te
c
tu
r
e
.
T
h
e
out
put
a
t
th
e
hi
dde
n
la
ye
r
i
n E
lm
a
n
-
R
N
N
c
a
n pr
ovi
de
f
e
e
dba
c
k i
n t
he
f
or
m
o
f
va
lu
e
t
o i
ts
e
lf
a
nd
ot
he
r
hi
dde
n
ne
ur
ons
t
hr
ough the
r
e
c
ur
r
e
nt
la
ye
r
[
24]
.
T
hi
s
f
e
e
dba
c
k
a
ll
ow
s
E
lm
a
n
-
R
N
N
to
s
tu
dy
pa
s
t
da
ta
.
E
a
c
h
ne
ur
on
in
th
e
E
lm
a
n
-
R
N
N
hi
dde
n
la
ye
r
ha
s
on
e
r
e
c
ur
r
e
nt
la
y
e
r
th
r
ough
th
e
li
nk
in
th
e
f
or
m
of
a
lo
a
d
va
lu
e
a
s
s
how
n
in
F
ig
ur
e
1.
R
e
c
ur
s
io
n
a
t
th
e
hi
dde
n
la
ye
r
c
a
n
a
c
t
a
s
a
m
e
m
or
y
m
e
c
ha
ni
s
m
[
25]
.
C
ha
nc
e
s
of
E
lm
a
n
-
R
N
N
ove
r
f
it
ti
ng
a
r
e
a
ls
o s
m
a
ll
e
r
i
n t
he
us
e
of
pe
r
io
ds
t
ha
t
a
r
e
not
t
oo l
a
r
ge
[
23]
.
T
he
s
e
c
ond
u
s
e
s
va
r
ia
bl
e
s
to
c
k
pr
ic
e
s
a
lo
ng
w
it
h
a
m
a
c
r
o
-
e
c
on
om
ic
f
a
c
to
r
(
m
ul
ti
va
r
ia
te
)
a
s
a
n
in
put
m
ode
l.
T
he
a
m
ount
of
in
put
s
in
th
e
in
put
la
ye
r
de
pe
nds
on
th
e
num
be
r
of
pe
r
io
ds
ba
c
kw
a
r
d
th
a
t
is
us
e
d
to
m
a
ke
pr
e
di
c
ti
ons
on f
ut
ur
e
pe
r
io
ds
(
ti
m
e
s
te
p)
. T
he
t
im
e
s
te
p a
m
ount
us
e
d de
pe
nds
on t
he
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
w
he
n doing r
e
s
e
a
r
c
h.
T
uni
ng pa
r
a
m
e
te
r
s
a
r
e
pe
r
f
or
m
e
d
w
he
r
e
t
he
va
lu
e
of
e
a
c
h pa
r
a
m
e
te
r
i
s
gi
ve
n,
a
s
s
how
n
in
T
a
bl
e
1
,
w
hi
c
h
w
il
l
b
e
a
n
e
xpe
r
im
e
nt
in
f
in
di
ng
th
e
be
s
t
m
ode
l
.
E
a
c
h
pa
r
a
m
e
te
r
ha
s
it
s
f
unc
ti
on
f
or
th
e
f
or
e
c
a
s
ti
ng
m
ode
l.
T
he
hi
dde
n
la
ye
r
node
pl
a
ys
a
r
ol
e
in
c
on
ve
r
ti
ng
va
lu
e
on
th
e
in
put
la
ye
r
to
th
e
out
pu
t
la
ye
r
, a
s
s
how
n by the
pur
pl
e
c
ir
c
le
i
n F
ig
ur
e
2.
D
r
opout c
a
n pr
e
ve
nt
f
or
e
c
a
s
t
r
e
s
ul
ts
f
r
om
be
in
g ove
r
f
it
ti
ng by r
a
ndoml
y s
e
le
c
ti
ng node
s
on a
hi
dde
n
la
ye
r
to
be
dr
oppe
d
ou
t
w
it
h
a
gi
ve
n
pr
oba
bi
li
ty
of
e
a
c
h
w
e
i
ght
upda
te
c
yc
le
[
26]
.
T
he
num
b
e
r
o
f
e
poc
h
r
e
pr
e
s
e
nt
s
th
e
num
b
e
r
of
it
e
r
a
ti
ons
of
th
e
m
ode
l
tr
a
in
e
d. T
he
a
c
ti
va
ti
on
pl
a
ys
a
r
ol
e
in
c
onve
r
ti
ng
in
put
va
lu
e
f
r
om
th
e
in
put
la
ye
r
to
s
om
e
r
a
ng
e
d
va
lu
e
,
ba
s
e
d
on
th
e
ty
p
e
of
a
c
ti
va
ti
on.
T
a
nh
a
c
ti
va
ti
on
ha
s
it
s
out
put
va
lu
e
be
twe
e
n
-
1 unti
l
1.
T
he
r
e
lu
a
nd
s
of
ts
ig
n
a
c
ti
va
ti
on
ha
s
th
e
a
dv
a
nt
a
ge
ov
e
r
th
e
ta
n
h
a
c
ti
va
ti
on
be
c
a
us
e
it
w
il
l
not
r
e
a
c
h
s
a
tu
r
a
ti
on
poi
nt
w
he
n
it
ge
ts
a
la
r
ge
in
put
va
lu
e
.
T
h
e
di
f
f
e
r
e
nc
e
be
twe
e
n
th
os
e
f
unc
ti
ons
is
r
e
lu
ha
s
a
n
out
put
va
lu
e
be
twe
e
n
0
unt
il
x
pos
it
iv
e
va
lu
e
a
n
d
s
of
ts
ig
n
h
a
s
a
n
out
p
ut
va
lu
e
f
r
om
-
1
unt
il
1.
T
he
opt
im
iz
e
r
is
u
s
e
d
to
upda
te
th
e
w
e
ig
ht
pa
r
a
m
e
te
r
in
th
e
R
N
N
m
ode
l
on
e
a
c
h
it
e
r
a
ti
on.
T
he
r
e
f
or
e
,
it
c
a
n
he
lp
m
in
i
m
iz
e
th
e
lo
s
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
oc
k
pr
ic
e
f
o
r
e
c
as
t
of
m
a
c
r
o
-
e
c
onomic
f
a
c
to
r
us
in
g
r
e
c
u
r
r
e
nt
ne
ur
al
ne
tw
or
k
(
M
. R
e
z
a P
ahl
aw
an
)
77
f
unc
ti
on.
T
he
b
a
tc
h
s
iz
e
d
e
f
in
e
s
th
e
num
be
r
of
tr
a
in
in
g
s
a
m
pl
e
s
to
w
or
k
th
r
oug
h
be
f
or
e
upda
ti
ng
th
e
m
ode
l
pa
r
a
m
e
te
r
s
[
27]
. E
a
c
h m
ode
l
pa
r
a
m
e
te
r
s
l
e
a
r
ne
d by us
in
g gr
a
di
e
nt
de
s
c
e
nt
l
e
a
r
ni
ng a
lg
or
it
hm
. T
hi
s
a
lg
or
it
hm
us
e
d f
or
t
r
a
in
in
g t
he
E
lm
a
n
-
R
N
N
m
ode
l
[
28]
.
E
xpe
r
im
e
nt
s
w
e
r
e
c
a
r
r
ie
d
out
ba
s
e
d
on
th
e
pa
r
a
m
e
te
r
s
c
e
na
r
io
in
T
a
bl
e
2
w
he
r
e
th
e
be
s
t
m
ode
l
f
o
r
e
a
c
h
ti
m
e
s
te
p
w
a
s
ta
k
e
n
to
he
lp
a
na
ly
z
e
th
e
s
e
le
c
ti
on
of
th
e
be
s
t
m
ode
l.
T
he
m
ode
l
a
na
ly
s
i
s
is
pe
r
f
or
m
e
d
to
de
te
r
m
in
e
th
e
le
ve
l
of
f
or
e
c
a
s
ti
ng
m
ode
l
pe
r
f
or
m
a
nc
e
th
a
t
ha
s
be
e
n
ge
ne
r
a
te
d
in
th
e
te
s
t
da
ta
.
T
he
ne
xt
s
te
p
is
to
c
onduc
t
a
f
in
a
l
a
na
ly
s
is
to
c
hoo
s
e
th
e
b
e
s
t
m
ode
l.
T
he
be
s
t
m
ode
l
s
e
le
c
ti
on
w
il
l
be
de
te
r
m
in
e
d
th
r
ough
th
e
s
m
a
ll
e
s
t
m
e
a
n a
b
s
ol
ut
e
pe
r
c
e
nt
a
g
e
e
r
r
or
(
M
A
P
E
)
va
lu
e
i
n e
a
c
h m
ode
l
ba
s
e
d on the
f
or
e
c
a
s
t
r
e
s
ul
t.
M
A
P
E
is
e
a
s
y
to
unde
r
s
ta
nd
be
c
a
us
e
it
r
e
pr
e
s
e
nt
s
e
r
r
or
r
a
te
by
pe
r
c
e
nt
a
ge
a
nd
of
te
n
us
e
d
in
f
in
a
nc
e
b
e
c
a
us
e
ga
in
s
a
nd
lo
s
s
e
s
a
r
e
of
te
n
m
e
a
s
ur
e
d
in
r
e
la
ti
ve
va
lu
e
s
[
29]
.
