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
-
AI
)
V
ol
. 14, No. 5, O
c
to
be
r
2025
, pp.
3624
~
3633
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
5
.pp
3624
-
3633
3624
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
S
t
oc
k
m
ar
k
e
t
l
i
q
u
i
d
i
t
y:
h
yb
r
i
d
d
e
e
p
l
e
ar
n
i
n
g ap
p
r
oac
h
e
s f
or
p
r
e
d
i
c
t
i
on
M
ar
ia
m
A
it
A
l
1
, S
ai
d
A
c
h
c
h
ab
1
, Y
ou
n
e
s
L
ah
r
ic
h
i
2
1
E
N
S
I
A
S
, M
oha
m
m
e
d V
U
ni
ve
r
s
i
t
y
i
n R
a
ba
t
, R
a
ba
t
, M
or
oc
c
o
2
T
he
H
i
ghe
r
I
ns
t
i
t
ut
e
of
C
om
m
e
r
c
e
a
nd B
us
i
ne
s
s
A
dm
i
ni
s
t
r
a
t
i
on (
G
r
oupe
I
S
C
A
E
)
, C
a
s
a
bl
a
nc
a
, M
or
oc
c
o
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
S
e
p 16, 202
4
R
e
vi
s
e
d
J
un 26, 2025
A
c
c
e
pt
e
d
J
ul
13
,
2025
Predic
ting
stock
marke
t
liquidity
espec
ially
in
emerg
ing
or
frontie
r
fi
nancia
l
markets,
such
as
the
Casablanc
a
stock
exchange
(CSE),
presents
sign
ificant
challenges
given
the
relative
narrowness
and
volatil
ity
of
these
m
ark
ets.
In
this
paper,
we
conduct
a
comprehensive
study
to
address
the
pred
ictions
accuracy gaps
between fiv
e main d
eep learnin
g model
s:
convolut
ional
neural
network
(CNN),
long
short
-
term
memory
(LSTM),
b
idirectiona
l
LSTM
(
B
iLSTM),
and
two
hybrid
architec
tures,
CNN
-
LSTM
and
CNN
-
B
i
LSTM.
The
proposed
methodology
focused
on
training
and
testing
these
mo
dels
on
historical
data
from
the
CSE,
with
precision
on
capturing
both
spati
al
and
temporal
market
dynamics.
The
models
were
fine
-
tuned
usin
g
key
hyperparameters
and
validated
on
20%
of
the
dataset
to
ensure
r
eliable
results.
The
evaluation
of
performance
was
conducted
using
error
metrics
such
as
mean
squared
error
(MSE),
root
mean
squared
error
(RMS
E),
and
mean
absolute
error
(M
AE).
The
study
demonstrates
that
the
hybrid
CNN
-
biLSTM
model
consistently
outperformed
all
standalone
and
other
hybrid
models
in
predictive
accuracy.
This
underscores
the
considerable
pro
mise
of
hybrid
deep
learning
architectures
for
addressing
the
unique
challen
ges
of
predicting stock market liquidity in volatile and emerging financial marke
ts.
K
e
y
w
o
r
d
s
:
C
a
s
a
bl
a
n
c
a
s
to
c
k
e
xc
ha
nge
D
e
e
p l
e
a
r
ni
ng
N
e
ur
a
l
ne
twor
k
S
to
c
k m
a
r
ke
t
pr
e
di
c
ti
on
V
ol
a
ti
le
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
:
M
a
r
ia
m
A
it
A
l
E
N
S
I
A
S
, M
oha
m
m
e
d V
U
ni
ve
r
s
it
y
in
R
a
ba
t
R
a
ba
t,
M
or
oc
c
o
E
m
a
il
:
m
a
r
ia
m
_a
it
a
l@um5.a
c
.m
a
1.
I
N
T
R
O
D
U
C
T
I
O
N
S
to
c
k
m
a
r
ke
t
li
qui
d
it
y
p
r
e
di
c
ti
on
in
vo
la
ti
le
e
nvi
r
onm
e
nt
s
,
s
u
c
h
a
s
th
e
C
a
s
a
bl
a
nc
a
s
to
c
k
e
xc
ha
nge
(
C
S
E
)
,
ho
ld
s
s
ig
ni
f
ic
a
nt
im
por
ta
nc
e
f
or
va
r
io
us
m
a
r
ke
t
pa
r
ti
c
ip
a
nt
s
,
in
c
lu
di
ng
in
ve
s
to
r
s
,
is
s
ue
r
s
,
a
nd
r
e
gul
a
to
r
s
.
H
ow
e
ve
r
,
c
onv
e
nt
io
na
l
pr
e
di
c
ti
on
te
c
hni
que
s
f
r
e
que
nt
ly
s
tr
uggl
e
to
a
d
e
qua
te
ly
c
a
pt
ur
e
th
e
c
om
pl
e
x
a
nd
nonl
in
e
a
r
dyna
m
ic
s
of
f
in
a
nc
ia
l
da
ta
,
pa
r
ti
c
ul
a
r
ly
in
e
m
e
r
gi
ng
o
r
f
r
ont
ie
r
m
a
r
ke
ts
c
ha
r
a
c
te
r
iz
e
d
by l
im
it
e
d t
r
a
ns
pa
r
e
nc
y. M
a
r
ke
t
m
ic
r
os
tr
uc
tu
r
e
a
nom
a
li
e
s
, s
uc
h
a
s
di
s
c
r
e
pa
nc
i
e
s
i
n bi
d
-
a
s
k s
pr
e
a
d
s
or
t
r
a
di
ng
vol
um
e
s
,
in
tr
oduc
e
a
ddi
ti
ona
l
c
om
pl
e
xi
ti
e
s
in
pr
e
di
c
ti
ng
li
qui
di
ty
,
a
lo
ng
w
it
h
vol
a
ti
li
ty
a
nd
di
s
ti
nc
t
s
tr
uc
tu
r
a
l
in
f
lu
e
nc
e
s
[
1]
.
C
on
s
e
q
ue
nt
l
y,
r
e
s
e
a
r
c
h
e
r
s
ha
ve
in
c
r
e
a
s
i
ngl
y
e
xpl
o
r
e
d
m
a
c
h
in
e
l
e
a
r
ni
n
g
m
o
de
l
s
,
p
a
r
ti
c
ul
a
r
ly
de
e
p
le
a
r
n
in
g
a
r
c
h
it
e
c
t
ur
e
s
, d
ue
to
th
e
ir
p
r
of
i
c
i
e
n
c
y
in
ha
ndl
in
g e
xt
e
n
s
iv
e
d
a
t
a
s
e
t
s
a
nd
c
o
m
pl
e
x
p
a
tt
e
r
n
s
[
2]
–
[
4]
.
I
n
th
is
c
ont
e
xt
,
hybr
id
m
ode
ls
th
a
t
c
om
bi
ne
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
C
N
N
)
a
nd
lo
ng
s
hor
t
-
te
r
m
m
e
m
or
y
ne
twor
ks
(
L
S
T
M
)
e
xpr
e
s
s
e
d
r
e
m
a
r
ka
bl
e
pr
e
di
c
ti
ve
p
e
r
f
or
m
a
nc
e
.
N
one
th
e
le
s
s
,
th
e
ir
a
ppl
ic
a
ti
on
in
de
ve
lo
pi
ng f
in
a
nc
ia
l
c
ont
e
xt
s
, s
u
c
h a
s
M
or
oc
c
o, r
e
m
a
in
s
und
e
r
e
xpl
or
e
d [
5]
, [
6]
.
D
e
e
p
le
a
r
ni
ng
m
ode
ls
h
a
ve
ga
in
e
d
s
ig
ni
f
ic
a
nt
a
tt
e
nt
io
n
f
or
s
to
c
k
m
a
r
ke
t
pr
e
di
c
ti
on
be
c
a
u
s
e
of
th
e
ir
c
a
pa
c
it
y
to
r
e
c
ogni
z
e
in
tr
ic
a
te
pa
tt
e
r
ns
w
it
hi
n
ti
m
e
s
e
r
ie
s
da
ta
.
M
ode
ls
s
uc
h
a
s
C
N
N
s
ha
v
e
be
e
n
hi
ghl
y
e
f
f
e
c
ti
ve
a
t
e
xt
r
a
c
ti
ng
f
e
a
tu
r
e
s
,
w
he
r
e
a
s
L
S
T
M
s
de
m
on
s
tr
a
te
s
tr
ong
c
a
pa
bi
li
ti
e
s
in
c
a
pt
ur
in
g
te
m
por
a
l
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
m
ar
k
e
t
li
qui
di
ty
:
hy
br
id
de
e
p l
e
ar
ni
ng app
r
oac
he
s
f
or
pr
e
di
c
ti
on
…
(
M
ar
ia
m
A
it
A
l
)
3625
de
pe
nde
nc
ie
s
,
pa
r
ti
c
ul
a
r
ly
in
dyn
a
m
ic
e
nvi
r
onm
e
nt
s
[
7]
.
H
ybr
id
m
ode
ls
,
s
uc
h
a
s
th
e
C
N
N
-
L
S
T
M
a
nd
C
N
N
-
bi
di
r
e
c
ti
ona
l
L
S
T
M
(
B
iL
S
T
M
)
,
ha
ve
s
how
n
pr
om
is
in
g
r
e
s
ul
ts
by
uni
ti
ng
th
e
s
tr
e
ngt
hs
of
s
pa
ti
a
l
a
n
d
te
m
por
a
l
pa
tt
e
r
n
r
e
c
ogni
ti
on.
W
hi
le
L
S
T
M
s
a
r
e
c
om
m
onl
y
us
e
d
f
or
s
to
c
k
pr
ic
e
f
or
e
c
a
s
ti
ng
du
e
to
th
e
ir
a
bi
li
ty
to
c
a
pt
ur
e
te
m
por
a
l
s
e
que
nc
e
s
,
th
e
y
of
te
n
s
tr
uggl
e
w
it
h
ge
ne
r
a
li
z
a
ti
on
w
he
n
a
ppl
ie
d
to
unf
a
m
il
ia
r
f
in
a
nc
ia
l
e
nvi
r
onm
e
nt
s
.
B
iL
S
T
M
m
od
e
l
s
e
nh
a
n
c
e
L
S
T
M
s
b
y e
f
f
e
c
ti
v
e
l
y c
a
pt
ur
in
g
d
e
p
e
n
de
nc
ie
s
f
r
om
bo
th
t
h
e
pa
s
t
a
n
d
f
ut
ur
e
.
H
ow
e
v
e
r
,
th
e
y
c
om
e
w
it
h
in
c
r
e
a
s
e
d
c
om
p
ut
a
ti
ona
l
c
o
s
t
s
a
n
d
e
xt
e
n
de
d
t
r
a
i
ni
ng
dur
a
ti
on
s
[
7]
.
H
ybr
id
C
N
N
-
L
S
T
M
m
ode
ls
of
f
e
r
a
pr
om
is
in
g
s
ol
ut
io
n
to
im
pr
ove
pr
e
di
c
ti
on
a
c
c
ur
a
c
y
by
c
a
pt
ur
in
g
bot
h
lo
c
a
l
pa
tt
e
r
ns
a
nd
lo
ng
-
te
r
m
de
pe
nde
nc
ie
s
.
N
one
th
e
le
s
s
,
th
e
s
e
m
ode
ls
a
r
e
hi
ghl
y
s
e
ns
it
iv
e
to
hype
r
pa
r
a
m
e
te
r
s
,
m
a
ki
ng
th
e
m
c
h
a
ll
e
ngi
ng
to
opt
im
iz
e
f
or
r
e
a
l
-
w
or
ld
a
ppl
ic
a
ti
ons
[
8]
–
[
13]
.
