I
nte
rna
t
io
na
l J
o
urna
l o
f
I
nfo
rm
a
t
ics a
nd
Co
m
m
un
ica
t
io
n T
ec
hn
o
lo
g
y
(
I
J
-
I
CT
)
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
,
p
p
.
1
6
9
6
~1
7
0
5
I
SS
N:
2252
-
8
7
7
6
,
DOI
:
1
0
.
1
1
5
9
1
/iji
c
t
.
v15
i
4
.
pp
169
6
-
1
7
0
5
1696
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ict.
ia
esco
r
e.
co
m
Unidirec
tiona
l an
d bidirec
tiona
l L
STM
f
o
r non
linea
r sy
ntheti
c
time seri
es foreca
sting
Nur
H
a
izum
Abd Ra
hm
a
n
1
,
H
a
ni Sy
a
hid
a
Z
ulk
a
f
li
2
1
C
e
n
t
r
e
f
o
r
M
a
t
h
e
m
a
t
i
c
a
l
S
c
i
e
n
c
e
s,
U
n
i
v
e
r
s
i
t
i
M
a
l
a
y
si
a
P
a
h
a
n
g
A
l
-
S
u
l
t
a
n
A
b
d
u
l
l
a
h
,
Le
b
u
h
P
e
r
si
a
r
a
n
T
u
n
K
h
a
l
i
l
Y
a
a
k
o
b
,
G
a
mb
a
n
g
,
M
a
l
a
y
s
i
a
2
D
e
p
a
r
t
me
n
t
o
f
M
a
t
h
e
m
a
t
i
c
s
a
n
d
S
t
a
t
i
st
i
c
s
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
,
U
n
i
v
e
r
si
t
i
P
u
t
r
a
M
a
l
a
y
s
i
a
,
S
e
r
d
a
n
g
,
M
a
l
a
y
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Feb
7
,
2
0
2
6
R
ev
is
ed
J
u
l 2
4
,
2
0
2
6
Acc
ep
ted
Au
g
2
8
,
2
0
2
6
De
e
p
lea
rn
in
g
m
o
d
e
ls
a
re
se
n
siti
v
e
to
a
rc
h
it
e
c
tu
re
c
h
o
ice
a
n
d
h
y
p
e
r
p
a
ra
m
e
ter
tu
n
in
g
.
Th
is
st
u
d
y
e
v
a
lu
a
tes
th
e
fo
re
c
a
stin
g
p
e
rf
o
r
m
a
n
c
e
o
f
u
n
i
d
irec
ti
o
n
a
l
lo
n
g
sh
o
rt
-
term
m
e
m
o
ry
(LS
TM
)
a
n
d
b
i
d
irec
ti
o
n
a
l
LS
TM
(BiL
S
TM
)
n
e
two
rk
s
u
sin
g
sy
n
th
e
ti
c
n
o
n
l
in
e
a
r
ti
m
e
se
ries
d
a
ta.
S
y
n
t
h
e
ti
c
d
a
ta
we
re
u
se
d
t
o
p
ro
v
i
d
e
a
c
o
n
tro
ll
e
d
e
x
p
e
rime
n
tal
se
tt
i
n
g
in
wh
ich
t
h
e
n
o
n
li
n
e
a
r
d
a
ta
-
g
e
n
e
ra
ti
n
g
p
ro
c
e
s
s,
sa
m
p
le
siz
e
,
a
n
d
h
y
p
e
rp
a
ra
m
e
ter
e
ffe
c
ts
c
o
u
ld
b
e
sy
ste
m
a
ti
c
a
ll
y
e
x
a
m
in
e
d
.
Th
re
sh
o
ld
a
u
to
re
g
re
ss
iv
e
(TAR)
a
n
d
se
lf
-
e
x
c
it
in
g
th
re
sh
o
ld
a
u
to
re
g
re
ss
iv
e
(S
ET
AR)
m
o
d
e
ls
we
re
sim
u
late
d
wit
h
sa
m
p
le
siz
e
s,
1
0
0
0
a
n
d
3
0
0
0
.
Ea
c
h
se
ries
wa
s
p
a
rti
ti
o
n
e
d
i
n
to
tra
in
in
g
a
n
d
tes
ti
n
g
se
ts
u
sin
g
a
n
8
0
:
2
0
sp
li
t.
Bo
t
h
n
e
two
rk
s
we
re
trai
n
e
d
u
sin
g
g
ri
d
se
a
rc
h
b
y
v
a
r
y
in
g
th
e
n
u
m
b
e
r
o
f
h
i
d
d
e
n
n
e
u
ro
n
s
(3
2
,
6
4
,
1
2
8
)
a
n
d
trai
n
in
g
e
p
o
c
h
s
(
2
0
,
5
0
,
1
0
0
).