I
f
th
e
be
s
t
m
ode
l
ha
s
be
e
n
c
hos
e
n,
th
e
n
th
e
c
om
pa
r
is
on
is
m
a
de
by
te
s
ti
ng
th
e
s
to
c
k
d
a
ta
of
ot
he
r
c
om
pa
ni
e
s
w
it
h
di
f
f
e
r
e
nt
c
ha
r
a
c
te
r
is
ti
c
s
a
nd
m
a
ki
ng
c
om
pa
r
is
on
s
w
it
h t
he
r
e
s
ul
ts
of
ot
he
r
m
e
th
ods
.
F
ig
ur
e
2
. E
lm
a
n
-
R
N
N
s
tr
uc
tu
r
e
T
a
bl
e
2. P
a
r
a
m
e
te
r
s
c
e
na
r
io
P
a
r
a
m
e
t
e
r
V
a
l
ue
H
i
dde
n L
a
ye
r
N
ode
F
r
om
n t
o 3n i
nput
D
r
opout
0.1 , 0.3 , 0.5
E
poc
h
50
A
c
t
i
va
t
i
on
t
a
nh, r
e
l
u,
s
of
t
s
i
gn
B
a
t
c
h S
i
z
e
32, 64, 96
O
pt
i
m
i
z
e
r
a
da
m
, r
m
s
pr
op, s
gd
C
om
pa
r
is
ons
a
r
e
m
a
de
w
it
h
th
e
f
e
e
d
-
f
or
w
a
r
d
ne
ur
a
l
ne
twor
k
(
F
F
N
N
)
w
hi
c
h
is
one
of
A
N
N
m
e
th
od
a
nd
th
e
ba
s
is
f
or
de
ve
lo
pi
ng
th
e
E
lm
a
n
-
R
N
N
a
nd
ot
he
r
R
N
N
a
r
c
hi
te
c
tu
r
a
l
m
e
th
ods
s
uc
h
a
s
lo
ng
s
hor
t
-
te
r
m
m
e
m
or
y
(
L
S
T
M
)
a
nd
ga
te
d
r
e
c
ur
r
e
nt
uni
ts
(
G
R
U
)
.
I
n
th
e
L
S
T
M
s
tr
uc
tu
r
e
,
th
e
r
e
is
a
n
a
ddi
ti
ona
l
f
o
r
ge
t
ga
te
a
nd
out
put
ga
te
,
w
hi
le
G
R
U
ha
s
a
n
a
ddi
ti
ona
l
c
ont
r
ol
le
r
ga
te
w
he
r
e
th
e
s
tr
uc
tu
r
e
a
ll
ow
s
th
e
de
te
r
m
in
a
ti
on
of
th
e
va
lu
e
t
ha
t
c
a
n be
p
a
s
s
e
d on
[
30]
.
3.
R
E
S
U
L
T
S
AND
D
I
S
C
U
S
S
I
O
N
T
he
e
xpe
r
im
e
nt
s
w
e
r
e
pe
r
f
or
m
e
d
on
a
n
I
nt
e
l
C
or
e
i5
7200U,
N
vi
di
a
G
T
930M
X
2G
B
gr
a
phi
c
c
a
r
d,
a
nd R
A
M
w
it
h 8G
B
w
hi
c
h
is
r
unni
ng on
W
in
dow
s
10 Ve
r
s
io
n
1909. T
he
s
of
twa
r
e
t
ool
us
e
d i
s
G
oogl
e
C
ol
a
b
a
nd P
yt
hon f
or
t
he
pr
ogr
a
m
m
in
g l
a
ngua
ge
.
3.1.
D
at
a
c
ol
le
c
t
io
n
T
he
da
ta
c
ol
le
c
ti
on
r
e
s
ul
t
s
a
r
e
X
Y
Z
'
s
s
to
c
k
pr
ic
e
da
ta
a
nd
hi
s
t
or
ic
a
l
da
ta
on
f
or
e
ig
n
e
xc
ha
nge
r
a
te
s
f
r
om
O
c
to
be
r
2014
to
S
e
pt
e
m
be
r
2019
on
w
or
ki
ng
da
y
s
.
T
a
bl
e
3
r
e
pr
e
s
e
nt
s
f
or
e
ig
n
e
xc
ha
ng
e
r
a
te
s
da
ta
f
or
th
e
la
s
t
f
iv
e
pe
r
io
ds
.
‘
P
r
ic
e
’
f
e
a
tu
r
e
c
ont
a
in
s
in
f
or
m
a
ti
on
a
bout
c
lo
s
in
g
pr
ic
e
on
e
a
c
h
d
a
y,
‘
O
pe
n’
c
ont
a
in
s
in
f
or
m
a
ti
on a
bout
ope
ni
ng pr
ic
e
on
e
a
c
h da
y, ‘
H
ig
h’
a
nd ‘
L
ow
’
c
ont
a
in
s
i
nf
or
m
a
ti
on a
bout
hi
ghe
s
t
a
nd l
ow
e
r
pr
ic
e
on
e
a
c
h
da
y,
‘
V
ol
.’
is
th
e
num
be
r
o
f
s
to
c
k
th
a
t
c
ha
nge
d
on
e
a
c
h
da
y,
a
nd
‘
C
ha
nge
%
’
is
th
e
di
f
f
e
r
e
nc
e
be
twe
e
n
th
e
c
ur
r
e
nt
pr
ic
e
a
nd
pr
e
vi
ous
d
a
y’
s
pr
ic
e
.
T
he
va
r
ia
b
le
us
e
d
in
th
is
r
e
s
e
a
r
c
h
is
'
P
r
ic
e
'
on
th
e
s
to
c
k
pr
ic
e
da
ta
a
nd t
he
dol
la
r
e
xc
ha
nge
r
a
t
e
a
ga
in
s
t
th
e
r
upi
a
h.
F
or
th
e
s
c
e
na
r
io
o
f
te
s
ti
ng
th
e
m
ode
l'
s
pe
r
f
or
m
a
nc
e
a
ga
in
s
t
th
e
s
to
c
k
da
ta
of
ot
he
r
c
om
pa
ni
e
s
th
a
t
ha
ve
di
f
f
e
r
e
nt
c
ha
r
a
c
te
r
is
ti
c
s
,
da
ta
is
ta
ke
n
f
r
om
th
e
s
to
c
k
da
ta
of
A
B
C
a
nd
D
E
F
in
th
e
s
a
m
e
pe
r
io
d.
A
B
C
'
s
s
to
c
k
da
ta
ha
s
a
n
in
c
r
e
a
s
e
in
th
e
la
s
t
th
r
e
e
ye
a
r
s
,
w
hi
le
D
E
F
'
s
s
to
c
k
da
ta
ha
s
de
c
li
ne
d
in
th
e
la
s
t
th
r
e
e
ye
a
r
s
F
ig
ur
e
3
s
how
s
th
a
t
X
Y
Z
,
A
B
C
, a
nd
D
E
F
s
to
c
ks
ha
ve
s
ig
ni
f
ic
a
nt
di
f
f
e
r
e
nc
e
s
in
pa
tt
e
r
ns
.
X
Y
Z
s
to
c
ks
te
nd
to
be
s
ta
bl
e
a
t
R
p
6,000
to
8,000,
w
hi
le
A
B
C
'
s
s
to
c
k
s
a
r
e
s
t
a
bl
e
in
th
e
pr
ic
e
r
a
nge
of
a
r
ound
hundr
e
ds
w
hi
le
D
E
F
'
s
s
to
c
k pr
ic
e
s
a
r
e
ve
r
y vola
ti
le
a
nd t
e
nd t
o de
c
li
ne
i
n t
he
l
a
s
t
th
r
e
e
ye
a
r
s
.
T
a
bl
e
3
. I
D
R
t
o U
S
dol
la
r
da
ta
P
r
i
c
e
O
pe
n
H
i
gh
L
ow
C
ha
nge
%
14195
14155
14195
14155
0.21
14165
14185
14200
14165
-
0.09
14177.5
14147.5
14177.5
14145.5
0.19
14150
14120
14155
14120
0.27
14112.5
14077.5
14112.5
14077.5
0.20
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
1, M
a
r
c
h
20
21
:
74
–
83
78
F
ig
ur
e
3
. S
to
c
k pr
ic
e
da
ta
pa
tt
e
r
n
3.2.
M
od
e
l
a
n
al
ys
is
I
n
th
e
uni
va
r
ia
te
m
ode
l
f
or
m
a
ti
on
e
xpe
r
im
e
nt
,
th
e
b
e
s
t
m
od
e
l
pr
oduc
e
s
M
A
P
E
of
1.546%
a
s
in
T
a
bl
e
4
s
o
th
a
t
th
e
be
s
t
m
ode
l
i
s
obt
a
in
e
d
w
it
h
th
e
pa
r
a
m
e
t
e
r
s
of
th
e
num
be
r
of
hi
dde
n
la
ye
r
node
s
30,
th
e
a
c
ti
va
ti
on
f
unc
ti
on
'
s
of
ts
ig
n'
,
th
e
num
be
r
of
ba
tc
h
s
iz
e
s
64,
th
e
ty
pe
of
opt
im
iz
e
r
‘
r
m
s
pr
op’
,
a
nd
th
e
dr
opou
t
r
a
te
of
0.1.