T
he
hybr
id
C
N
N
-
B
iL
S
T
M
a
r
c
hi
te
c
tu
r
e
,
w
hi
c
h
c
om
bi
ne
s
C
N
N
'
s
s
p
a
ti
a
l
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
w
it
h
B
iL
S
T
M
s
te
m
por
a
l
m
ode
li
ng,
ha
s
s
how
n
e
ve
n
gr
e
a
te
r
pr
om
is
e
,
but
it
s
a
ppl
ic
a
ti
on
in
dyna
m
ic
a
ll
y
s
hi
f
ti
ng
a
nd
gr
ow
th
-
or
ie
nt
e
d
m
a
r
ke
ts
r
e
m
a
in
s
l
a
r
ge
ly
une
xpl
or
e
d [
14]
–
[
22]
.
T
he
obj
e
c
ti
ve
of
th
is
s
tu
dy
is
to
e
va
lu
a
t
e
a
nd
c
om
pa
r
e
th
e
pr
e
di
c
ti
ve
a
c
c
ur
a
c
y
of
va
r
io
us
de
e
p
le
a
r
ni
ng
a
r
c
hi
te
c
tu
r
e
s
f
or
f
or
e
c
a
s
ti
ng
s
to
c
k
m
a
r
ke
t
li
qui
di
ty
in
e
m
e
r
gi
ng
f
in
a
nc
ia
l
m
a
r
ke
ts
,
us
in
g
th
e
C
S
E
a
s
a
c
a
s
e
s
tu
dy. S
pe
c
if
ic
a
ll
y, w
e
i
nve
s
ti
g
a
te
f
iv
e
m
ode
ls
:
C
N
N
, L
S
T
M
, B
iL
S
T
M
, a
nd t
w
o hybr
id
m
ode
ls
, C
N
N
-
L
S
T
M
a
nd
C
N
N
-
B
iL
S
T
M
.
U
s
in
g
th
e
f
in
a
nc
ia
l
ti
m
e
s
s
to
c
k
e
xc
ha
nge
(
F
T
S
E
)
C
S
E
M
or
oc
c
o
15
in
de
x
a
s
a
ke
y
in
di
c
a
to
r
of
th
e
C
S
E
,
w
e
e
v
a
lu
a
te
th
e
s
e
m
ode
ls
th
r
ough e
r
r
or
m
e
tr
ic
s
s
uc
h
a
s
m
e
a
n
s
qua
r
e
d
e
r
r
or
(
M
S
E
)
,
r
oot
m
e
a
n
s
qua
r
e
d
e
r
r
or
(
R
M
S
E
)
,
a
nd
m
e
a
n
a
bs
ol
ut
e
e
r
r
or
(
M
A
E
)
to
de
te
r
m
in
e
th
e
ir
e
f
f
ic
a
c
y
in
f
or
e
c
a
s
ti
ng
m
a
r
ke
t
li
qui
di
ty
.
T
he
c
hoi
c
e
of
th
e
s
e
m
od
e
ls
is
ba
s
e
d
on
th
e
i
r
a
bi
li
ty
to
a
c
c
ur
a
te
ly
r
e
pr
e
s
e
nt
th
e
te
m
por
a
l
de
pe
nde
nc
ie
s
a
nd
s
pa
ti
a
l
pa
tt
e
r
ns
pr
e
s
e
nt
in
f
in
a
nc
ia
l
ti
m
e
s
e
r
ie
s
da
ta
.
T
he
goa
l
is
to
de
te
r
m
in
e
w
hi
c
h
m
ode
l
be
s
t
c
a
pt
ur
e
s
th
e
c
om
pl
e
x
te
m
por
a
l
a
nd
s
pa
ti
a
l
pa
tt
e
r
ns
in
s
to
c
k
m
a
r
ke
t
da
ta
,
pr
ovi
di
ng
a
r
e
li
a
bl
e
f
o
r
e
c
a
s
ti
ng
to
ol
f
or
m
a
r
ke
t
pa
r
ti
c
ip
a
nt
s
i
n vola
ti
le
e
nvi
r
onm
e
nt
s
.
T
he
f
in
di
ngs
a
im
to
f
il
l
a
s
ig
ni
f
ic
a
nt
ga
p
in
th
e
l
it
e
r
a
tu
r
e
r
e
ga
r
di
ng
de
e
p
le
a
r
ni
ng
a
ppl
ic
a
ti
ons
in
th
e
c
ont
e
xt
of
de
ve
lo
pi
ng
c
ount
r
ie
s
.
T
hi
s
s
tu
dy
s
tr
iv
e
s
to
s
ys
te
m
a
ti
c
a
ll
y
a
s
s
e
s
s
v
a
r
io
us
m
ode
l
de
s
ig
n
s
to
pr
ovi
de
e
s
s
e
nt
ia
l
gui
da
nc
e
f
or
f
in
a
nc
ia
l
pr
of
e
s
s
io
na
ls
,
m
a
r
ke
t
r
e
gul
a
to
r
s
,
a
nd
in
ve
s
to
r
s
,
a
id
in
g
th
e
ir
de
c
is
io
n
-
m
a
ki
ng
in
unpr
e
di
c
ta
bl
e
s
c
e
na
r
io
s
.
T
he
p
a
pe
r
is
s
tr
uc
tu
r
e
d
a
s
f
ol
lo
w
s
:
s
e
c
ti
on
2
de
s
c
r
ib
e
s
th
e
d
a
ta
s
our
c
e
s
,
m
od
e
l
a
r
c
hi
te
c
tu
r
e
s
,
tr
a
in
in
g
a
nd
va
li
da
ti
on
s
e
tu
p,
a
nd
e
va
lu
a
ti
on
m
e
tr
ic
s
us
e
d
to
m
e
a
s
ur
e
m
ode
l
pe
r
f
or
m
a
nc
e
.
S
e
c
ti
on
3
pr
e
s
e
nt
s
a
d
e
ta
il
e
d
c
om
pa
r
is
on
of
th
e
m
ode
l
s
,
in
te
r
pr
e
ti
ng
th
e
pe
r
f
or
m
a
nc
e
m
e
tr
ic
s
a
nd
c
om
pa
r
in
g
th
e
m
a
ga
in
s
t
e
xi
s
ti
ng
li
te
r
a
tu
r
e
to
hi
ghl
ig
ht
th
e
s
tr
e
ngt
hs
a
nd
li
m
it
a
ti
ons
of
e
a
c
h
m
ode
l,
w
hi
le
s
e
c
ti
on
4
s
um
m
a
r
iz
e
s
th
e
ke
y
r
e
s
ul
ts
,
e
xa
m
in
e
s
pr
a
c
ti
c
a
l
c
ons
e
qu
e
nc
e
s
,
a
nd
pr
opos
e
s
a
ve
nue
s
f
or
f
ur
th
e
r
s
tu
dy
to
im
pr
ove
s
to
c
k m
a
r
ke
t
li
qui
di
ty
pr
e
di
c
ti
on i
n de
ve
lo
pi
ng ma
r
ke
ts
.
2.
M
E
T
H
O
D
O
L
O
G
Y
T
hi
s
s
e
c
ti
on
de
li
ne
a
te
s
a
c
om
pr
e
he
n
s
iv
e
gui
de
li
ne
f
or
th
e
m
e
th
odol
ogi
c
a
l
pha
s
e
s
of
th
e
s
tu
dy,
s
ta
r
ti
ng
f
r
om
da
ta
ba
s
e
c
ol
le
c
ti
on
a
nd
pr
e
-
pr
oc
e
s
s
in
g
,
f
ol
lo
w
e
d
by
m
ode
l’
s
a
r
c
hi
te
c
tu
r
e
s
,
tr
a
in
in
g,
te
s
ti
ng
a
n
d
hype
r
pa
r
a
m
e
te
r
s
f
in
e
-
tu
ni
ng.
2
.1.
D
at
as
e
t
d
e
s
c
r
ip
t
io
n
T
he
da
ta
s
e
t
ut
il
iz
e
d
in
th
is
s
tu
dy
is
publ
ic
ly
a
c
c
e
s
s
ib
le
on
th
e
of
f
ic
ia
l
w
e
bs
it
e
of
th
e
C
S
E
,
c
ove
r
in
g
th
e
m
os
t
r
e
c
e
nt
5
ye
a
r
s
pe
r
io
d
[
23]
.
T
a
bl
e
1
pr
e
s
e
nt
s
a
s
a
m
pl
e
of
th
e
f
e
a
tu
r
e
s
e
xt
r
a
c
te
d
f
r
om
th
e
da
ta
s
e
t.
T
he
f
e
a
tu
r
e
na
m
e
s
ha
ve
be
e
n t
r
a
ns
la
te
d f
r
om
F
r
e
nc
h i
nt
o E
ngl
is
h f
o
r
c
la
r
it
y. B
e
f
or
e
m
ovi
ng t
o t
he
ne
xt
s
te
p i
n t
he
da
ta
s
e
t
pr
oc
e
s
s
in
g,
it
is
im
por
ta
nt
to
hi
ghl
ig
ht
th
a
t
in
a
dd
it
io
n
to
th
e
de
s
c
r
ib
e
d
f
e
a
tu
r
e
s
w
e
c
a
lc
ul
a
te
d
a
k
e
y
f
e
a
tu
r
e
na
m
e
d
th
e
“
B
id
-
A
s
k
-
S
pr
e
a
d”
,
w
hi
c
h
r
e
pr
e
s
e
nt
s
a
ke
y
m
e
a
s
ur
e
u
s
e
d
to
g
a
uge
tr
a
di
ng
f
r
ic
ti
ons
,
w
he
r
e
a
la
r
ge
r
s
pr
e
a
d
in
di
c
a
te
s
lo
w
e
r
li
qui
di
ty
a
nd
hi
ghe
r
im
pl
ic
it
c
os
ts
f
or
tr
a
de
r
s
by
le
ve
r
a
gi
ng
c
om
pl
e
te
ope
n,
hi
gh,
lo
w
,
a
nd
c
lo
s
e
pr
ic
e
da
ta
w
hi
c
h
e
ns
ur
e
m
or
e
a
c
c
ur
a
te
s
pr
e
a
d
e
s
ti
m
a
te
s
,
e
s
s
e
nt
ia
l
f
or
e
m
pi
r
ic
a
l
f
in
a
nc
e
a
nd
pr
a
c
ti
c
a
l
a
ppl
ic
a
ti
ons
in
a
s
s
e
t
pr
ic
in
g
a
nd
r
e
gul
a
to
r
y
a
na
l
ys
is
,
pr
ovi
di
ng
a
c
om
pr
e
he
ns
iv
e
vi
e
w
of
th
e
s
it
ua
ti
on [
24]
, [
25]
. T
a
bl
e
2 s
na
ps
hot
t
he
da
ta
s
e
t
f
or
m
a
t.
A
f
te
r
w
a
r
ds
,
a
nd
to
e
ns
ur
e
th
e
m
ode
ls
r
e
li
a
bi
li
ty
,
w
e
d
e
vot
e
d
8
0%
of
da
ta
f
or
tr
a
in
in
g
a
nd
20%
w
a
s
e
ve
nl
y
di
vi
de
d
be
twe
e
n
te
s
ti
ng
a
nd
va
li
da
ti
on.
T
h
e
tr
a
in
in
g
s
e
que
nc
e
s
s
uppor
ts
c
om
pr
e
he
ns
iv
e
le
a
r
ni
ng,
w
hi
le
th
e
va
li
da
ti
on
s
e
t
e
n
a
bl
e
s
f
in
e
-
tu
ni
ng.