I
n
th
e
m
ul
ti
va
r
ia
te
m
ode
l
f
or
m
a
ti
on
e
xpe
r
i
m
e
nt
,
th
e
be
s
t
m
ode
l
pr
oduc
e
s
M
A
P
E
of
1.558%
a
s
in
T
a
bl
e
5
s
o
th
a
t
th
e
be
s
t
m
ode
l
i
s
obt
a
in
e
d
w
it
h
th
e
pa
r
a
m
e
t
e
r
s
of
th
e
num
be
r
of
hi
dde
n
la
ye
r
node
s
52,
th
e
a
c
ti
va
ti
on f
unc
ti
on ‘
ta
nh’
,
t
he
numbe
r
of
ba
tc
h s
iz
e
s
32, the
t
yp
e
of
opt
im
iz
e
r
‘
r
m
s
pr
op
’
, a
nd t
he
d
r
opout
r
a
te
of
0.1.
I
t
c
a
n
be
s
e
e
n
th
a
t
th
e
r
e
is
a
te
nde
nc
y
of
s
e
ve
r
a
l
pa
r
a
m
e
te
r
s
in
th
e
E
lm
a
n
-
R
N
N
m
ode
l
f
r
om
e
a
c
h
ti
m
e
s
te
p.
F
ir
s
t,
8
out
of
10
uni
va
r
ia
te
m
ode
ls
a
nd
7
out
of
1
0
m
ul
ti
va
r
ia
te
m
ode
ls
h
a
ve
‘
r
m
s
pr
op’
a
s
th
e
opt
im
iz
e
r
’
s
pa
r
a
m
e
te
r
.
S
e
c
ond,
a
ll
of
th
e
be
s
t
m
od
e
l
f
r
om
e
a
c
h
m
ode
l
ha
s
0.1
a
s
th
e
dr
opout
r
a
te
.
F
r
om
th
a
t
te
nde
nc
y of
s
e
ve
r
a
l
pa
r
a
m
e
te
r
s
, i
t
in
di
c
a
te
s
t
ha
t
th
e
p
a
r
a
m
e
te
r
i
s
s
ui
ta
bl
e
f
or
t
hi
s
s
tu
dy c
a
s
e
i
n t
hi
s
r
e
s
e
a
r
c
h.
T
a
bl
e
4
. U
ni
va
r
ia
te
be
s
t
m
ode
l
S
t
e
p
P
a
r
a
m
e
t
e
r
M
A
P
E
(%)
M
S
E
V
a
l
i
d
R
unt
i
m
e
(
s
)
1
[
1, '
r
e
l
u'
, 96, '
a
da
m
'
, 0.1]
2.221
0.0014
3.84
2
[
3, '
r
e
l
u'
, 64, '
s
gd'
, 0.1]
2.515
0.0121
3.807
3
[
7, '
t
a
nh'
, 32, '
r
m
s
pr
op
'
, 0.1]
1.891
0.0042
2.575
4
[
8, '
t
a
nh'
, 32, '
r
m
s
pr
op
'
, 0.1]
1.92
0.0036
5.07
5
[
14, '
t
a
nh'
, 96, '
r
m
s
pr
op
'
, 0.1]
1.732
0.0020
7.294
6
[
14,
'
t
a
nh'
, 32, '
r
m
s
pr
op
'
, 0.1]
1.704
0.0028
5.543
7
[
21, '
r
e
l
u'
, 64, '
r
m
s
pr
op'
, 0.1]
1.622
0.0011
18.14
8
[
21, '
s
of
t
s
i
gn'
, 96, '
r
m
s
pr
op'
, 0.1]
1.592
0.0022
6.408
9
[
23, '
s
of
t
s
i
gn'
, 32, '
r
m
s
pr
op'
, 0.1]
1.552
0.0027
5.994
10
[
30, '
s
of
t
s
i
gn'
, 64, '
r
m
s
pr
op'
, 0.1]
1.546
0.0025
3.974
T
a
bl
e
5
. M
ul
ti
va
r
ia
te
be
s
t
m
ode
l
S
t
e
p
P
a
r
a
m
e
t
e
r
M
A
P
E
(%)
M
S
E
V
a
l
i
d
R
unt
i
m
e
(
s
)
1
[
5, '
t
a
nh'
, 96, '
s
gd'
, 0.1]
2.664
0.0148
4,927
2
[
4, '
r
e
l
u'
, 32, '
a
da
m
'
, 0.1]
2.42
0.0184
2.94
3
[
18, '
s
of
t
s
i
gn'
, 32, '
r
m
s
pr
op'
,
0.1]
2.341
0.0011
2,444
4
[
8, '
s
of
t
s
i
gn'
, 32, '
a
da
m
'
, 0.1]
1.933
0.0014
2.002
5
[
12, '
t
a
nh'
, 32, '
r
m
s
pr
op
'
, 0.1]
1.688
0.0004
9,564
6
[
16, '
s
of
t
s
i
gn'
, 32, '
r
m
s
pr
op'
, 0.1]
1.559
0.0001
4.237
7
[
19, '
t
a
nh'
, 32, '
r
m
s
pr
op
'
, 0.1]
1.61
0.0017
6.784
8
[
22, '
s
of
t
s
i
gn'
, 64, '
r
m
s
pr
op'
, 0.1]
1.679
0.0002
2.963
9
[
52, '
t
a
nh'
, 32, '
r
m
s
pr
op
'
, 0.1]
1.558
0.0023
4.020
10
[
33, '
r
e
l
u'
, 64, '
r
m
s
pr
op'
, 0.1]
1.773
0.0019
2.729
3.3.
C
om
p
ar
is
on
o
f
u
n
iv
ar
ia
t
e
-
m
u
lt
iv
ar
ia
t
e
m
od
e
l
P
e
r
f
or
m
a
nc
e
c
om
pa
r
is
ons
w
e
r
e
m
a
de
f
r
om
th
e
uni
va
r
ia
te
a
nd
m
ul
ti
va
r
ia
te
m
ode
ls
.
F
ig
ur
e
4
s
how
s
th
a
t
th
e
be
s
t
uni
va
r
ia
te
m
ode
l
pe
r
f
or
m
a
nc
e
in
ti
m
e
s
te
p
10
w
it
h
a
M
A
P
E
va
lu
e
of
1.546%
a
s
s
how
n
in
F
ig
ur
e
4
(
a
)
a
nd
1.558
%
f
o
r
th
e
be
s
t
m
ul
ti
va
r
ia
te
m
ode
l
in
ti
m
e
s
te
p
9
a
s
s
how
n
in
F
ig
ur
e
4
(
b)
.
F
ig
ur
e
5
c
om
pa
r
e
s
a
c
tu
a
l
a
nd
f
or
e
c
a
s
t
da
ta
us
in
g
th
e
be
s
t
uni
va
r
ia
te
m
ode
l
a
s
s
how
n
in
F
ig
ur
e
5
(
a
)
a
nd
m
ul
ti
va
r
ia
te
m
ode
l
a
s
s
how
n
in
F
ig
ur
e
5
(
b)
f
r
om
A
pr
il
2018
to
S
e
pt
e
m
be
r
2019.
T
he
uni
va
r
ia
te
m
od
e
l
ha
s
be
tt
e
r
pe
r
f
or
m
a
nc
e
th
a
n
th
e
m
ul
ti
va
r
ia
te
m
ode
l,
w
he
r
e
th
e
di
f
f
e
r
e
nc
e
is
0.012%
.
F
ig
ur
e
4
s
ho
w
is
th
e
uni
va
r
ia
te
m
ode
l
ha
s
a
s
ta
nda
r
d
de
vi
a
ti
on
of
0.303
a
nd
th
e
m
ul
ti
va
r
ia
te
m
ode
l
ha
s
a
s
ta
nda
r
d
de
vi
a
ti
on
of
0.384
w
hi
c
h
in
di
c
a
te
s
t
ha
t
th
e
m
ul
ti
va
r
ia
te
m
ode
l
ha
s
m
or
e
di
ve
r
s
e
da
ta
di
s
tr
ib
ut
io
n.
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
O
c
t
01, 2014
N
ov 14, 2014
D
e
c
30, 2014
F
e
b 12, 2015
M
a
r
30, 2015
M
a
y 13, 2015
J
un 26, 2015
A
ug 11, 2015
S
e
p 24, 2015
N
ov 09, 2015
D
e
c
23, 2015
F
e
b 05, 2016
M
a
r
22, 2016
M
a
y 05, 2016
J
un 20, 2016
A
ug 03, 2016
S
e
p 16, 2016
N
ov 01, 2016
D
e
c
15, 2016
J
a
n 30, 2017
M
a
r
15, 2017
A
pr
28, 2017
J
u
n
1
3
,
2
0
1
7
J
ul
27, 2017
S
e
p 11, 2017
O
c
t
25, 2017
D
e
c
08, 2017
J
a
n 23, 2018
M
a
r
08, 2018
A
pr
23, 2018
J
un 06, 2018
J
u
l
2
0
,
2
0
1
8
S
e
p 04, 2018
O
c
t
18, 2018
D
e
c
03, 2018
J
a
n 16, 2019
M
a
r
01, 2019
A
pr
16, 2019
M
a
y 30, 2019
J
ul
15, 2019
A
ug 28, 2019
P
r
i
c
e
D
a
t
e
X
Y
Z
A
B
C
D
E
F
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
oc
k
pr
ic
e
f
o
r
e
c
as
t
of
m
a
c
r
o
-
e
c
onomic
f
a
c
to
r
us
in
g
r
e
c
u
r
r
e
nt
ne
ur
al
ne
tw
or
k
(
M
. R
e
z
a P
ahl
aw
an
)
79
T
he
s
e
r
e
s
ul
t
s
in
di
c
a
te
th
a
t
th
e
a
ddi
ti
on
of
a
m
a
c
r
o
-
e
c
onomi
c
f
a
c
to
r
in
th
e
f
or
e
c
a
s
t
m
ode
l
doe
s
not
ha
ve
a
s
ig
ni
f
ic
a
nt
im
pa
c
t,
a
nd
in
s
te
a
d,
th
e
r
e
s
ul
ts
a
r
e
lo
w
e
r
m
o
de
l
pe
r
f
or
m
a
nc
e
,
a
lt
hough
th
e
s
to
c
k
pr
ic
e
a
nd
th
e
e
xc
ha
nge
r
a
te
ha
ve
a
pos
it
iv
e
c
or
r
e
la
ti
on. Als
o, i
t
c
a
n be
a
n
a
ddi
ti
ona
l
in
s
ig
ht
f
or
i
nve
s
to
r
s
i
f
t
he
y w
a
nt
t
o
c
ons
id
e
r
t
he
e
xc
ha
ng
e
r
a
te
f
a
c
to
r
i
n s
to
c
k i
nve
s
ti
ng.