T
h
e
s
e
p
a
r
a
te
te
s
ti
ng
s
e
t,
pr
ovi
de
s
a
n
obj
e
c
ti
ve
m
e
a
s
ur
e
of
ge
ne
r
a
li
z
a
ti
on t
o uns
e
e
n d
a
ta
.
2
.
2
.
M
od
e
ls
s
e
le
c
t
io
n
I
n
th
is
s
tu
dy,
w
e
e
xa
m
in
e
m
ul
ti
pl
e
de
e
p
le
a
r
ni
ng
a
r
c
hi
te
c
tu
r
e
s
to
e
f
f
e
c
ti
ve
ly
c
a
pt
ur
e
bot
h
lo
c
a
l
a
nd
lo
ng
-
te
r
m
de
pe
nde
nc
ie
s
pr
e
s
e
nt
i
n t
he
s
to
c
k m
a
r
ke
t
da
ta
. T
h
e
s
e
le
c
ti
on of
e
a
c
h m
ode
l
a
r
c
hi
te
c
tu
r
e
w
a
s
dr
iv
e
n
by
it
s
a
bi
li
ty
to
ha
ndl
e
di
f
f
e
r
e
nt
f
a
c
e
ts
of
f
in
a
nc
ia
l
da
ta
.
T
he
pa
r
a
m
e
te
r
iz
a
ti
on
of
e
a
c
h
m
ode
l
w
a
s
a
ls
o
c
a
r
e
f
ul
ly
c
ons
id
e
r
e
d.
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
. 14, No. 5, O
c
to
be
r
2025
:
3624
-
3633
3626
T
a
bl
e
1.
D
a
ta
s
e
t
de
f
a
ul
t
f
e
a
tu
r
e
s
F
e
a
t
ur
e
T
r
a
ns
l
a
t
i
on
D
e
s
c
r
i
pt
i
on
S
é
a
nc
e
S
e
s
s
i
on da
t
e
T
he
da
t
e
of
t
he
s
e
s
s
i
on
V
a
l
e
ur
i
ndi
c
e
I
nde
x va
l
ue
S
e
s
s
i
on s
t
oc
k i
nde
x v
a
l
ue
P
l
us
ha
ut
H
i
ghe
s
t
va
l
ue
H
i
ghe
s
t
s
e
s
s
i
on i
nde
x va
l
ue
P
l
us
ba
s
L
ow
e
s
t
va
l
u
e
L
ow
e
s
t
s
e
s
s
i
on i
nde
x va
l
ue
V
a
r
i
a
t
i
on
ve
i
l
l
e
P
r
e
vi
ous
da
y
c
ha
nge
T
he
pe
r
c
e
nt
a
ge
c
ha
nge
i
n t
he
i
nde
x va
l
ue
c
om
pa
r
e
d t
o t
he
pr
e
vi
ous
da
y
V
a
r
i
a
t
i
on
31/
12
Y
e
a
r
t
o da
t
e
c
ha
nge
T
he
ye
a
r
-
to
-
da
t
e
pe
r
c
e
nt
a
ge
c
ha
nge
i
n t
he
i
nde
x va
l
ue
,
pr
e
s
um
a
bl
y m
e
a
s
ur
e
d f
r
om
D
e
c
e
m
be
r
31s
t
of
t
he
pr
e
vi
ous
ye
a
r
T
a
bl
e
2
.
O
ve
r
vi
e
w
of
t
he
C
S
E
da
ta
s
e
t
f
e
a
tu
r
e
s
f
or
m
a
t
S
é
a
nc
e
V
a
l
e
ur
i
ndi
c
e
P
l
us
ha
ut
P
l
us
ba
s
V
a
r
i
a
t
i
on ve
i
l
l
e
V
a
r
i
a
t
i
on 31/
12
B
i
d
-
A
s
k S
pr
e
a
d
2021
-
01
-
04
10317.51
1031.75
10225.50
0.90
0.90
92.25
2021
-
01
-
05
10261.66
10323.29
10261.66
-
0.54
0.35
61.63
2021
-
01
-
06
10233.11
10304.18
10219.33
-
0.28
0.07
84.85
2021
-
01
-
07
10288.08
10329.4
10233.11
0.54
0.61
96.83
2021
-
01
-
08
10269.53
10311.30
10266.89
-
0.18
0.43
44.41
2.
2
.1. C
on
vol
u
t
io
n
al
n
e
u
r
al
n
e
t
w
or
k
s
F
ig
ur
e
1
il
lu
s
tr
a
te
s
th
e
a
r
c
hi
te
c
tu
r
e
of
C
N
N
,
a
c
la
s
s
of
de
e
p
l
e
a
r
ni
ng,
s
pe
c
if
ic
a
ll
y
ne
ur
a
l
ne
twor
ks
,
de
s
ig
ne
d
f
or
im
a
ge
pr
oc
e
s
s
in
g
a
nd c
om
put
e
r
vi
s
io
n
ta
s
ks
de
s
ig
ne
d
f
or
im
a
ge
pr
oc
e
s
s
in
g
a
nd
c
om
put
e
r
vi
s
io
n
ta
s
ks
.
T
he
s
e
ne
twor
ks
a
r
e
p
a
r
ti
c
ul
a
r
ly
a
de
pt
a
t
r
e
c
ogni
z
in
g
s
p
a
ti
a
l
pa
tt
e
r
ns
w
it
hi
n
gr
id
-
li
ke
da
ta
,
w
hi
c
h
w
a
s
a
da
pt
e
d t
o ha
ndl
e
t
im
e
-
s
e
r
ie
s
na
tu
r
e
f
or
s
to
c
k m
a
r
ke
t
p
r
e
di
c
ti
o
ns
by c
a
pt
ur
in
g s
hor
t
-
te
r
m
pa
tt
e
r
ns
i
n f
in
a
nc
ia
l
da
ta
s
e
que
n
c
e
s
[
26]
,
[
27]
.
T
he
a
r
c
hi
te
c
tu
r
e
of
th
e
C
N
N
is
c
om
pos
e
d
of
th
r
e
e
m
a
in
la
ye
r
s
:
c
onvolut
io
na
l,
pool
in
g,
a
nd
f
ul
ly
c
onne
c
te
d
l
a
ye
r
s
.
C
onvolut
io
na
l
l
a
ye
r
s
c
on
s
ti
tu
te
th
e
f
unda
m
e
nt
a
l
f
r
a
m
e
w
or
k
of
C
N
N
s
.
F
il
te
r
s
,
a
ls
o
known
a
s
ke
r
ne
ls
,
a
r
e
a
ppl
ie
d
a
s
th
e
y
tr
a
ve
r
s
e
th
e
in
put
da
ta
,
r
e
s
ul
ti
ng
in
th
e
c
r
e
a
ti
on
of
f
e
a
tu
r
e
m
a
ps
,
w
hi
c
h
e
m
pha
s
iz
e
s
ig
ni
f
ic
a
nt
lo
c
a
l
pa
tt
e
r
ns
in
th
e
da
t
a
,
s
uc
h
a
s
s
w
if
t
c
ha
nge
s
in
s
to
c
k
pr
ic
e
s
,
by
c
onc
e
nt
r
a
ti
ng
on
s
pa
ti
a
ll
y
a
dj
a
c
e
nt
d
a
ta
poi
nt
s
.
I
n
th
is
m
ode
l,
th
e
f
il
te
r
s
iz
e
w
a
s
c
hos
e
n
th
r
ough
e
m
pi
r
ic
a
l
tr
ia
ls
, s
tr
ik
in
g a
ba
la
nc
e
be
twe
e
n
c
a
pt
ur
in
g r
e
le
va
nt
pa
tt
e
r
ns
a
n
d m
a
in
ta
in
in
g c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y, w
hi
c
h
is
e
s
s
e
nt
ia
l
to
pr
e
s
e
r
ve
im
por
ta
nt
f
e
a
tu
r
e
s
on
th
e
da
t
a
s
e
t.
F
in
a
ll
y,
th
e
f
ul
ly
c
onne
c
te
d
la
ye
r
s
c
on
s
ol
id
a
te
in
put
f
r
om
t
he
c
onvolut
io
na
l
a
nd pooli
ng l
a
ye
r
s
t
o pr
oduc
e
a
s
in
gl
e
n
e
ur
on t
ha
t
f
or
e
c
a
s
ts
s
to
c
k m
a
r
ke
t
li
qui
di
ty
.
F
ig
ur
e
1
.
T
he
C
N
N
a
r
c
hi
te
c
tu
r
e
T
he
C
N
N
de
s
ig
ne
d
in
th
is
s
t
ud
y,
in
it
ia
te
s
w
i
th
a
C
o
nv
1D
la
ye
r
th
a
t
i
nc
lu
d
e
s
6
4
f
il
te
r
s
a
nd
a
ke
r
ne
l
s
iz
e
of
3,
a
pp
ly
i
ng
th
e
r
e
c
ti
f
ie
d
li
ne
a
r
un
it
(
R
e
L
U
)
a
c
ti
va
t
io
n
f
unc
ti
on.
T
hi
s
c
on
vol
ut
io
na
l
la
y
e
r
de
te
c
ts
lo
c
a
l
pa
t
te
r
ns
i
n
t
he
in
pu
t
d
a
ta
t
hr
ou
gh
th
e
a
p
pl
ic
a
ti
on
o
f
f
i
lt
e
r
s
a
c
r
os
s
th
e
t
im
e
s
te
ps
,
a
ls
o
c
a
pt
ur
in
g
s
ig
ni
f
ic
a
n
t
s
ho
r
t
-
t
e
r
m
t
r
e
n
ds
.
N
e
x
t,
a
M
a
x
P
o
ol
in
g
1
D
l
a
ye
r
is
i
m
pl
e
m
e
nt
e
d
w
i
th
a
po
ol
s
iz
e
of
2
,
w
hi
c
h
e
f
f
e
c
t
iv
e
ly
r
e
du
c
e
s
t
he
di
m
e
ns
io
na
l
it
y
o
f
th
e
f
e
a
t
ur
e
m
a
ps
t
hr
ou
gh
dow
n
-
s
a
m
pl
in
g
,
th
us
p
r
e
s
e
r
v
in
g
e
s
s
e
nt
ia
l
f
e
a
t
u
r
e
s
w
hi
le
m
in
im
iz
i
ng
c
o
m
p
ut
a
ti
ona
l
c
om
pl
e
xi
t
y.
T
he
o
ut
p
ut
f
r
om
t
he
po
ol
in
g
la
ye
r
is
s
ubs
e
q
ue
n
tl
y
f
la
t
te
n
e
d,
tr
a
ns
f
o
r
m
in
g
th
e
1D
f
e
a
t
ur
e
m
a
ps
in
t
o
a
s
i
ngu
la
r
ve
c
to
r
s
u
it
a
bl
e
f
o
r
i
npu
t
i
nt
o
f
u
ll
y
c
onn
e
c
te
d
la
ye
r
s
.
S
ubs
e
que
nt
ly
,
a
de
ns
e
la
ye
r
c
o
m
p
r
is
in
g
5
0
ne
u
r
ons
w
it
h
R
e
L
U
a
c
ti
va
t
io
n
f
a
c
il
it
a
te
s
th
e
m
ode
l
'
s
a
bi
li
ty
to
c
o
m
p
r
e
h
e
nd
i
nt
r
ic
a
t
e
i
nt
e
r
a
c
t
io
ns
a
m
on
g
f
e
a
tu
r
e
s
.
T
he
c
on
c
l
udi
ng
l
a
ye
r
is
a
de
ns
e
o
ut
pu
t
la
ye
r
f
e
a
tu
r
i
ng
a
s
i
ng
le
ne
u
r
o
n,
w
h
ic
h
de
li
v
e
r
s
t
he
m
ode
l
'
s
pr
e
di
c
t
io
n
f
o
r
r
e
g
r
e
s
s
io
n
ta
s
ks
.
T
he
m
od
e
l
ut
i
li
z
e
s
th
e
A
da
m
op
ti
m
iz
e
r
to
f
a
c
i
li
ta
te
e
f
f
e
c
t
iv
e
l
e
a
r
ni
ng
vi
a
a
da
pt
iv
e
g
r
a
d
ie
n
t
e
s
ti
m
a
ti
o
n,
w
h
il
e
M
S
E
is
s
e
le
c
t
e
d
a
s
t
he
lo
s
s
f
unc
ti
on
to
e
m
pha
s
iz
e
p
r
e
c
is
e
p
r
e
di
c
ti
ons
by
im
pos
in
g
pe
na
l
ti
e
s
o
n
la
r
ge
r
e
r
r
o
r
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
m
ar
k
e
t
li
qui
di
ty
:
hy
br
id
de
e
p l
e
ar
ni
ng app
r
oac
he
s
f
or
pr
e
di
c
ti
on
…
(
M
ar
ia
m
A
it
A
l
)
3627
F
u
r
t
he
r
m
o
r
e
,
th
e
M
A
E
is
in
c
or
po
r
a
te
d
a
s
a
pe
r
f
o
r
m
a
nc
e
m
e
tr
ic
to
e
n
ha
nc
e
in
t
e
r
pr
e
ta
bi
li
ty
r
e
ga
r
d
in
g
t
he
a
ve
r
a
ge
de
v
ia
ti
on
f
r
o
m
a
c
tu
a
l
va
lu
e
s
.
2.
2
.
2
.
L
on
g
s
h
or
t
-
t
e
r
m
m
e
m
o
r
y
U
nl
ik
e
to
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
(
R
N
N
)
,
w
hi
c
h
e
xpe
r
ie
nc
e
c
ha
ll
e
nge
s
w
it
h
th
e
v
a
ni
s
hi
ng
gr
a
di
e
nt
pr
obl
e
m
,
L
S
T
M
ne
twor
ks
a
r
e
s
pe
c
if
ic
a
ll
y
de
s
ig
ne
d
a
s
a
va
r
ia
nt
of
R
N
N
s
to
e
f
f
e
c
ti
ve
ly
c
a
pt
ur
e
lo
ng
-
te
r
m
de
pe
nde
nc
ie
s
in
s
e
que
nt
ia
l
da
t
a
.
F
ig
ur
e
2
por
tr
a
ys
th
e
a
r
c
hi
te
c
tu
r
e
of
th
is
c
a
te
gor
y
of
ne
ur
a
l
ne
twor
ks
th
a
t
le
ve
r
a
ge
m
e
m
or
y
c
e
ll
s
to
pr
e
s
e
r
ve
c
r
uc
ia
l
in
f
or
m
a
ti
on
a
c
r
os
s
pr
ol
onge
d
s
e
qu
e
nc
e
s
,
r
e
nde
r
in
g
th
e
m
a
ppr
opr
ia
te
f
or
ti
m
e
-
s
e
r
ie
s
da
ta
.
T
he
L
S
T
M
c
e
ll
c
ons
is
te
nt
ly
in
vol
ve
s
th
r
e
e
f
unda
m
e
nt
a
l
g
a
te
s
:
in
put
,
f
or
ge
t,
a
nd output
. T
he
i
nput
ga
te
de
te
r
m
in
e
s
t
he
in
f
or
m
a
ti
on
t
o s
to
r
e
,
th
e
f
or
ge
t
ga
te
i
de
nt
i
f
ie
s
w
hi
c
h i
nf
o
r
m
a
ti
on t
o
e
li
m
in
a
te
,
a
nd
th
e
out
put
ga
te
r
e
gul
a
te
s
th
e
f
in
a
l
out
put
a
c
c
or
di
ng
to
th
e
c
ur
r
e
nt
c
e
ll
s
ta
te
a
nd
pr
io
r
out
put
s
[
28]
–
[
30]
.
T
he
a
r
c
hi
te
c
tu
r
e
of
th
e
s
tu
dy
s
e
ts
up
w
it
h
a
n
L
S
T
M
la
y
e
r
c
om
pr
is
in
g
100
uni
ts
,
ut
il
iz
in
g
th
e
de
f
a
ul
t
r
e
tu
r
n
s
e
que
nc
e
s
pa
r
a
m
e
te
r
.
T
hi
s
c
onf
ig
ur
a
ti
on
pe
r
m
it
s
th
e
m
ode
l
to
pr
oduc
e
th
e
c
om
pl
e
te
s
e
que
nc
e
of
hi
dde
n
s
ta
te
s
,
w
hi
c
h
s
im
pl
if
ie
s
th
e
s
ubs
e
que
nt
L
S
T
M
la
y
e
r
'
s
a
c
c
e
s
s
to
s
e
que
nt
ia
l
in
f
or
m
a
ti
on,
w
hi
c
h
is
im
por
ta
nt
f
or
a
c
qui
r
in
g
in
tr
ic
a
te
te
m
por
a
l
de
pe
nde
n
c
ie
s
a
c
r
os
s
va
r
io
us
ti
m
e
s
t
e
ps
,
a
nd
to
f
ur
th
e
r
e
va
lu
a
te
th
is
in
f
or
m
a
ti
on
a
nd
ge
ne
r
a
te
a
f
in
a
l
hi
dde
n
s
ta
te
th
a
t
e
nc
a
ps
ul
a
te
s
t
he
le
a
r
ne
d
te
m
por
a
l
f
e
a
tu
r
e
s
up
to
th
e
c
ur
r
e
nt
ti
m
e
s
te
p,
a
n
a
ddi
ti
ona
l
L
S
T
M
l
a
ye
r
is
c
on
s
tr
uc
te
d.
T
hi
s
la
y
e
r
c
ons
is
ts
of
one
hundr
e
d
uni
ts
a
nd
doe
s
not
e
nga
ge
th
e
r
e
tu
r
n
s
e
qu
e
nc
e
s
pa
r
a
m
e
te
r
.
S
ubs
e
que
nt
ly
,
th
e
out
put
f
r
om
th
e
L
S
T
M
la
ye
r
s
is
s
e
nt
to
a
D
e
n
s
e
la
ye
r
c
ons
is
ti
ng
of
a
s
ol
it
a
r
y
ne
ur
on.
T
hi
s
la
ye
r
ge
n
e
r
a
te
s
th
e
ul
ti
m
a
te
pr
e
di
c
ti
on
f
or
r
e
gr
e
s
s
io
n
pr
obl
e
m
s
.
T
hi
s
m
ode
l
w
a
s
c
on
s
tr
uc
te
d
w
it
h
th
e
A
da
m
opt
im
iz
e
r
,
s
e
le
c
te
d
f
or
it
s
c
us
to
m
iz
a
bl
e
le
a
r
ni
ng
r
a
te
a
tt
r
ib
ut
e
s
.
T
he
s
e
c
h
a
r
a
c
te
r
is
ti
c
s
r
e
s
ul
t
in
tr
a
in
in
g
m
e
th
odol
ogi
e
s
th
a
t
a
r
e
bot
h
e
f
f
e
c
ti
ve
a
nd
r
e
li
a
bl
e
.
F
ur
th
e
r
m
or
e
,
th
e
M
S
E
s
e
r
ve
s
a
s
th
e
lo
s
s
f
unc
ti
on
to
in
f
li
c
t
a
gr
e
a
te
r
pe
na
lt
y
on
s
ubs
ta
nt
ia
l
pr
e
di
c
ti
on
e
r
r
or
s
,
he
nc
e
e
nha
nc
in
g
th
e
a
c
c
ur
a
c
y
of
s
to
c
k
m
a
r
ke
t
pr
e
di
c
ti
ons
.
T
he
M
A
E
is
a
pe
r
f
or
m
a
nc
e
m
e
tr
ic
th
a
t
e
lu
c
id
a
te
s
a
ve
r
a
ge
de
vi
a
ti
ons
, he
nc
e
i
m
pr
ovi
ng c
om
pr
e
he
ns
io
n of
t
he
m
ode
l'
s
pr
e
di
c
ti
ve
e
f
f
ic
a
c
y.
F
ig
ur
e
2
.
T
he
L
S
T
M
ne
twor
k
a
r
c
hi
te
c
tu
r
e
2.
2
.
3
.
B
id
ir
e
c
t
io
n
al
lo
n
g s
h
or
t
-
t
e
r
m
m
e
m
or
y
B
iL
S
T
M
ne
twor
ks
a
r
e
a
n
a
dv
a
nc
e
d
f
or
m
of
L
S
T
M
de
s
ig
ne
d
t
o
id
e
nt
if
y
pa
tt
e
r
ns
in
s
e
que
nt
ia
l
da
ta
by
pr
oc
e
s
s
in
g
it
in
f
or
w
a
r
d
a
s
w
e
ll
a
s
ba
c
kw
a
r
d
di
r
e
c
ti
ons
.
T
h
is
bi
di
r
e
c
ti
ona
l
pr
oc
e
s
s
in
g
pe
r
m
it
s
L
S
T
M
s
to
a
c
qui
r
e
m
or
e
e
xt
e
n
s
iv
e
t
e
m
por
a
l
pa
tt
e
r
ns
,
r
e
nde
r
in
g
th
e
m
e
s
pe
c
ia
ll
y
us
e
f
ul
in
s
it
ua
ti
ons
w
he
r
e
bot
h
pa
s
t
a
nd
f
ut
ur
e
c
ont
e
xt
s
im
pr
ove
pr
e
di
c
ti
on
a
c
c
ur
a
c
y
f
or
s
to
c
k
m
a
r
ke
t
f
or
e
c
a
s
ti
ng
a
s
il
lu
s
tr
a
te
d
in
F
ig
ur
e
3.
T
he
y
a
r
e
pa
r
ti
c
ul
a
r
ly
e
f
f
e
c
ti
ve
,
be
c
a
u
s
e
th
e
y
c
a
pt
ur
e
de
p
e
nde
nc
ie
s
t
ha
t
m
ig
ht
not
be
a
ppa
r
e
nt
w
he
n
onl
y
pa
s
t
in
f
or
m
a
ti
on
is
c
ons
id
e
r
e
d
[
30]
–
[
32]
.
T
o
c
r
e
a
te
th
is
m
ode
l,
w
e
r
e
ta
in
e
d
th
e
L
S
T
M
s
e
tu
p
out
li
ne
d
in
th
e
pr
e
c
e
di
ng s
e
c
ti
on a
nd j
u
s
t
tr
ig
ge
r
e
d t
he
bi
di
r
e
c
ti
ona
l
pr
oc
e
s
s
.
2
.
2
.
4
.
CNN
-
L
S
T
M
an
d
C
N
N
-
B
iL
S
T
M
T
he
hybr
id
C
N
N
-
L
S
T
M
or
C
N
N
-
B
iL
S
T
M
m
ode
ls
c
om
bi
ne
s
th
e
a
dva
nt
a
ge
s
of
C
N
N
a
nd
L
S
T
M
/B
iL
S
T
M
w
hi
le
a
dhe
r
in
g
to
th
e
c
onf
ig
ur
a
ti
on
de
s
c
r
ib
e
d
in
th
e
pr
e
vi
ous
s
e
c
ti
on.
B
ot
h
a
ppr
oa
c
he
s
us
e
C
N
N
f
or
s
pa
ti
a
l
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
a
nd
L
S
T
M
f
or
te
m
por
a
l
l
e
a
r
ni
ng.