(
a
)
(
b)
F
ig
ur
e
4
. M
A
P
E
c
om
pa
r
is
on of
e
a
c
h t
im
e
s
te
p, (
a
)
U
ni
va
r
ia
te
(
b)
M
ul
ti
va
r
ia
te
(
a
)
(
b)
F
ig
ur
e
5
. C
om
pa
r
is
on of
a
c
tu
a
l
a
nd f
or
e
c
a
s
t
da
ta
, (
a
)
U
ni
va
r
ia
te
(
b)
M
ul
ti
va
r
ia
t
e
3.4.
M
od
e
l
r
ob
u
s
t
n
e
s
s
t
e
s
t
T
e
s
ti
ng
th
e
be
s
t
m
ode
l
th
a
t
ha
s
be
e
n
f
or
m
e
d
is
done
by
te
s
ti
ng
by
f
or
e
c
a
s
ti
ng
th
e
s
ha
r
e
s
of
ot
he
r
c
om
pa
ni
e
s
a
nd f
ut
ur
e
da
ta
.
T
he
m
e
a
s
ur
e
m
e
nt
of
m
ode
l
pe
r
f
or
m
a
nc
e
a
ls
o u
s
e
s
M
A
P
E
.
3.4.1.
F
u
t
u
r
e
d
at
a
F
ig
ur
e
6
is
a
c
om
pa
r
is
on
c
ha
r
t
of
f
or
e
c
a
s
ti
ng
da
ta
pr
oduc
e
d
by
th
e
m
od
e
l
a
nd
a
c
tu
a
l
d
a
ta
f
r
om
O
c
to
be
r
2019
to
M
a
y
2020
u
s
in
g
be
s
t
uni
va
r
ia
te
a
s
s
how
n
in
F
ig
ur
e
6
(
a
)
a
nd
m
ul
ti
va
r
ia
te
m
ode
l
a
s
s
how
n
in
F
ig
ur
e
6
(b
)
.
T
he
m
ode
l
c
a
nnot
f
ol
lo
w
th
e
pa
tt
e
r
n
of
da
ta
in
th
e
A
pr
il
to
M
a
y
2020
pe
r
io
d.
P
e
r
ha
ps
th
e
r
e
a
s
on
is
th
a
t
th
e
s
ig
ni
f
ic
a
nt
r
e
duc
ti
on
w
he
r
e
th
e
m
od
e
l
c
a
nnot
f
ol
lo
w
t
he
da
ta
pa
tt
e
r
n
pr
ope
r
ly
.
T
h
e
ongoing
C
ovi
d
-
19
pa
nde
m
ic
is
one
of
th
e
c
a
us
e
s
of
th
is
.
T
he
da
ta
us
e
d
in
th
e
s
c
e
na
r
io
of
f
ut
ur
e
da
ta
te
s
ti
ng
is
f
r
om
O
c
to
be
r
2019
to
M
a
y
2020.
T
he
uni
va
r
ia
te
m
od
e
l
ha
s
a
s
ig
ni
f
ic
a
nt
de
c
r
e
a
s
e
in
pe
r
f
or
m
a
nc
e
w
he
r
e
th
e
M
A
P
E
pr
oduc
e
d
is
10.25%
(
S
e
pt
2014
-
M
a
y
2020)
f
r
om
1.546
%
in
a
pr
e
vi
ous
pe
r
io
d
(
S
e
pt
2014
-
O
c
t
2019
)
.
S
im
il
a
r
to
t
he
uni
va
r
ia
te
m
ode
l,
t
he
m
ul
ti
va
r
ia
te
m
ode
l'
s
pe
r
f
or
m
a
nc
e
h
a
s
de
c
r
e
a
s
e
d s
ig
ni
f
ic
a
nt
ly
, w
he
r
e
t
he
r
e
s
ul
ti
ng
M
A
P
E
i
s
11.43%
f
r
om
1.558%
.
F
ig
ur
e
6
. C
om
pa
r
is
on of
a
c
tu
a
l
a
nd f
or
e
c
a
s
t
da
ta
, (
a
)
U
ni
va
r
ia
te
, (
b)
M
ul
ti
va
r
ia
te
on f
ut
ur
e
da
ta
2.221
2.515
1.891
1.92
1.732
1.704
1.622
1.592
1.552
1
.5
4
6
0
0.5
1
1.5
2
2.5
3
1
2
3
4
5
6
7
8
9
10
M
A
P
E
(%)
T
IM
E
S
T
E
P
2.664
2.42
2.341
1.933
1.688
1.559
1.61
1.679
1
.5
5
8
1.773
0
0.5
1
1.5
2
2.5
3
1
2
3
4
5
6
7
8
9
10
M
A
P
E
(%)
T
IM
E
S
T
E
P
6000
6500
7000
7500
8000
8500
9000
A
pr
02, 2018
A
pr
17, 2018
M
a
y 02, 2018
M
a
y 17, 2018
J
un 01, 2018
J
un 18, 2018
J
ul
03, 2018
J
u
l
1
8
,
2
0
1
8
A
ug 02, 2018
A
ug 17, 2018
S
e
p 03, 2018
S
e
p 18, 2018
O
c
t
03, 2018
O
c
t
18, 2018
N
ov 02, 2018
N
ov 19, 2018
D
e
c
04, 2018
D
e
c
19, 2018
J
a
n 03, 2019
J
a
n 18, 2019
F
e
b 04, 2019
F
e
b 19, 2019
M
a
r
06, 2019
M
a
r
21, 2019
A
pr
05, 2019
A
pr
22, 2019
M
a
y 07, 2019
M
a
y 22, 2019
J
un 06, 2019
J
u
n
2
1
,
2
0
1
9
J
ul
08, 2019
J
ul
23, 2019
A
ug 07, 2019
A
ug 22, 2019
S
e
p 06, 2019
S
e
p 23, 2019
P
r
i
c
e
D
a
t
e
A
c
t
ua
l
F
or
e
c
a
s
t
6000
6500
7000
7500
8000
8500
A
pr
02, 2018
A
pr
17, 2018
M
a
y 02, 2018
M
a
y 17, 2018
J
un 01, 2018
J
un 18, 2018
J
ul
03, 2018
J
u
l
1
8
,
2
0
1
8
A
ug 02, 2018
A
ug 17, 2018
S
e
p 03, 2018
S
e
p 18, 2018
O
c
t
03, 2018
O
c
t
18, 2018
N
ov 02, 2018
N
ov 19, 2018
D
e
c
04, 2018
D
e
c
19, 2018
J
a
n 03, 2019
J
a
n 18, 2019
F
e
b 04, 2019
F
e
b 19, 2019
M
a
r
06, 2019
M
a
r
21, 2019
A
pr
05, 2019
A
pr
22, 2019
M
a
y 07, 2019
M
a
y 22, 2019
J
un 06, 2019
J
un 21, 2019
J
ul
08, 2019
J
ul
23, 2019
A
ug 07, 2019
A
ug 22, 2019
S
e
p 06, 2019
S
e
p 23, 2019
P
r
i
c
e
D
a
t
e
A
c
t
ua
l
F
or
e
c
a
s
t
(
a
)
(
b)
3000
3500
4000
4500
5000
5500
6000
6500
7000
7500
O
c
t
03, 2019
O
c
t
10, 2019
O
c
t
17, 2019
O
c
t
24, 2019
O
c
t
31, 2019
N
ov 07, 2019
N
ov 14, 2019
N
ov 21, 2019
N
ov 28, 2019
D
e
c
05, 2019
D
e
c
12, 2019
D
e
c
19, 2019
D
e
c
26, 2019
J
a
n 02, 2020
J
a
n 09, 2020
J
a
n 16, 2020
J
a
n 23, 2020
J
a
n 30, 2020
F
e
b 06, 2020
F
e
b 13, 2020
F
e
b 20, 2020
F
e
b 27, 2020
M
a
r
05, 2020
M
a
r
12, 2020
M
a
r
19, 2020
M
a
r
27, 2020
A
pr
03, 2020
A
pr
13, 2020
A
pr
20, 2020
A
pr
27, 2020
M
a
y 05, 2020
M
a
y 13, 2020
M
a
y
2
0
,
2
0
2
0
P
r
i
c
e
D
a
t
e
A
c
t
ua
l
F
or
e
c
a
s
t
3000
3500
4000
4500
5000
5500
6000
6500
7000
7500
O
c
t
0
3
,
2
0
1
9
O
c
t
10, 2019
O
c
t
17, 2019
O
c
t
24, 2019
O
c
t
31, 2019
N
ov 07, 2019
N
ov 14, 2019
N
o
v
2
1
,
2
0
1
9
N
ov 28, 2019
D
e
c
05, 2019
D
e
c
12, 2019
D
e
c
19, 2019
D
e
c
26, 2019
J
a
n 02, 2020
J
a
n 09, 2020
J
a
n 16, 2020
J
a
n 23, 2020
J
a
n 30, 2020
F
e
b 06, 2020
F
e
b 13, 2020
F
e
b 20, 2020
F
e
b 27, 2020
M
a
r
05, 2020
M
a
r
12, 2020
M
a
r
19, 2020
M
a
r
27, 2020
A
pr
03, 2020
A
pr
13, 2020
A
pr
20, 2020
A
pr
27, 2020
M
a
y 05, 2020
M
a
y 13, 2020
M
a
y 20, 2020
P
r
i
c
e
D
a
t
e
A
c
t
ua
l
F
or
e
c
a
s
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
1, M
a
r
c
h
20
21
:
74
–
83
80
3.4.2.
C
om
p
an
y s
t
oc
k
c
om
p
ar
i
s
on
T
a
bl
e
6
s
how
i
s
th
e
be
s
t
uni
va
r
ia
te
m
ode
l
is
u
s
e
d
to
te
s
t
t
he
pe
r
f
or
m
a
nc
e
of
ot
he
r
c
om
pa
ni
e
s
.
F
ig
ur
e
7
c
om
pa
r
e
s
th
e
a
c
tu
a
l
da
ta
a
nd
f
or
e
c
a
s
t
r
e
s
ul
ts
of
A
B
C
a
s
s
how
n
in
F
ig
ur
e
7
(
a
)
a
nd
D
E
F
a
s
s
how
n
in
F
ig
ur
e
7
(
b)
s
to
c
k
pr
ic
e
s
us
in
g
th
e
be
s
t
uni
va
r
ia
te
m
ode
l.
T
he
be
s
t
uni
va
r
ia
te
m
ode
l
pe
r
f
or
m
a
nc
e
w
he
n
u
s
e
d
to
f
or
e
c
a
s
t
A
B
C
'
s
s
to
c
k
pr
ic
e
is
3.03%
,
w
hi
le
th
e
m
ode
l'
s
pe
r
f
or
m
a
nc
e
on
D
E
F
'
s
s
ha
r
e
s
ha
s
a
p
e
r
f
or
m
a
nc
e
of
6.02%
.
T
a
bl
e
7
s
how
is
th
e
be
s
t
m
ul
ti
va
r
ia
te
m
ode
l
pe
r
f
or
m
a
nc
e
te
s
ti
ng
is
on
ot
he
r
c
om
pa
ni
e
s
.
F
ig
ur
e
8
c
om
pa
r
e
s
t
he
a
c
tu
a
l
da
ta
a
nd
th
e
f
or
e
c
a
s
ti
ng
of
A
B
C
a
s
s
how
n
in
F
ig
ur
e
8
(
a
)
a
nd
D
E
F
a
s
s
how
n
in
F
ig
ur
e
8
(
b)
s
to
c
k pr
ic
e
s
us
in
g t
he
be
s
t
m
ul
ti
va
r
ia
te
m
ode
l.
T
he
be
s
t
m
ul
ti
va
r
ia
te
m
ode
l
pe
r
f
or
m
a
nc
e
us
e
d t
o f
or
e
c
a
s
t
A
B
C
'
s
s
to
c
k pr
ic
e
i
s
6.382%
, w
hi
le
t
he
m
ode
l'
s
pe
r
f
or
m
a
nc
e
on
D
E
F
'
s
s
to
c
k ha
s
a
n a
c
c
ur
a
c
y of
4.466%
.
T
a
bl
e
6
. M
od
e
l
pe
r
f
or
m
a
nc
e
on othe
r
c
om
pa
ni
e
s
'
s
to
c
k (
U
ni
va
r
ia
te
)
S
t
oc
k
M
A
P
E
(
%
)
M
S
E
V
a
l
i
da
t
i
on
R
unt
i
m
e
(
s
)
A
B
C
3.03
0.00017
2.771
D
E
F
6.02
0.0025
2.399
T
a
bl
e
7
. M
od
e
l
pe
r
f
or
m
a
nc
e
on othe
r
c
om
pa
ni
e
s
'
s
to
c
k (
M
ul
ti
va
r
ia
te
)
S
t
oc
k
M
A
P
E
(
%
)
M
S
E
V
a
l
i
da
t
i
on
R
unt
i
m
e
(
s
)
A
B
C
6.382
0.00025
3.603
D
E
F
4.466
0.0017
3.574
(
a
)
(
b)
F
ig
ur
e
7
. C
om
pa
r
is
on of
a
c
tu
a
l
a
nd f
or
e
c
a
s
t
da
ta
on A
B
C
’
s
s
to
c
k (
a
)
a
nd D
E
F
’
s
s
to
c
k
(
b)
(
U
ni
va
r
ia
te
)
(
a
)
(
b)
F
ig
ur
e
8
. C
om
pa
r
is
on of
a
c
tu
a
l
a
nd f
or
e
c
a
s
t
da
ta
, (
a
)
A
B
C
’
s
s
to
c
k, (
b)
D
E
F
’
s
s
to
c
k (
M
ul
ti
va
r
ia
te
)
3.5.
M
e
t
h
od
c
om
p
ar
is
on
I
n
T
a
bl
e
8,
th
e
E
lm
a
n
-
R
N
N
m
e
th
od
ha
s
a
hi
ghe
r
pe
r
f
or
m
a
nc
e
th
a
n
F
F
N
N
.
I
t
c
a
n
b
e
s
e
e
n
f
r
om
th
e
M
A
P
E
pr
oduc
e
d
by
E
lm
a
n
-
R
N
N
is
1.546%
,
w
hi
le
F
F
N
N
is
1.835%
.
E
ve
n
th
ough
th
e
p
e
r
f
or
m
a
nc
e
of
th
e
m
ode
l
pr
oduc
e
d
by
th
e
E
lm
a
n
-
R
N
N
i
s
lo
w
e
r
th
a
n
L
S
T
M
a
nd G
R
U
,
w
he
r
e
bot
h
m
e
th
ods
c
a
n
pr
oduc
e
M
A
P
E
va
lu
e
s
of
1.345%
a
nd
1.352
%
c
ons
e
c
ut
iv
e
ly
,
t
he
r
unt
im
e
o
f
E
lm
a
n
-
R
N
N
is
s
ig
ni
f
ic
a
nt
ly
f
a
s
te
r
.
T
he
E
lm
a
n
-
R
N
N
r
e
qui
r
e
s
3.974 s
e
c
onds
, w
hi
le
t
he
L
S
T
M
a
nd G
R
U
r
e
qui
r
e
38.956 a
nd 37.84 s
e
c
ond
s
.
I
n
T
a
bl
e
9,
th
e
E
lm
a
n
-
R
N
N
m
e
th
od
ha
s
a
hi
ghe
r
pe
r
f
or
m
a
nc
e
th
a
n
F
F
N
N
.
I
t
c
a
n
b
e
s
e
e
n
f
r
om
th
e
M
A
P
E
pr
oduc
e
d
by
th
e
E
lm
a
n
-
R
N
N
is
1.558%
,
w
hi
le
F
F
N
N
i
s
2.961%
.
T
he
pe
r
f
or
m
a
nc
e
is
lo
w
e
r
th
a
n
th
e
L
S
T
M
a
nd
G
R
U
,
w
he
r
e
bot
h
m
e
th
ods
c
a
n
pr
oduc
e
M
A
P
E
va
lu
e
s
of
1.351%
a
nd
1.375%
.
H
ow
e
ve
r
,
in
te
r
m
s
of
s
pe
e
d,
E
lm
a
n
-
R
N
N
is
s
ig
ni
f
ic
a
nt
ly
be
tt
e
r
th
a
n
th
e
L
S
T
M
a
nd
G
R
U
m
e
th
ods
be
c
a
u
s
e
it
r
e
qui
r
e
s
4.020
s
e
c
onds
,
w
hi
le
th
e
two
m
e
th
ods
r
e
qui
r
e
33.533
a
nd
19.409
s
e
c
onds
c
on
s
e
c
ut
iv
e
ly
.
A
c
om
pa
r
is
on
of
th
e
m
ode
l
pe
r
f
or
m
a
nc
e
be
twe
e
n univa
r
ia
te
a
nd mul
ti
va
r
ia
te
m
ode
ls
a
s
s
how
n i
n
F
ig
ur
e
9
(
a
-
b
)
.