H
ow
e
ve
r
,
th
e
C
N
N
-
bi
L
S
T
M
m
ode
l
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
. 14, No. 5, O
c
to
be
r
2025
:
3624
-
3633
3628
of
f
e
r
s
a
s
e
c
ond vie
w
poi
nt
on s
e
que
nc
e
da
ta
, pos
s
ib
ly
e
nha
nc
in
g a
c
c
ur
a
c
y i
n s
e
tt
in
gs
w
he
r
e
c
om
pr
e
he
ns
io
n of
bot
h pa
s
t
a
nd f
ut
ur
e
i
s
be
ne
f
ic
ia
l.
F
ig
ur
e
3
.
T
he
Bi
L
S
T
M
a
r
c
hi
te
c
tu
r
e
2
.
3
.
H
yp
e
r
p
ar
am
e
t
e
r
t
u
n
in
g an
d
op
t
im
iz
a
t
io
n
O
ve
r
th
e
tr
a
in
in
g
pr
oc
e
s
s
,
w
e
opt
im
iz
e
d
hype
r
pa
r
a
m
e
te
r
s
in
a
c
c
or
da
nc
e
w
it
h
de
e
p
le
a
r
ni
ng
pr
in
c
ip
le
s
,
pe
r
f
or
m
in
g
r
e
pe
a
te
d
e
xpe
r
im
e
nt
s
to
e
nha
n
c
e
pe
r
f
or
m
a
nc
e
by
id
e
nt
if
yi
ng
th
e
id
e
a
l
c
onf
ig
ur
a
ti
ons
f
or
ba
tc
h
s
iz
e
,
dr
opout
r
a
te
s
,
a
nd
le
a
r
ni
ng
r
a
te
s
,
th
e
r
e
f
or
e
ba
la
nc
in
g
c
om
put
in
g
e
f
f
ic
ie
nc
y
w
it
h
m
ode
l
e
f
f
ic
a
c
y
.
T
he
ba
tc
h
s
iz
e
w
a
s
s
e
t
to
30,
ba
la
nc
in
g
c
om
put
a
t
io
na
l
e
f
f
ic
ie
nc
y
w
it
h
th
e
m
ode
l'
s
a
bi
li
ty
to
ge
ne
r
a
li
z
e
.
W
e
f
ix
e
d
th
e
num
b
e
r
of
e
poc
hs
a
t
100,
pr
ovi
di
ng
e
nough
it
e
r
a
ti
ons
f
or
th
e
m
ode
l
to
th
or
oughly
le
a
r
n
th
e
unde
r
ly
in
g
da
ta
pa
tt
e
r
ns
w
it
hout
ove
r
f
it
ti
ng
.
T
he
hype
r
pa
r
a
m
e
te
r
s
w
e
r
e
f
in
e
-
tu
ne
d
th
r
ough
m
ul
ti
pl
e
it
e
r
a
ti
ons
,
te
s
ti
ng
va
r
io
us
c
onf
ig
ur
a
ti
ons
to
e
ns
ur
e
opt
im
a
l
pe
r
f
or
m
a
nc
e
a
c
r
os
s
bot
h
tr
a
in
in
g
a
nd
va
li
da
ti
on
da
ta
s
e
ts
.
T
hi
s
pr
oc
e
s
s
in
vol
ve
d
c
ont
in
uous
m
oni
to
r
in
g
of
not
ju
s
t
th
e
f
in
a
l
pr
e
di
c
ti
on
a
c
c
ur
a
c
y
but
a
l
s
o
th
e
m
ode
l'
s
c
a
pa
c
it
y
to
g
e
ne
r
a
li
z
e
w
e
ll
to
unobs
e
r
ve
d
d
a
ta
,
th
us
r
e
duc
in
g
th
e
r
is
k
of
ove
r
f
it
ti
ng.
T
he
m
a
nua
l
it
e
r
a
ti
on
te
c
hni
que
f
a
c
il
it
a
te
d
m
ode
s
t
m
odi
f
ic
a
ti
ons
to
hyp
e
r
pa
r
a
m
e
te
r
s
,
a
ll
ow
in
g
a
c
onc
e
nt
r
a
te
d
a
na
ly
s
is
of
th
e
im
pa
c
t
of
e
a
c
h
c
onf
ig
ur
a
ti
on
on
m
ode
l
c
or
r
e
c
tn
e
s
s
a
nd
s
ta
bi
li
ty
.
T
hi
s
te
c
hni
que
ha
s
in
he
r
e
nt
dr
a
w
ba
c
ks
,
e
s
pe
c
ia
ll
y
th
e
ti
m
e
-
c
on
s
um
in
g
pr
oc
e
s
s
of
m
a
nua
ll
y
a
dj
us
ti
ng
s
e
ve
r
a
l
pa
r
a
m
e
te
r
s
.
M
or
e
ove
r
,
m
a
nua
l
tu
ni
ng
m
a
y
f
a
il
to
id
e
nt
if
y
th
e
de
f
in
it
iv
e
id
e
a
l
s
e
tt
in
gs
due
to
hum
a
n
li
m
it
a
ti
ons
in
m
e
th
odi
c
a
ll
y
e
x
a
m
in
in
g
th
e
w
hol
e
hype
r
pa
r
a
m
e
te
r
s
p
a
c
e
.
F
or
e
va
lu
a
ti
ng
m
ode
l
pe
r
f
or
m
a
nc
e
dur
in
g
tr
a
in
in
g,
M
S
E
w
a
s
s
e
le
c
te
d
a
s
th
e
lo
s
s
f
unc
ti
on.
M
S
E
pe
na
li
z
e
s
la
r
ge
r
e
r
r
or
s
m
or
e
he
a
vi
ly
,
m
a
ki
ng
it
a
n
id
e
a
l
m
e
tr
ic
f
or
r
e
gr
e
s
s
io
n
pr
obl
e
m
s
li
ke
s
to
c
k
m
a
r
ke
t
pr
e
di
c
ti
on,
w
h
e
r
e
pr
e
c
is
e
f
or
e
c
a
s
ti
ng
i
s
c
r
uc
ia
l.
B
y
f
oc
us
in
g
on
m
in
im
iz
in
g
s
ig
ni
f
ic
a
nt
de
vi
a
ti
ons
be
twe
e
n
pr
e
di
c
te
d
a
nd
a
c
tu
a
l
va
lu
e
s
,
M
S
E
he
lp
s
im
pr
ove
th
e
m
ode
l'
s
a
c
c
ur
a
c
y
a
nd
r
e
li
a
bi
li
ty
in
r
e
a
l
-
w
or
ld
a
ppl
ic
a
ti
ons
.
T
he
s
e
le
c
ti
on
of
M
S
E
a
ls
o
e
ns
ur
e
d
th
a
t
th
e
m
ode
l
w
oul
d
pr
io
r
it
iz
e
r
e
duc
in
g
la
r
ge
r
e
r
r
or
s
, e
nha
nc
in
g i
ts
r
obus
tn
e
s
s
i
n pr
e
di
c
ti
ng s
to
c
k m
a
r
ke
t
li
qui
di
ty
a
c
c
ur
a
te
ly
.
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
T
hi
s
s
e
c
ti
on
pr
e
s
e
nt
s
a
c
om
pa
r
a
ti
ve
p
e
r
f
or
m
a
nc
e
a
na
ly
s
i
s
of
th
e
pr
opos
e
d
de
e
p
le
a
r
ni
ng
a
r
c
hi
te
c
tu
r
e
s
,
in
c
lu
di
ng
C
N
N
,
L
S
T
M
,
B
iL
S
T
M
,
a
nd
hybr
id
C
N
N
-
L
S
T
M
a
nd
C
N
N
-
B
iL
S
T
M
m
ode
ls
,
in
f
or
e
c
a
s
ti
ng s
to
c
k m
a
r
ke
t
li
qui
di
ty
w
it
hi
n t
he
vol
a
ti
le
c
ont
e
xt
o
f
th
e
C
S
E
. O
ur
f
in
di
ngs
va
li
da
te
pr
io
r
r
e
s
e
a
r
c
h,
de
m
ons
tr
a
ti
ng
th
a
t
C
N
N
-
B
iL
S
T
M
m
ode
ls
e
xc
e
l
in
e
nvi
r
onm
e
nt
s
r
e
qui
r
in
g
th
e
id
e
nt
i
f
ic
a
ti
on
of
bo
th
s
pa
ti
a
l
a
nd
te
m
por
a
l
pa
tt
e
r
ns
,
s
uc
h
a
s
f
lu
c
tu
a
ti
ng
f
in
a
nc
ia
l
m
a
r
ke
ts
.
T
he
e
xi
s
ti
ng
li
te
r
a
tu
r
e
hi
ghl
ig
ht
s
a
not
a
bl
e
li
m
it
a
ti
on
in
th
e
ut
il
iz
a
ti
on
o
f
de
e
p
le
a
r
ni
ng
f
or
pr
e
di
c
ti
ng
vol
a
ti
le
s
to
c
k
m
a
r
ke
ts
.
T
hi
s
s
tu
dy
a
ddr
e
s
s
e
s
th
is
ga
p
by
de
m
ons
tr
a
ti
ng
th
a
t
th
e
C
N
N
-
B
iL
S
T
M
m
ode
l
c
a
n e
f
f
e
c
ti
ve
ly
m
a
na
ge
th
e
di
s
ti
nc
t
c
ha
ll
e
ng
e
s
pr
e
s
e
nt
e
d
by
de
ve
lo
pi
ng
m
a
r
ke
ts
s
uc
h
a
s
M
or
oc
c
o'
s
,
w
hi
c
h
in
c
lu
de
r
a
pi
d
f
lu
c
tu
a
ti
ons
a
nd
unus
ua
l
m
a
r
ke
t
be
ha
vi
or
s
.
T
a
bl
e
3
s
um
m
a
r
iz
e
s
th
e
pe
r
f
or
m
a
nc
e
of
m
ode
ls
in
te
r
m
s
of
ke
y
e
r
r
or
m
e
tr
ic
s
(
M
S
E
,
R
M
S
E
,
a
nd
M
A
E
)
a
nd
F
ig
ur
e
s
4
to
8 de
m
ons
tr
a
te
s
t
he
l
e
a
r
ni
ng c
ur
ve
s
w
hi
c
h pr
ove
s
th
e
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
of
C
N
N
-
B
iL
S
T
M
a
nd
in
di
c
a
te
s
a
vi
a
bl
e
a
ve
nue
f
or
f
in
a
nc
ia
l
a
na
ly
s
ts
a
nd
in
ve
s
to
r
s
lo
oki
ng
f
or
de
pe
nda
bl
e
li
qui
di
ty
f
or
e
c
a
s
ts
in
de
ve
lo
pi
ng or
e
m
e
r
gi
ng ma
r
ke
ts
.
T
a
bl
e
3
.
P
e
r
f
or
m
a
nc
e
c
om
pa
r
is
on of
t
r
a
in
e
d m
ode
ls
M
ode
l
M
S
E
R
M
S
E
M
A
E
C
N
N
0
.014874
0.121957
0.092361
L
S
T
M
0.
005839
0.076415
0.050488
B
i
L
S
T
M
0.
005972
0.077280
0.048909
C
N
N
-
L
S
T
M
0.
004838
0.069556
0.044244
C
N
N
-
B
i
L
S
T
M
0.004742
0.068859
0.048855
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
m
ar
k
e
t
li
qui
di
ty
:
hy
br
id
de
e
p l
e
ar
ni
ng app
r
oac
he
s
f
or
pr
e
di
c
ti
on
…
(
M
ar
ia
m
A
it
A
l
)
3629
F
ig
ur
e
4
.
T
he
C
N
N
l
e
a
r
ni
ng c
ur
ve
F
ig
ur
e
5
.
T
he
L
S
T
M
l
e
a
r
ni
ng c
ur
ve
F
ig
ur
e
6
.