600
700
800
900
1000
1100
1200
1300
1400
A
pr
02, 2018
A
pr
16, 2018
A
pr
30, 2018
M
a
y 14, 2018
M
a
y 28, 2018
J
un 11, 2018
J
un 25, 2018
J
ul
09, 2018
J
ul
23, 2018
A
u
g
0
6
,
2
0
1
8
A
ug 20, 2018
S
e
p 03, 2018
S
e
p
1
7
,
2
0
1
8
O
c
t
01, 2018
O
c
t
15, 2018
O
c
t
29, 2018
N
ov 12, 2018
N
ov 26, 2018
D
e
c
10, 2018
D
e
c
24, 2018
J
a
n 07, 2019
J
a
n 21, 2019
F
e
b 04, 2019
F
e
b 18, 2019
M
a
r
04, 2019
M
a
r
18, 2019
A
pr
01, 2019
A
pr
15, 2019
A
pr
29, 2019
M
a
y 13, 2019
M
a
y 27, 2019
J
un 10, 2019
J
un 24, 2019
J
ul
08, 2019
J
ul
22, 2019
A
ug 05, 2019
A
ug 19, 2019
S
e
p 02, 2019
S
e
p 16, 2019
S
e
p 30, 2019
P
r
i
c
e
D
a
t
e
A
c
t
ua
l
F
or
e
c
a
s
t
1000
2000
3000
4000
5000
6000
7000
A
pr
02, 2018
A
pr
16, 2018
A
pr
30, 2018
M
a
y
1
4
,
2
0
1
8
M
a
y 28, 2018
J
un 11, 2018
J
un 25, 2018
J
ul
09, 2018
J
ul
23, 2018
A
ug 06, 2018
A
ug 20, 2018
S
e
p
0
3
,
2
0
1
8
S
e
p 17, 2018
O
c
t
01, 2018
O
c
t
15, 2018
O
c
t
29, 2018
N
ov 12, 2018
N
ov 26, 2018
D
e
c
10, 2018
D
e
c
2
4
,
2
0
1
8
J
a
n 07, 2019
J
a
n 21, 2019
F
e
b 04, 2019
F
e
b 18, 2019
M
a
r
04, 2019
M
a
r
18, 2019
A
pr
01, 2019
A
p
r
1
5
,
2
0
1
9
A
pr
29, 2019
M
a
y 13, 2019
M
a
y 27, 2019
J
un 10, 2019
J
un 24, 2019
J
ul
08, 2019
J
ul
22, 2019
A
u
g
0
5
,
2
0
1
9
A
ug 19, 2019
S
e
p 02, 2019
S
e
p 16, 2019
S
e
p 30, 2019
P
r
i
c
e
D
a
t
e
A
c
t
ua
l
F
or
e
c
a
s
t
600
700
800
900
1000
1100
1200
1300
1400
A
pr
02, 2018
A
pr
16, 2018
A
pr
30, 2018
M
a
y 14, 2018
M
a
y 28, 2018
J
u
n
1
1
,
2
0
1
8
J
un 25, 2018
J
ul
09, 2018
J
ul
23, 2018
A
ug 06, 2018
A
ug 20, 2018
S
e
p 03, 2018
S
e
p 17, 2018
O
c
t
01, 2018
O
c
t
15, 2018
O
c
t
29, 2018
N
ov 12, 2018
N
o
v
2
6
,
2
0
1
8
D
e
c
10, 2018
D
e
c
24, 2018
J
a
n 07, 2019
J
a
n 21, 2019
F
e
b 04, 2019
F
e
b 18, 2019
M
a
r
04, 2019
M
a
r
18, 2019
A
pr
01, 2019
A
pr
15, 2019
A
pr
29, 2019
M
a
y 13, 2019
M
a
y 27, 2019
J
un 10, 2019
J
un 24, 2019
J
ul
08, 2019
J
ul
22, 2019
A
ug 05, 2019
A
u
g
1
9
,
2
0
1
9
S
e
p 02, 2019
S
e
p 16, 2019
S
e
p 30, 2019
P
r
i
c
e
D
a
t
e
A
c
t
ua
l
F
or
e
c
a
s
t
1000
2000
3000
4000
5000
6000
7000
A
pr
02, 2018
A
pr
17, 2018
M
a
y 02, 2018
M
a
y 17, 2018
J
un 01, 2018
J
un 18, 2018
J
ul
03, 2018
J
ul
18, 2018
A
ug 02, 2018
A
ug 17, 2018
S
e
p 03, 2018
S
e
p 18, 2018
O
c
t
03, 2018
O
c
t
18, 2018
N
o
v
0
2
,
2
0
1
8
N
ov 19, 2018
D
e
c
04, 2018
D
e
c
19, 2018
J
a
n
0
3
,
2
0
1
9
J
a
n 18, 2019
F
e
b 04, 2019
F
e
b 19, 2019
M
a
r
06, 2019
M
a
r
21, 2019
A
pr
05, 2019
A
pr
22, 2019
M
a
y 07, 2019
M
a
y 22, 2019
J
un 06, 2019
J
un 21, 2019
J
ul
08, 2019
J
ul
23, 2019
A
ug 07, 2019
A
ug 22, 2019
S
e
p 06, 2019
S
e
p 23, 2019
P
r
i
c
e
D
a
t
e
A
c
t
ua
l
F
or
e
c
a
s
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
oc
k
pr
ic
e
f
o
r
e
c
as
t
of
m
a
c
r
o
-
e
c
onomic
f
a
c
to
r
us
in
g
r
e
c
u
r
r
e
nt
ne
ur
al
ne
tw
or
k
(
M
. R
e
z
a P
ahl
aw
an
)
81
T
a
bl
e
8
. U
ni
va
r
ia
te
m
ode
l
pe
r
f
or
m
a
nc
e
M
e
t
hod
M
A
P
E
(
%
)
R
unt
i
m
e
(
s
)
E
l
m
a
n
-
R
N
N
1.546
3.974
F
F
N
N
1.835
2.137
L
S
T
M
1.345
38.956
G
R
U
1.352
37.84
T
a
bl
e
9
. M
ul
ti
va
r
ia
te
m
ode
l
pe
r
f
or
m
a
nc
e
M
e
t
hod
M
A
P
E
(
%
)
R
unt
i
m
e
(
s
)
E
l
m
a
n
-
R
N
N
1.558
4.020
F
F
N
N
2.961
3.193
L
S
T
M
1.351
33.533
G
R
U
1.375
19.409
T
o
s
um
up,
E
lm
a
n
-
R
N
N
pe
r
f
or
m
a
nc
e
f
or
bot
h
uni
va
r
ia
te
a
nd
m
ul
ti
va
r
ia
te
m
ode
ls
is
good
a
nd
f
a
s
t
e
nough
c
om
pa
r
e
to
th
e
ot
he
r
m
e
th
ods
be
c
a
us
e
in
te
r
m
s
of
M
A
P
E
is
not
th
e
w
or
s
t
a
nd
th
e
r
unt
im
e
is
not
th
e
lo
nge
s
t.
S
im
il
a
r
to
th
e
uni
va
r
ia
te
m
ode
l,
th
e
m
ul
ti
va
r
ia
t
e
m
ode
l
pr
oduc
e
d
by
th
e
E
lm
a
n
-
R
N
N
is
s
ti
ll
lo
w
e
r
th
a
n t
he
L
S
T
M
a
nd G
R
U
w
he
r
e
bot
h
m
e
th
ods
c
a
n pr
oduc
e
M
A
P
E
va
lu
e
s
of
1.351%
a
nd 1.375%
.
I
n t
e
r
m
s
o
f
s
pe
e
d,
E
lm
a
n
-
R
N
N
i
s
be
tt
e
r
th
a
n
th
e
L
S
T
M
a
nd
G
R
U
m
e
th
od
s
be
c
a
us
e
it
r
e
qui
r
e
s
4.020
s
e
c
ond
s
w
hi
le
th
e
two
m
e
t
hods
r
e
qui
r
e
33.533
a
nd
19.409
s
e
c
onds
.
H
ow
e
ve
r
,
E
lm
a
n
-
R
N
N
is
s
ti
ll
s
lo
w
e
r
th
a
n
F
F
N
N
w
hi
c
h
r
e
qui
r
e
s
3.193
s
e
c
ond
s
.
I
t
c
a
n
be
c
onc
lu
de
d
th
a
t
th
e
a
ddi
ti
on
of
a
m
a
c
r
o
-
e
c
onomi
c
f
a
c
to
r
doe
s
not
pr
ovi
de
a
be
tt
e
r
m
ode
l
pe
r
f
or
m
a
nc
e
.
(
a
)
(
b)
F
ig
ur
e
9
. C
om
pa
r
is
on of
t
he
m
ode
l
pe
r
f
or
m
a
nc
e
, (
a
)
U
ni
va
r
ia
te
, (
b)
M
ul
ti
va
r
ia
te
4.
C
O
N
C
L
U
S
I
O
N
F
r
om
th
e
r
e
r
e
s
ul
t,
th
e
r
e
is
a
n
in
di
c
a
ti
on
th
a
t
th
e
e
xc
ha
nge
r
a
t
e
doe
s
not
ha
ve
a
s
ig
ni
f
ic
a
nt
im
pa
c
t,
a
lt
hough
th
e
s
to
c
k
pr
ic
e
a
nd
th
e
e
xc
ha
nge
r
a
te
ha
ve
a
pos
i
ti
ve
c
or
r
e
la
ti
on.
F
ur
th
e
r
m
or
e
,
it
be
c
om
e
s
a
n
a
ddi
ti
ona
l
in
s
ig
ht
f
or
in
ve
s
to
r
s
if
th
e
y
w
a
nt
to
c
ons
id
e
r
th
e
e
xc
ha
nge
r
a
te
f
a
c
to
r
in
s
to
c
k
in
ve
s
ti
ng,
e
s
pe
c
ia
ll
y
w
he
n
th
e
R
upi
a
h
is
w
e
a
ke
n
e
d
a
ga
in
s
t
th
e
U
S
D
ol
la
r
.
F
or
f
ur
th
e
r
r
e
s
e
a
r
c
h,
one
c
a
n
c
ons
id
e
r
ot
he
r
m
a
c
r
o
-
e
c
onomi
c
s
e
c
to
r
s
s
u
c
h
a
s
oi
l
pr
ic
e
s
,
in
f
la
ti
on,
a
nd
gr
os
s
dom
e
s
ti
c
pr
oduc
t
(
G
D
P
)
or
th
e
im
pa
c
t
of
publ
ic
s
oc
ia
l
s
e
nt
im
e
nt
a
nd gov
e
r
nm
e
nt
pol
it
ic
a
l
pol
ic
ie
s
on s
to
c
k pr
ic
e
s
. I
t
is
c
a
us
e
d by ma
ny f
a
c
to
r
s
t
ha
t
c
a
n
a
f
f
e
c
t
th
e
s
to
c
k
pr
ic
e
.