T
he
B
iL
S
T
M
l
e
a
r
ni
ng c
ur
ve
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
. 14, No. 5, O
c
to
be
r
2025
:
3624
-
3633
3630
F
ig
ur
e
7
.
T
he
C
N
N
-
L
S
T
M
l
e
a
r
ni
ng
c
ur
ve
F
ig
ur
e
8
.
T
he
C
N
N
-
B
iL
S
T
M
l
e
a
r
ni
ng c
ur
v
e
T
he
de
m
on
s
tr
a
te
d
a
c
c
ur
a
c
y
of
hybr
id
m
ode
ls
,
s
p
e
c
if
ic
a
ll
y
C
N
N
-
B
iL
S
T
M
,
s
ugge
s
ts
a
pr
om
is
in
g
di
r
e
c
ti
on
w
it
hi
n
de
ve
lo
pi
ng
m
a
r
ke
ts
.
T
he
hybr
id
de
e
p
le
a
r
ni
ng
m
ode
ls
c
a
n
in
f
or
m
in
ve
s
tm
e
nt
s
tr
a
te
gi
e
s
a
nd
r
e
gul
a
to
r
y
de
c
is
io
ns
,
e
ve
nt
ua
ll
y
im
pr
ovi
ng
m
a
r
ke
t
s
ta
bi
li
ty
a
nd
in
ve
s
to
r
c
onf
id
e
nc
e
on
th
e
m
a
r
ke
t
m
ove
m
e
nt
s
.
A
ddi
ti
ona
ll
y,
de
pl
oyi
ng
th
is
s
ol
ut
io
n
in
a
r
e
a
l
-
w
or
ld
e
nvi
r
onm
e
nt
w
il
l
f
a
c
il
it
a
te
be
tt
e
r
de
c
is
io
n
-
m
a
ki
ng,
e
s
pe
c
ia
ll
y
unde
r
uns
ta
bl
e
e
c
onomi
c
c
ondi
ti
ons
.
W
hi
le
th
is
s
tu
dy
of
f
e
r
s
va
lu
a
bl
e
in
s
ig
ht
s
,
s
e
ve
r
a
l
li
m
it
a
ti
ons
s
houl
d
be
not
e
d.
F
ir
s
t,
th
e
a
na
ly
s
is
f
oc
us
e
s
onl
y
o
n
one
s
to
c
k
e
xc
ha
nge
C
S
E
,
w
hi
c
h
m
a
y
li
m
it
how
w
e
ll
th
e
r
e
s
ul
ts
a
ppl
y
to
ot
he
r
de
ve
lo
pi
ng
m
a
r
ke
ts
w
it
h
di
f
f
e
r
e
nt
e
c
onomi
c
c
ondi
ti
ons
.
A
ddi
ti
ona
ll
y,
th
e
m
ode
l'
s
hype
r
pa
r
a
m
e
te
r
s
w
e
r
e
s
e
t
m
a
nua
ll
y,
w
hi
c
h
m
ig
ht
not
be
opt
im
a
l;
f
ut
ur
e
w
or
k
c
oul
d
im
pr
ove
th
is
by
us
in
g
a
ut
om
a
te
d
tu
ni
ng
m
e
th
ods
.
I
nc
lu
di
ng
e
xt
e
r
na
l
e
c
onomi
c
f
a
c
to
r
s
,
c
oul
d
a
ls
o
m
a
ke
th
e
m
ode
l
m
or
e
r
obus
t
a
nd be
tt
e
r
r
e
f
le
c
t
th
e
br
oa
de
r
e
c
onomi
c
e
nvi
r
onm
e
nt
a
f
f
e
c
ti
ng ma
r
ke
t
li
qui
di
ty
.
F
ig
ur
e
9 r
e
p
r
e
s
e
nt
s
t
he
us
e
of
th
e
C
N
N
-
B
iL
S
T
M
m
ode
l
to
f
or
e
c
a
s
t
s
to
c
k
m
a
r
ke
t
l
i
qui
di
ty
f
or
m
ont
hs
e
xc
lu
de
d
f
r
om
th
e
in
it
ia
l
da
ta
s
e
t,
w
hi
c
h
pr
ove
s
th
e
m
ode
l'
s
pr
e
di
c
ti
ve
pr
e
c
is
io
n
a
nd
de
pe
nda
bi
li
ty
in
pr
a
c
ti
c
a
l
c
ont
e
xt
s
.
T
he
di
s
c
r
e
pa
nc
ie
s
be
twe
e
n
th
e
a
c
tu
a
l
a
nd
pr
oj
e
c
te
d
va
lu
e
s
ove
r
th
is
le
ngt
hy
dur
a
ti
on
a
r
e
of
te
n
m
in
im
a
l,
s
ugge
s
ti
ng
th
a
t
th
e
m
ode
l
e
f
f
e
c
ti
ve
ly
e
xt
e
nd
s
it
s
a
ppl
ic
a
bi
li
ty
to
nove
l
da
ta
w
it
hout
m
uc
h
d
e
vi
a
ti
on.
T
hi
s
a
ppl
ic
a
ti
on
pr
ovi
de
s
a
va
lu
a
bl
e
c
om
pa
r
is
on
poi
nt
w
it
h
th
e
or
i
gi
na
l
da
ta
s
e
t'
s
r
e
s
id
ua
ls
,
s
how
in
g
a
s
im
il
a
r
ly
na
r
r
ow
s
pr
e
a
d
a
r
ound
z
e
r
o
w
hi
c
h
r
e
in
f
or
c
e
s
th
e
r
obus
tn
e
s
s
of
th
e
C
N
N
-
B
iL
S
T
M
m
ode
l,
de
m
ons
tr
a
ti
ng
it
s
pot
e
nt
ia
l
f
or
a
c
c
ur
a
te
f
or
e
c
a
s
ti
ng i
n vola
ti
le
a
nd dyna
m
ic
m
a
r
k
e
ts
.
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
m
ar
k
e
t
li
qui
di
ty
:
hy
br
id
de
e
p l
e
ar
ni
ng app
r
oac
he
s
f
or
pr
e
di
c
ti
on
…
(
M
ar
ia
m
A
it
A
l
)
3631
F
ig
ur
e
9
.
R
e
s
id
ua
l
p
lo
t
of
C
N
N
-
bi
L
S
T
M
p
r
e
di
c
ti
ons
4.
C
O
N
C
L
U
S
I
O
N
G
iv
e
n
th
e
a
c
a
de
m
ic
a
nd
pr
of
e
s
s
io
na
l
im
por
ta
nc
e
of
th
e
s
to
c
k
m
a
r
ke
t
li
qui
di
ty
,
e
s
pe
c
ia
ll
y
w
it
hi
n
e
m
e
r
gi
ng
c
ount
r
ie
s
,
w
e
tr
y
in
th
is
s
tu
dy
to
e
va
lu
a
te
a
nd
c
om
p
a
r
e
di
f
f
e
r
e
nt
ne
ur
a
l
ne
twor
ks
a
r
c
hi
te
c
tu
r
e
s
to
pr
e
di
c
t
th
e
f
ut
ur
e
le
ve
l
o
f
th
is
s
tr
a
te
gi
c
f
in
a
nc
ia
l
va
r
ia
bl
e
.
A
s
a
r
e
s
ul
t,
w
e
a
r
gue
th
a
t
th
e
hybr
id
de
e
p
le
a
r
ni
n
g
m
ode
ls
a
r
e
m
or
e
s
ui
ta
bl
e
,
e
s
pe
c
ia
ll
y
th
e
c
om
bi
na
ti
on
of
C
N
N
a
nd
B
i
-
L
S
T
M
w
it
hi
n
a
hybr
id
f
r
a
m
e
w
or
k
.
W
e
a
ppl
y
th
is
pr
om
is
in
g
a
ppr
oa
c
h
to
one
of
m
o
s
t
a
c
ti
ve
s
to
c
k
m
a
r
ke
ts
in
M
id
dl
e
E
a
s
t
a
nd
N
or
th
A
f
r
ic
a
(
M
E
N
A
)
r
e
gi
on
i.
e
.,
th
e
C
S
E
a
nd
pr
ovi
de
e
vi
d
e
nc
e
of
it
s
pe
r
f
or
m
a
nc
e
f
or
e
f
f
e
c
ti
ve
ly
ta
c
kl
in
g
th
is
pr
e
s
s
in
g
is
s
ue
.
D
e
s
pi
te
th
e
s
ig
ni
f
ic
a
nt
na
tu
r
e
of
th
e
r
e
s
ul
ts
,
th
is
s
tu
dy
s
houl
d
b
e
c
om
pl
e
te
d
a
nd
e
nr
ic
he
d.
I
n
th
is
c
ont
e
xt
,
th
e
pr
opos
e
d
a
ppr
oa
c
h
c
a
n
s
e
r
ve
a
s
a
ba
s
is
f
or
ongoing
th
e
or
e
ti
c
a
l
im
pr
ove
m
e
nt
a
nd
e
m
pi
r
ic
a
l
br
oa
de
r
in
ve
s
ti
ga
ti
on.
M
or
e
ove
r
,
a
nd
gi
ve
n
th
e
s
tr
a
te
gi
c
im
por
ta
nc
e
o
f
m
a
r
ke
t
li
qui
di
ty
p
r
e
di
c
ti
on
f
or
in
ve
s
to
r
s
,
it
w
oul
d
be
im
por
ta
nt
to
f
ur
th
e
r
e
xt
e
nd
a
nd
de
e
pe
n
th
e
s
tu
dy
to
pr
ovi
de
a
pr
a
c
ti
c
a
l
to
ol
th
a
t
c
a
n
b
e
a
ppl
ie
d
in
s
to
c
k m
a
r
ke
t
de
c
is
io
n
-
m
a
ki
ng pr
oc
e
s
s
e
s
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
T
he
a
ut
hor
s
de
c
la
r
e
t
ha
t
no f
undi
ng w
a
s
r
e
c
e
iv
e
d f
or
c
onduc
ti
ng t
hi
s
r
e
s
e
a
r
c
h
.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
M
a
r
ia
m
A
it
A
l
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
S
a
id
A
c
hc
ha
b
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Y
oune
s
L
a
hr
ic
hi
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
A
ut
hor
s
s
ta
te
no c
onf
li
c
t
of
i
nt
e
r
e
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
. 14, No. 5, O
c
to
be
r
2025
:
3624
-
3633
3632
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
T
he
da
ta
th
a
t
s
uppor
t
th
e
f
in
di
ngs
of
th
is
s
tu
dy
a
r
e
ope
nl
y
a
va
il
a
bl
e
a
t
ht
tp
s
:/
/ww
w
.c
a
s
a
bl
a
nc
a
-
bour
s
e
.c
om
, r
e
f
e
r
e
nc
e
numbe
r
[
23]
.
R
E
F
E
R
E
N
C
E
S
[
1]
P
.
Q
.
K
ha
ng
e
t
al
.
,
“
M
a
c
hi
ne
l
e
a
r
ni
ng
f
or
l
i
qu
i
di
t
y
pr
e
di
c
t
i
on
on
V
i
e
t
na
m
e
s
e
s
t
oc
k
m
a
r
ke
t
,”
P
r
oc
e
di
a
C
om
put
e
r
Sc
i
e
nc
e
,
vol
.
192, pp. 3590
–
3597, 2021, doi
:
10.1016/
j
.pr
oc
s
.2021.09.132.
[
2]
H
.
B
a
e
k,
“
A
C
N
N
-
L
S
T
M
s
t
oc
k
pr
e
di
c
t
i
on
m
ode
l
ba
s
e
d
on
ge
ne
t
i
c
a
l
gor
i
t
hm
o
pt
i
m
i
z
a
t
i
on,”
A
s
i
a
-
P
ac
i
f
i
c
F
i
nanc
i
al
M
ar
k
e
t
s
,
vol
.