T
he
E
lm
a
n
-
R
N
N
out
pe
r
f
or
m
s
F
F
N
N
in
te
r
m
s
of
m
ode
l
pe
r
f
or
m
a
nc
e
e
ve
n
th
ough
it
is
lo
w
e
r
in
t
he
r
unt
im
e
m
ode
l.
E
ve
n
th
ough
th
e
pe
r
f
or
m
a
nc
e
of
th
e
uni
v
a
r
ia
te
a
nd
m
ul
ti
va
r
ia
te
m
ode
l
pr
oduc
e
d
by
th
e
E
lm
a
n
-
R
N
N
is
l
ow
e
r
th
a
n
L
S
T
M
a
nd
G
R
U
,
th
e
r
unt
im
e
of
El
m
a
n
-
R
N
N
is
s
ig
ni
f
ic
a
nt
ly
f
a
s
te
r
.
D
e
te
r
m
in
in
g
th
e
va
lu
e
th
a
t
c
a
n
be
e
nt
e
r
e
d
or
not
on
th
e
ga
te
s
tr
uc
tu
r
e
in
th
e
L
S
T
M
a
nd
G
R
U
m
e
th
ods
pl
a
ys
a
r
ol
e
in
w
hy
th
e
r
unt
im
e
m
ode
l
is
l
onge
r
.
A
C
K
N
O
WL
E
D
G
E
M
E
N
T
S
T
he
a
ut
hor
s
w
oul
d
li
ke
to
th
a
nk
D
a
ta
E
ngi
ne
e
r
in
g
a
nd
B
us
in
e
s
s
I
nt
e
ll
ig
e
nc
e
L
a
bor
a
to
r
y
a
nd
I
nf
or
m
a
ti
on
S
ys
te
m
s
D
e
pa
r
tm
e
nt
,
I
ns
ti
tu
t
T
e
knol
ogi
S
e
pul
u
h
N
ope
m
be
r
,
I
ndone
s
i
a
f
or
s
uppor
ti
ng
th
is
r
e
s
e
a
r
c
h.
T
he
a
ut
hor
s
a
ls
o
th
a
nk
I
nt
e
r
na
ti
ona
l
J
our
na
l
of
A
r
ti
f
ic
ia
l
I
nt
e
ll
ig
e
nc
e
’
s
R
e
vi
e
w
e
r
f
or
th
e
ir
va
lu
a
bl
e
c
om
m
e
nt
on
th
is
r
e
s
e
a
r
c
h.
R
E
F
E
R
E
N
C
E
S
[1]
PT
Bursa
Efek
Indonesia
,
“
41
Years
Of
Contributing
to
Modern
Ca
pital
Markets
in
the
Digital
Economy
Era
,”
available:
https://www.idx.co.id/id
-
id/berita/pr
ess
-
release
-
detail/?emitenCode=
936 (accessed Nov. 29, 2019).
[2]
D.
Wartati
,
N.
A.
Masruroh,
“Application
of
Artificial
Neural
Net
works
and
Particle
Swarm
Optimization
for
Predic
ting
the
Indone
sia
Stock
Excha
nge
Stock
P
rice
Index,”
J.
Tek
nosains
,
vol.
6,
no.
1,
p
p
.
22
-
30
,
201
6
,
doi:
10.22146/teknosains.27616.
[3]
A.
Dingli
,
K.
S.
Fournier,
“Financial
time
series
forecasting
-
a
de
ep
learning
approach,”
Int.
J.
Mach.
Learn.
Comput.
, vol. 7, no. 5, pp. 118
-
122, 2017, doi:
10.18178/ijmlc.2017.7.5.632.
[4]
A. S. Sukamt
o
,
W. Seti
awan, “St
ock Forecas
ting B
ased on P
ast Data
with F
uzzy Time
Series Ap
proach”
J. Edukas
i
dan Penelit. Inform.
, vol. 4, no. 2, p
p
. 192
-
196
, 2018, doi: 10.26418/jp.v4i2.29469.
1.546
1.835
1.345000
1.352000
0
0.5
1
1.5
2
E
l
m
a
n-
R
N
N
F
F
N
N
L
S
T
M
G
R
U
M
A
P
E
(
%
)
Me
t
ho
d
1.558
2
.9
6
1
1.351
1.375
0
,0
0
1
0
,0
0
2
0
,0
0
2
0,0
03
0
,0
0
3
E
l
m
a
n-RN
N
F
F
N
N
L
S
T
M
G
RU
M
A
P
E
(
%
)
M
e
th
od
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
1, M
a
r
c
h
20
21
:
74
–
83
82
[5]
H.
Maqsood
et
al.
,
“A
lo
cal
and
global
event
sentiment
based
efficient
stock
exchange
forecasting
using
deep
learning,”
Int.
J.
Inf.
Manage.
,
vol.
50,
no.
December
2018,
pp.
432
-
451,
2020,
doi:
10.1016/j.ijinfomgt.2019.07.011.
[6]
I.
Komang
Gede
Darma
Putra
Sadia
,
I.
Gde
Ary
Wira
jaya,
“The
Effect
of
Macroeconomic
Variables
on
the
Intrinsic
Value
of
Stock,”
E
-
Jurnal
Akunt.
Univ.
Udayana
,
vol.
27,
no.
3,
pp.
2406
-
2431
,
201
9
DOI:
10.24843/EJA.2019.v27.i03.p28
[7]
M.
Tachibana,
“Relationship
between
stock
and
currency
markets
conditio
nal
on
the
US
stock
returns:
A
vine
copula app
roach,”
J. Multi
natl.
Financ.
Manag.
, vol. 46, pp. 75
-
106, 2018, doi: 10.1016/j.mulfin.2018.05.001.
[8]
K.
Kumar.
J
,
K.
A,
“Prediction
of
Future
Stock
Close
Price
using
Prop
osed
Hybrid
ANN
Model
of
Functional
L
ink
Fuzzy
Logic
Neural
Model
(FLFNM),”
IAES
Int.
J.
Artif.
Intell.
,
vol.
1,
no.
1,
pp.
25
-
30,
2012,
doi:
10.11591/ij
-
ai.v1i1
.362.
[9]
C.
Voyant
et
al.
,
“Machine
learning
methods
for
solar
radiation
forec
asting:
A
review,”
Renew.
Energy
,
vol.
105,
pp.
569
-
582, 2017, doi: 10.1016/j.renene.2016.12.095.
[10]
N.
H.
A.
Rahman
,
M.
H.
Lee,
“Artificial
neural
network
forecasting
p
erformance
with
missing
value
imputations,”
IAES Int. J.
Artif. In
tell.
, vol. 9, no. 1, pp. 33
-
39, 2020, doi: 10.11591/ijai.v9.i1.pp3
3
-
39.
[11]
R.
Gupta,
A.
N.
Singh,
and
A.
Singhal,
“Application
of
ANN
for
water
quality
index,”
Int.
J.
Mach.
Learn.
Comput.
, vol. 9, no. 5, pp. 688
-
693, 2019, doi: 10.18178/ijmlc.2019.9.5.859.
[12]
Y.
Zhang,
X.
Wang,
and
H.
Tang,
“An
improved
Elman
neural
netwo
rk
with
piecewise
weighted
gradient
for
time
series predic
tion,”
Neurocom
puting
, vol. 359, pp. 199
-
208, 2019, doi: 10.1016/j.neucom.2019.06.001.
[13]
J.
Wang
,
J.
Wang,
“Forecasting
energy
marke
t
indices
with
recurre
nt
neural
networks
:
Case
study
of
crude
oil
price fluctuations,”
Energy
, vol. 102, pp. 365
-
374, 2016, doi: 10.1016/j.energy.2016.02.098.
[14]
N. Mohamed, M. H.
Ahmad, Z. Ismail,
and K. A. Arshad, “An Art
ifici
al Neural Networks Foreca
sting for Malaysia
’ s Load,”
Stat. Forum Teor. dan Apl. Stat.
, vol. 8, no. 2, pp. 111
-
118, 2008.
[15]
L.
Di
Persio
and
O.
Honchar,
“Artificial
neural
networks
architectures
for
stock
price
prediction:
Comparisons
an
d
applicati
ons,”
Int. J. Ci
rcuits, S
yst.
Signal Process.
, vol. 10, pp. 403
-
413, 2016.
[16]
A.
J.
P.
Samarawickrama
and
T.
G.
I.
Fernando,
“A
recurrent
neura
l
network
approach
in
predicting
daily
stock
prices an application to the Sri Lankan stock
market,”
2017 IEEE Int.
Conf. Ind. Inf. Syst. ICI
IS 2017
-
Proc.
, pp. 1
-
6, 2018, doi: 10.1109/ICIINFS.2017.8300345.
[17]
J.
Zhao,
D.
Zeng,
S.
Liang,
H.
Kang,
and
Q.
Liu,
“Prediction
model
f
or
stock
price
trend
based
on
recurrent
neural
network,”
J. Ambien
t Intel
l. Human
iz. Co
mput.
, 2020, doi: 10.1007/s12
652
-
020
-
02057
-
0.
[18]
J. Wang, J. Wa
ng, W.