31, no. 2, pp. 205
–
220, J
un. 2024, doi
:
10.1007/
s
10690
-
023
-
09412
-
z.
[
3]
H
.
S
hi
,
A
.
W
e
i
,
X
.
X
u,
Y
.
Z
hu,
H
.
H
u,
a
nd
S
.
T
a
ng,
“
A
C
N
N
-
L
S
T
M
ba
s
e
d
de
e
p
l
e
a
r
ni
ng
m
ode
l
w
i
t
h
hi
gh
a
c
c
ur
a
c
y
a
n
d
r
obus
t
ne
s
s
f
or
c
a
r
bon
pr
i
c
e
f
or
e
c
a
s
t
i
ng:
a
c
a
s
e
of
S
he
nz
he
n’
s
c
a
r
bon
m
a
r
ke
t
i
n
C
hi
na
,”
J
our
nal
of
E
nv
i
r
onm
e
nt
al
M
anage
m
e
nt
,
vol
. 352, F
e
b. 2024, doi
:
10.1016/
j
.j
e
nvm
a
n.2024.120131.
[
4]
H
.
N
.
B
h
a
nda
r
i
,
B
.
R
i
m
a
l
,
N
.
R
.
P
okhr
e
l
,
R
.
R
i
m
a
l
,
K
.
R
.
D
a
ha
l
,
a
nd
R
.
K
.
C
.
K
ha
t
r
i
,
“
P
r
e
di
c
t
i
ng
s
t
oc
k
m
a
r
ke
t
i
nde
x
us
i
n
g
L
S
T
M
,”
M
ac
hi
ne
L
e
ar
ni
ng w
i
t
h A
ppl
i
c
at
i
ons
, vol
. 9, S
e
p. 2022, doi
:
10.1016/
j
.m
l
w
a
.2022.100320.
[
5]
W
. L
u, J
.
L
i
, Y
.
L
i
, A
.
S
un, a
nd J
. W
a
ng,
“
A
C
N
N
-
L
S
T
M
-
ba
s
e
d
m
ode
l
t
o f
or
e
c
a
s
t
s
t
oc
k pr
i
c
e
s
,
”
C
om
pl
e
x
i
t
y
, vol
. 2020, pp.
1
–
10,
N
ov. 2020, doi
:
10.1155/
2020/
6622927.
[
6]
S
.
M
e
ht
a
b,
J
.
S
e
n,
a
nd
S
.
D
a
s
gupt
a
,
“
R
obus
t
a
n
a
l
ys
i
s
of
s
t
oc
k
pr
i
c
e
t
i
m
e
s
e
r
i
e
s
us
i
ng
C
N
N
a
nd
L
S
T
M
-
ba
s
e
d
de
e
p
l
e
a
r
ni
ng
m
ode
l
s
,”
i
n
P
r
oc
e
e
di
ngs
of
t
he
4t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
E
l
e
c
t
r
oni
c
s
,
C
om
m
uni
c
at
i
on
and
A
e
r
os
pac
e
T
e
c
hnol
ogy
,
I
C
E
C
A
2020
, N
ov. 2020, pp. 1481
–
1486, doi
:
10.1109/
I
C
E
C
A
49313.2020.9297652.
[
7]
M
. Y
a
ng a
nd J
. W
a
ng, “
A
d
a
pt
a
bi
l
i
t
y of
f
i
na
nc
i
a
l
t
i
m
e
s
e
r
i
e
s
pr
e
di
c
t
i
on ba
s
e
d o
n B
i
L
S
T
M
,”
P
r
oc
e
di
a
C
om
put
e
r
Sc
i
e
nc
e
, vol
. 199
,
pp. 18
–
25, 2021, doi
:
10.1016/
j
.pr
oc
s
.2022.01.003.
[
8]
S
.
M
e
ht
a
b
a
nd
J
.
S
e
n,
“
S
t
oc
k
pr
i
c
e
pr
e
di
c
t
i
on
us
i
ng
C
N
N
a
nd
L
S
T
M
-
ba
s
e
d
de
e
p
l
e
a
r
ni
ng
m
ode
l
s
,”
i
n
2020
I
nt
e
r
nat
i
ona
l
C
onf
e
r
e
nc
e
on
D
e
c
i
s
i
on
A
i
d
Sc
i
e
nc
e
s
and
A
ppl
i
c
at
i
on,
D
A
SA
2020
,
N
ov.
2020,
pp.
447
–
453,
doi
:
10.1109/
D
A
S
A
51403.2020.9317207.
[
9]
J
.
R
a
s
he
e
d,
A
.
J
a
m
i
l
,
A
.
A
l
i
H
a
m
e
e
d,
M
.
I
l
ya
s
,
A
.
O
z
ya
va
s
,
a
nd
N
.
A
j
l
ouni
,
“
I
m
pr
ovi
ng
s
t
oc
k
pr
e
di
c
t
i
on
a
c
c
ur
a
c
y
us
i
ng
C
N
N
a
nd
L
S
T
M
,”
i
n
2020
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
D
at
a
A
nal
y
t
i
c
s
f
or
B
us
i
ne
s
s
and
I
ndus
t
r
y
:
W
ay
T
ow
ar
ds
a
Sus
t
ai
nabl
e
E
c
onom
y
, I
C
D
A
B
I
2020
, O
c
t
. 2020, pp. 1
–
5, doi
:
10.1109/
I
C
D
A
B
I
51230.2020.9325597.
[
10]
X
.
Z
ha
o,
Y
.
L
i
u,
a
nd
Q
.
Z
ha
o,
“
G
e
ne
r
a
l
i
z
e
d
l
o
s
s
-
ba
s
e
d
C
N
N
-
B
i
L
S
T
M
f
or
s
t
oc
k
m
a
r
ke
t
pr
e
di
c
t
i
on,”
I
nt
e
r
nat
i
onal
J
ou
r
nal
o
f
F
i
nanc
i
al
St
udi
e
s
, vol
. 12, no. 3, p. 61, J
un. 2024, doi
:
10.3390/
i
j
f
s
12030061.
[
11]
M
.
R
.
S
a
hi
b,
H
.
E
l
ki
na
,
a
nd
T
.
Z
a
ki
,
“
F
r
om
t
e
c
hni
c
a
l
i
ndi
c
a
t
or
s
t
o
t
r
a
di
ng
de
c
i
s
i
ons
:
a
de
e
p
l
e
a
r
ni
ng
m
ode
l
c
om
bi
ni
ng
C
N
N
a
n
d
L
S
T
M
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
A
dv
anc
e
d
C
om
put
e
r
Sc
i
e
nc
e
and
A
ppl
i
c
at
i
ons
,
vol
.
15,
no.
6,
pp.
847
–
855,
2024,
doi
:
10.14569/
I
J
A
C
S
A
.2024.0150685.
[
12]
D
.
P
a
t
na
i
k,
N
.
V
.
J
.
R
a
o,
B
.
P
a
dhi
a
r
i
,
a
nd
S
.
P
a
t
na
i
k,
“
O
pt
i
m
i
s
e
d
hybr
i
d
C
N
N
-
L
S
T
M
m
ode
l
f
o
r
s
t
oc
k
pr
i
c
e
p
r
e
di
c
t
i
on,”
I
nt
e
r
nat
i
onal
J
our
nal
of
M
anage
m
e
nt
and
D
e
c
i
s
i
on
M
ak
i
ng
,
vo
l
.
23,
no.
4,
pp.
438
–
460,
2024,
doi
:
10.1504/
I
J
M
D
M
.2024.139387.
[
13]
A
. V
i
da
l
a
nd W
. K
r
i
s
t
j
a
npol
l
e
r
, “
G
ol
d vol
a
t
i
l
i
t
y pr
e
di
c
t
i
on us
i
ng a
C
N
N
-
L
S
T
M
a
ppr
oa
c
h,”
E
x
pe
r
t
Sy
s
t
e
m
s
w
i
t
h A
ppl
i
c
at
i
ons
, vol
.
157, N
ov. 2020, doi
:
10.1016/
j
.e
s
w
a
.2020.113481.
[
14]
Y
.
C
he
n
e
t
al
.
,
“
S
t
oc
k
p
r
i
c
e
f
or
e
c
a
s
t
ba
s
e
d
on
C
N
N
-
B
i
L
S
T
M
-
E
C
A
m
ode
l
,”
Sc
i
e
nt
i
f
i
c
P
r
ogr
am
m
i
ng
,
vol
.
2021,
pp.
1
–
20,
J
u
l
.
2021, doi
:
10.1155/
2021/
2446543.
[
15]
M
.
T
.
P
a
w
i
t
r
a
,
H
.
F
a
khr
ur
r
oj
a
,
a
nd
L
.
A
bdur
r
a
hm
a
n,
“
P
r
e
di
c
t
i
ng
s
t
oc
k
m
a
r
ke
t
us
i
ng
C
N
N
a
nd
B
i
L
S
T
M
m
od
e
l
,”
i
n
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
C
om
put
e
r
,
C
ont
r
ol
,
I
nf
or
m
at
i
c
s
and
i
t
s
A
ppl
i
c
at
i
ons
,
I
C
3I
N
A
,
O
c
t
.
2024,
no.
2024,
pp.
267
–
272,
doi
:
10.1109/
I
C
3I
N
A
64086.2024.10732245.
[
16]
A
.
S
ha
r
m
a
,
C
.
K
.
V
e
r
m
a
,
a
nd
P
.
S
i
ngh,
“
E
nha
nc
i
ng
opt
i
on
pr
i
c
i
ng
a
c
c
ur
a
c
y
i
n
t
he
I
ndi
a
n
m
a
r
ke
t
:
a
C
N
N
-
B
i
L
S
T
M
a
ppr
oa
c
h,”
C
om
put
at
i
onal
E
c
onom
i
c
s
, vol
. 65, no. 6, pp. 3751
–
3778, J
un. 2025, doi
:
10.1007/
s
10614
-
024
-
10689
-
z.
[
17]
X
.
Z
ha
ng,
Y
.
M
a
,
a
nd
M
.
W
a
ng,
“
A
n
a
t
t
e
nt
i
on
-
ba
s
e
d
l
ogi
s
t
i
c
-
C
N
N
-
B
i
L
S
T
M
hybr
i
d
ne
ur
a
l
ne
t
w
or
k
f
or
c
r
e
di
t
r
i
s
k
pr
e
di
c
t
i
on
of
l
i
s
t
e
d r
e
a
l
e
s
t
a
t
e
e
nt
e
r
pr
i
s
e
s
,”
E
x
pe
r
t
Sy
s
t
e
m
s
, vol
. 41, no. 2, F
e
b. 2024, doi
:
10.1111/
e
xs
y.13299.
[
18]
R
.
Z
ha
o,
Z
.
L
e
i
,
a
nd Z
. Z
ha
o,
“
R
e
s
e
a
r
c
h
on
t
he
a
ppl
i
c
a
t
i
on
of
de
e
p
l
e
a
r
ni
ng
t
e
c
hni
que
s
i
n
s
t
oc
k
m
a
r
ke
t
pr
e
di
c
t
i
on
a
nd i
nve
s
t
m
e
nt
de
c
i
s
i
on
-
m
a
ki
ng i
n f
i
na
nc
i
a
l
m
a
na
ge
m
e
nt
,”
F
r
ont
i
e
r
s
i
n E
ne
r
gy
R
e
s
e
a
r
c
h
, vol
.
12, M
a
y 2024, doi
:
10.3389/
f
e
nr
g.2024.1376677.
[
19]
A
.
L
uo,
L
.
Z
hong,
J
.
W
a
ng,
Y
.
W
a
ng,
S
.
L
i
,
a
nd
W
.
T
a
i
,
“
S
hor
t
-
t
e
r
m
s
t
oc
k
c
or
r
e
l
a
t
i
on
f
or
e
c
a
s
t
i
ng
ba
s
e
d
on
C
N
N
-
B
i
L
S
T
M
e
nha
nc
e
d by a
t
t
e
nt
i
on m
e
c
ha
ni
s
m
,”
I
E
E
E
A
c
c
e
s
s
, vol
. 12, pp. 29617
–
29632, 2024, doi
:
10.1109/
A
C
C
E
S
S
.2024.3369419.
[
20]
A
.
N
guye
n,
S
.
H
a
,
a
nd
P
.
N
guye
n,
“
B
i
C
onN
e
t
:
a
t
i
m
e
-
de
l
a
ye
d
hybr
i
d
de
e
p
l
e
a
r
ni
ng
m
ode
l
f
or
e
qui
t
y
pr
i
c
e
p
r
e
di
c
t
i
on,”
SSR
N
E
l
e
c
t
r
oni
c
J
our
nal
, 2024, doi
:
10.2139/
s
s
r
n.4729187.
[
21]
M
.
S
.
A
kt
e
r
,
H
.
S
ha
hr
i
a
r
,
R
.
C
how
dhur
y,
a
nd
M
.
R
.
C
.
M
a
hdy,
“
F
or
e
c
a
s
t
i
ng
t
he
r
i
s
k
f
a
c
t
or
of
f
r
on
t
i
e
r
m
a
r
ke
t
s
:
a
nove
l
s
t
a
c
ki
n
g
e
ns
e
m
bl
e
of
ne
ur
a
l
ne
t
w
or
k a
ppr
oa
c
h,”
F
ut
ur
e
I
nt
e
r
ne
t
, vol
. 14, no. 9, p. 252,
A
ug. 2022, doi
:
10.3390/
f
i
14090252.
[
22]
H
.
W
a
ng,
J
.
W
a
ng,
L
.
C
a
o,
Y
.
L
i
,
Q
.
S
un,
a
nd
J
.
W
a
ng,
“
A
s
t
oc
k
c
l
os
i
ng
pr
i
c
e
pr
e
di
c
t
i
on
m
ode
l
ba
s
e
d
on
C
N
N
-
B
i
S
L
S
T
M
,”
C
om
pl
e
x
i
t
y
, vol
. 2021, no. 1, J
a
n. 2021, doi
:
10.1155/
2021/
5360828.
[
23]
“
C
om
pos
i
t
i
on
e
t
hi
s
t
or
i
que
de
s
i
ndi
c
e
s
,
”
B
our
s
e
d
e
C
a
s
abl
anc
a
,
2024.
ht
t
p
s
:
/
/
w
w
w
.c
a
s
a
bl
a
nc
a
-
bour
s
e
.c
om
/
f
r
/
c
om
pos
i
t
i
on
-
et
-
hi
s
t
or
i
que
-
de
s
-
i
ndi
c
e
s
[
24]
M
.
R
.
J
a
ha
n
-
P
a
r
va
r
a
nd
F
.
Z
i
ke
s
,
“
W
he
n
do
l
ow
-
f
r
e
que
nc
y
m
e
a
s
ur
e
s
r
e
a
l
l
y
m
e
a
s
ur
e
e
f
f
e
c
t
i
ve
s
pr
e
a
ds
?
e
vi
de
nc
e
f
r
om
e
qui
t
y
a
n
d
f
or
e
i
gn e
xc
ha
nge
m
a
r
ke
t
s
,”
R
e
v
i
e
w
of
F
i
nanc
i
al
St
udi
e
s
, vol
. 36, no. 10, pp. 41
90
–
4232, S
e
p. 2023, doi
:
10.1093/
r
f
s
/
hha
d028.
[
25]
A
.
J
.
M
e
n
kve
l
d
e
t
a
l
.
, “
N
ons
t
a
n
da
r
d
e
r
r
o
r
s
,”
J
our
nal
o
f
F
i
na
nc
e
,
v
ol
. 7
9,
no
.
3,
pp
. 2
339
–
2
390
,
J
un.
2
024
,
doi
:
10
.11
11
/
j
of
i
.
133
37
.
[
26]
Y
.
Z
he
ng,
Q
.
L
i
u,
E
.
C
he
n,
Y
.
G
e
,
a
nd
J
.
L
.
Z
ha
o,
“
T
i
m
e
s
e
r
i
e
s
c
l
a
s
s
i
f
i
c
a
t
i
on
us
i
ng
m
ul
t
i
-
c
ha
nne
l
s
de
e
p
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
ks
,”
I
nt
e
r
nat
i
onal
C
onf
e
r
e
n
c
e
on
W
e
b
-
age
I
nf
or
m
at
i
on
M
ana
ge
m
e
nt
,
vol
.
8485,
2014,
pp.
298
–
310,
doi
:
10.1007/
978
-
3
-
319
-
08010
-
9_33.
[
27]
Z
.
W
a
ng,
W
.
Y
a
n,
a
nd
T
.
O
a
t
e
s
,
“
T
i
m
e
s
e
r
i
e
s
c
l
a
s
s
i
f
i
c
a
t
i
on
f
r
om
s
c
r
a
t
c
h
w
i
t
h
de
e
p
ne
ur
a
l
ne
t
w
or
ks
:
a
s
t
r
ong
ba
s
e
l
i
ne
,”
i
n
P
r
oc
e
e
di
ngs
of
t
h
e
I
nt
e
r
nat
i
onal
J
oi
nt
C
onf
e
r
e
n
c
e
on
N
e
ur
al
N
e
t
w
or
k
s
,
M
a
y
2017,
pp.
1578
–
1585,
doi
:
10.1109/
I
J
C
N
N
.2017.7966039.
[
28]
M
.
F
a
i
q
e
t
al
.
,
“
P
r
e
di
c
t
i
on
of
e
ne
r
gy
c
ons
um
pt
i
on
i
n
c
a
m
pus
bui
l
di
ngs
us
i
ng
l
ong
s
hor
t
-
t
e
r
m
m
e
m
or
y,”
A
l
e
x
andr
i
a
E
ngi
ne
e
r
i
ng
J
our
nal
, vol
. 67, pp. 65
–
76, M
a
r
. 2023, doi
:
10.1016/
j
.a
e
j
.2022.12.015.
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
m
ar
k
e
t
li
qui
di
ty
:
hy
br
id
de
e
p l
e
ar
ni
ng app
r
oac
he
s
f
or
pr
e
di
c
ti
on
…
(
M
ar
ia
m
A
it
A
l
)
3633
[
29]
J
.
P
a
t
e
l
,
S
.
S
ha
h,
P
.
T
ha
kka
r
,
a
nd
K
.
K
ot
e
c
h
a
,
“
P
r
e
di
c
t
i
ng
s
t
oc
k
a
nd
s
t
oc
k
pr
i
c
e
i
nde
x
m
ove
m
e
nt
us
i
ng
t
r
e
nd
de
t
e
r
m
i
ni
s
t
i
c
da
t
a
pr
e
pa
r
a
t
i
on
a
nd
m
a
c
hi
ne
l
e
a
r
ni
ng
t
e
c
hni
que
s
,”
E
x
pe
r
t
Sy
s
t
e
m
s
w
i
t
h
A
ppl
i
c
at
i
ons
,
vol
.
42,
no.
1,
pp.
259
–
268,
J
a
n.
2015,
doi
:
10.1016/
j
.e
s
w
a
.2014.07.040.
[
30]
T
.
F
i
s
c
he
r
a
nd
C
.
K
r
a
us
s
,
“
D
e
e
p
l
e
a
r
ni
ng
w
i
t
h
l
ong
s
hor
t
-
t
e
r
m
m
e
m
or
y
ne
t
w
or
ks
f
or
f
i
na
nc
i
a
l
m
a
r
ke
t
pr
e
di
c
t
i
ons
,”
E
ur
ope
an
J
our
nal
of
O
pe
r
at
i
onal
R
e
s
e
a
r
c
h
, vol
. 270, no. 2, pp. 654
–
669, O
c
t
. 2018, doi
:
10.1016/
j
.e
j
or
.2017.11.054.
[
31]
S
.
M
oot
ha
,
S
.
S
r
i
dha
r
,
R
.
S
e
e
t
ha
r
a
m
a
n,
a
nd
S
.
C
hi
t
r
a
ka
l
a
,
“
S
t
oc
k
pr
i
c
e
pr
e
di
c
t
i
on
us
i
ng
bi
-
di
r
e
c
t
i
ona
l
L
S
T
M
ba
s
e
d
s
e
que
nc
e
t
o
s
e
que
nc
e
m
ode
l
i
ng
a
nd
m
ul
t
i
t
a
s
k
l
e
a
r
ni
ng,”
2020
11t
h
I
E
E
E
A
nnual
U
bi
qui
t
ous
C
om
put
i
ng,
E
l
e
c
t
r
oni
c
s
and
M
obi
l
e
C
om
m
uni
c
at
i
on C
onf
e
r
e
nc
e
, U
E
M
C
O
N
2020
, pp. 78
–
86, 2020, doi
:
10.1109/
U
E
M
C
O
N
51285.2020.9298066.
[
32]
R
.
L
.
A
bdul
j
a
bba
r
,
H
.
D
i
a
,
a
nd
P
.
W
.
T
s
a
i
,
“
D
e
ve
l
opm
e
nt
a
nd
e
va
l
ua
t
i
on
of
bi
di
r
e
c
t
i
ona
l
L
S
T
M
f
r
e
e
w
a
y
t
r
a
f
f
i
c
f
or
e
c
a
s
t
i
ng
m
ode
l
s
us
i
ng s
i
m
ul
a
t
i
on da
t
a
,”
S
c
i
e
nt
i
f
i
c
R
e
por
t
s
, vo
l
. 11, no. 1
, D
e
c
. 2021, doi
:
10.1038/
s
41598
-
021
-
03282
-
z.
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Mariam
Ait
Al
is
currently
pursuing
Ph.D.
research
at
ENSIAS,
Mohammed
V
University
in
Rabat
,
Rabat,
Morocco.
She
is
also
a
Senior
Software
Engineer
and
IT
Project
Manager
with
extensive
experience
in
consultancy
for
complex
t
echnical
scenarios.
Her
professional
expertise
informs
her
academic
work,
focusing
on
bridging
practical
industry
applicati
ons
with
theoreti
cal
advancement
s
in
technolo
gy.
Her
re
search
interest
s
include
artificial
intell
igence,
deep
learning,
and
their
applicati
ons
in
financ
ial
markets.
She
can
be
contacted
at email
: mariam
_
aital@
um5.ac.ma
.
Said
Achchab
is
a
p
rofessor
of
digital
finance,
artificial
intellig
ence
and
risk
manageme
nt
at
ENSIAS,
Mohamed
V
University
in
Rabat
,
and
the
co
ordinator
of
the
"
Digital
Engineering
for
Finance"
engineering
program.
He
is
the
founding
chairman
of
the
African
Fintech
Institu
te. He
can
be co
ntacte
d at e
mail:
s.achchab@um5s.net.ma
.
Younes
Lahrichi
is
a
Ph.D.,
full
professor
and
senior
lecturer
in
Finance
a
t
ISCAE
Casablanca.
Aut
h
or
of
several
academic
publications,
he
is
t
he
head
of
finance
and
accountin
g
department
and
m
aster
in
digital
finance
program
directo
r.
He
can
be
contacted
at
email:
ylahrichi
@
groupeis
cae.ma
.
Evaluation Warning : The document was created with Spire.PDF for Python.