Fang, and H.
Niu, “Financial Time
Series Predi
ction Using Elman Recu
rrent Random
Neural
Networks,”
Hindawi
Publ.
Corp.
Comput.
Intell.
Neurosci.
,
vol.
2016,
pp.
1
-
14,
2016,
[Online].
Available:
http://dx.doi.org/1
0.1155/2016/4742515.
[19]
W.
Chen,
C.
K.
Yeo,
C.
T.
Lau,
and
B.
S.
Lee,
“Leveraging
social
media
news
to
predict
stock
index
movement
using RNN
-
boost,”
Data Knowl
. Eng.
, vol. 118, no. August, pp. 14
-
24, 2018, doi: 10.1016/j.datak.2018.08.003.
[20]
H.
Hassa
ni,
M.
Kalantari,
and
Z.
Ghodsi,
“Evaluating
the
Perfor
mance
of
Multiple
Imputation
Methods
for
Handling
Missing
Values
in
Time
Series
Data:
A
Study
Focused
o
n
East
Africa,
Soil
-
Carbona
te
-
Stable
Isotope
Data,”
Stats
, vol. 2, no. 4, pp. 457
-
467, 2019,
doi: 10.3390/stats2040032.
[21]
L. Al Shalabi, Z. Shaaban, and B.
Kasasbeh, “Data Mining: A
Preprocessing Engine,”
J. Compu
t. Sci.
, vol. 2,
no. 9
,
pp. 735
-
739, 2006, doi: 10.3844/jcssp.2006.735.739.
[22]
T.
Jerez
and
W.
Kristjanpoller,
“Effects
of
the
vali
dation
set
on
stock
returns
forecasting,”
Expert
Syst.
Appl.
,
vol.
150, 2020, doi: 10.1016/j.eswa.2020.113271.
[23]
A.
Graves,
“Supervised
Sequence
Labelling
with
Recurrent
Neural
Networks
[Ph.
D.
dissertation],”
Tech.
Univ.
Munich, Ger.
, 2008
,
DOI: 10.100
7/978
-
3
-
642
-
24797
-
2
.
[24]
P.
Mishra
and
P.
Nath
Pandey,
“Elman
RNN
based
classification
of
p
roteins
sequences
on
account
of
their
mutual
information,”
J. Theor
. Biol.
, vol. 311, pp. 40
-
45, 2012, doi: 10.1016/j.jtbi.2012.06.039.
[25]
M.
Rhanoui,
S.
Yousfi,
M.
Mikram,
and
H.
Merizak,
“Forecasting
fi
nancial
budget
time
series
:
ARIMA
random
walk
vs
LSTM
neural
network,”
IAES
Int.
J.
Artif.
Intell.
,
vol.
8,
no.
4,
pp.
317
-
327,
2019,
doi:
10.11591/ijai.v8.i4.pp317
-
327.
[26]
A.
P.
Piotrowski,
J.
J.
Napiorkowski,
and
A.
E.
Piotrowska,
“Impac
t
of
deep
learning
-
based
dropout
on
shallow
neural
networks
applied
to
stream
temperature
modelling,”
Earth
-
S
cience
Rev.
,
vol.
201,
no.
August
2019,
p
.
103076, 2020, doi: 10.1016/j.earscirev.2019.103076.
[27]
P.
M.
Radiuk,
“Impa
ct
of
Training
Set
Batch
Size
on
the
Perform
ance
of
Convolutional
Neural
Networks
fo
r
Diverse Datasets,”
Inf. Techno
l. Manag.
Sci.
, vol. 20, no. 1, pp. 20
-
24, 2018, doi: 10.1515/itms
-
2017
-
0003.
[28]
A.
Sherstinsky,
“Fundamentals
of
Recurrent
Neural
Network
(RN
N)
and
Long
Short
-
Term
Memory
(LSTM
)
network,”
Phys. D Nonlinear Phenom.
, vol. 404, p. 132306, 2020, doi: 10.1016/j.physd.2019.132306.
[29]
A.
de
Myttenaere,
B.
Golden,
B.
Le
G
rand,
and
F.
Rossi,
“Mean
Abso
lute
Percentage
Error
for
regression
models,”
Neurocom
puting
, vol. 192, pp. 38
-
48, 2016, doi: 10.1016/j.neucom.2015.12.114.
[30]
S.
Gao
et
al.
,
“Short
-
term
runoff
prediction
with
GRU
and
LS
TM
networks
without
requiring
time
ste
p
optimization
during
sample
generation,”
J.
Hydrol.
,
vol.
5
89,
no.
June,
p
.
125188,
2020,
doi:
10.1016/j.jhydrol.2020.125188.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
St
oc
k
pr
ic
e
f
o
r
e
c
as
t
of
m
a
c
r
o
-
e
c
onomic
f
a
c
to
r
us
in
g
r
e
c
u
r
r
e
nt
ne
ur
al
ne
tw
or
k
(
M
. R
e
z
a P
ahl
aw
an
)
83
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Muhammad
Reza
Pahlawa
n
received
his
bachelor’s
degree
in
I
nformation
Systems
from
Institut
Teknologi
Sepuluh
Nopember,
Surabaya,
Indonesia.
In
c
ollege,
he
was
a
grading
assistan
t in t
he Model
Driven DS
S, Dig
ital
Startu
p, and Dat
abase Tec
hnology
course. In
his fi
nal
year,
he
tak
es
the
field
of
interest
in
Data
Engineering
and
Business
Intelligence
Laboratory
(RDIB).
He
currently
joined
Information
Systems
from
Institut
Tekno
logi
Sepuluh
Nopember
in
his Master study.
Edwin
Riksakomara
received
his
bachelor
degree
in
computin
g
and Master
degree
in
industrial
engineerin
g
from
Instit
ut
Teknolog
i
Sepuluh
Nopember,
Surabaya,
In
donesia,
in
1994
and
2001,
respectively.
He
is
currently
a
lecturer
within
the
Data
Engineering
and
Business
Intelligen
ce
research
group
in
Information
Syst
ems
Department,
Institut
Teknologi
Sepuluh
Nopember.
His
main rese
arch inte
rests are
Data Mining
and E
-
Business.
Raras
Tyasnurita
received
her
bachelor'
s
degree
in
computing
from
Institut
Teknologi
Sepuluh
Nopember,
Surabaya,
Indonesia,
and
the
MBA
degree
from
Natio
nal
Taiwan
University
of
Scienc
e
and
Techn
ology,
Taipe
i,
Taiwa
n,
in
2008
and
2012,
respe
ctively
.
She
joined
Information
Technology
Applicatio
n
and
Integration
Laboratory
in
her
Master
study.
She
is
a
PhD
candidate
in
Information
Techno
logy,
University
o
f
Nottingham
.,
United
Kingdom.
She
is
currently
a
lecturer
within
the
Data
Engineeri
ng
and
Busines
s
Intel
ligence
research
group
in
Information
Systems
Department,
Institut
Teknologi
Sepuluh
Nope
mber
.
Her
main
research
interests
are
Artificial
Intelligenc
e,
Combinatorial
Optimisation,
Machine
Learning,
and
Forec
asting.
Ahmad
Muklason
received
his
bachelor'
s
degree
in
computing
from
Institut
Teknologi
Sepuluh
Nopember,
Surabaya,
Indonesia,
in
200
6,
Master’s
degree
in
C
omputer
and
Information
Scienc
es
Depar
tment
from
Univer
siti
Teknol
ogi
Petrona
s,
Malays
ia,
i
n
2009,
and
PhD
degre
e
in
Information
Technology,
School
of
Computer
Sci
ence,
from
The
University
of
Nottin
gham,
United
Kingdom,
in
2017.
He
i
s
currently
a
Head
of
Data Engineering
and
Business
Intelligence
Laboratory in Information Systems Department, Insti
tut Teknologi Se
puluh Nopember. His main
research
interests
are
Hyper/Meta
-
heuristics,
Examination
Timetablin
g
Problems,
Combinatorial
Optim
isations Proble
ms, and Fairn
ess Urba
n Computing.
Faizal
Mahananto
received
the
S.Kom.
degree
from
the
Institut
Tek
nologi
Sepuluh
Nopember,
Indonesia,
in
2008
and
the
M.Eng.
and
Ph.D.
degree
from
the
Kuma
moto
University
,
Japan,
in
2012
and
2015,
respe
ctively.
He
is
currently
working
in
the
Departme
nt
of
Informati
on
System
s,
Institut
Teknologi
Sepuluh
Nopember,
Indonesia.
His
current
r
esearch
interests
include
healthcare
Information
Systems,
Business
Intelligence
in
Medicine
a
nd
Healthcare,
Heart
Rate
V
ariabili
ty, and
Biomedi
cal Sig
nal Proces
sing.
Retno
Aulia
Vinarti
received
her
bachelor
and
Master
degree
in
computing
from
Institut
Teknologi
Sepuluh
Nopember
,
Surabaya,
Indonesia,
in
2009
and
2011,
respectively.
She
received
her
PhD
degree
from
Trinity
College
Dublin,
Dublin,
Irelan
d,
in
2019.
She
is
currently
a
lecturer
within
the
Data
Engineering
and
Business
Intelligence
res
e
arch
group
in
Information
System
s
Department,
Institut
Teknologi
Sepuluh
Nopember.
Her
main
research
interests
are
Healthcare
Informatics,
Rule
-
based
Systems,
Bayesian
Networks,
Knowledge
-
driven
Models,
and Predi
ction.
Evaluation Warning : The document was created with Spire.PDF for Python.