F
o
re
c
a
st
a
c
c
u
ra
c
y
wa
s
e
v
a
lu
a
ted
u
sin
g
m
e
a
n
sq
u
a
re
d
e
rro
r
(M
S
E)
a
n
d
m
e
a
n
a
b
so
l
u
te
e
rro
r
(M
AE),
a
v
e
ra
g
e
d
a
c
ro
ss
3
0
re
p
li
c
a
ti
o
n
s.
T
h
e
re
su
lt
s
sh
o
w
t
h
a
t
in
c
re
a
sin
g
e
p
o
c
h
s
a
n
d
h
i
d
d
e
n
n
e
u
r
o
n
s
imp
ro
v
e
s
a
c
c
u
ra
c
y
,
b
u
t
th
e
i
m
p
ro
v
e
m
e
n
ts
d
e
c
li
n
e
a
t
h
ig
h
e
r
se
tt
in
g
s
.
BiL
S
TM
a
c
h
iev
e
s
l
o
we
r
fo
re
c
a
stin
g
e
rr
o
rs
th
a
n
LS
TM
fo
r
b
o
t
h
TAR
a
n
d
S
ET
AR
d
a
tas
e
ts,
a
n
d
t
h
e
Wi
lco
x
o
n
tes
t
c
o
n
firms
th
a
t
th
e
d
iffer
e
n
c
e
s
a
re
sta
ti
stica
ll
y
sig
n
ifi
c
a
n
t
.
T
h
e
m
a
i
n
c
o
n
tri
b
u
ti
o
n
o
f
t
h
is
st
u
d
y
is
a
c
o
n
tro
ll
e
d
e
m
p
iri
c
a
l
fra
m
e
wo
rk
th
a
t
li
n
k
s
th
e
a
rc
h
it
e
c
tu
re
,
sa
m
p
le
siz
e
,
a
n
d
h
y
p
e
r
p
a
ra
m
e
ter
se
tt
in
g
s
to
n
o
n
li
n
e
a
r
re
g
ime
-
sw
it
c
h
in
g
fo
re
c
a
stin
g
p
e
rfo
rm
a
n
c
e
.
Th
e
fin
d
in
g
s
su
g
g
e
st
th
a
t
B
iL
S
TM
is
p
re
fe
ra
b
le
wh
e
n
h
i
g
h
e
r
a
c
c
u
ra
c
y
is
re
q
u
ired
,
w
h
il
e
L
S
TM
re
m
a
in
s
a
p
ra
c
ti
c
a
l
a
lt
e
rn
a
ti
v
e
fo
r
sm
o
o
th
e
r
n
o
n
li
n
e
a
r
se
ries
o
r
re
so
u
rc
e
-
c
o
n
stra
in
e
d
a
p
p
l
ica
ti
o
n
s.
K
ey
w
o
r
d
s
:
T
h
r
esh
o
ld
a
u
to
r
eg
r
ess
iv
e
m
o
d
el
Self
-
ex
citin
g
th
r
esh
o
l
d
au
to
r
eg
r
ess
iv
e
m
o
d
el
T
im
e
s
er
ies
Hy
p
er
p
ar
a
m
eter
tu
n
in
g
Gr
id
s
ea
r
ch
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Nu
r
Haiz
u
m
Ab
d
R
ah
m
an
C
en
tr
e
f
o
r
Ma
th
em
atica
l Scie
n
ce
s
,
Un
iv
er
s
iti Ma
lay
s
ia
Pah
an
g
Al
-
Su
ltan
Ab
d
u
llah
L
eb
u
h
Per
s
iar
an
T
u
n
Kh
alil Y
aa
k
o
b
,
Gam
b
an
g
,
Ma
lay
s
ia
E
m
ail: h
aizu
m
@
u
m
p
s
a.
ed
u
.
m
y
1.
I
NT
RO
D
UCT
I
O
N
Acc
u
r
ate
tim
e
s
er
ies
f
o
r
ec
asti
n
g
is
cr
u
cial
f
o
r
d
ec
is
io
n
-
m
ak
in
g
in
s
ec
to
r
s
s
u
ch
as
b
an
k
in
g
,
tr
an
s
p
o
r
tatio
n
,
en
er
g
y
,
an
d
en
v
ir
o
n
m
e
n
tal
s
cien
ce
,
wh
e
r
e
p
r
ed
ictin
g
f
u
tu
r
e
tr
en
d
s
a
n
d
an
o
m
alies
is
ess
en
tial
f
o
r
o
p
er
atio
n
al
ef
f
icien
cy
an
d
r
is
k
m
itig
atio
n
[
1
]
-
[
3
]
.
T
h
e
i
n
cr
ea
s
in
g
co
m
p
lex
ity
o
f
tem
p
o
r
al
co
n
n
ec
ti
o
n
s
in
r
ea
l
-
wo
r
ld
d
atasets
h
as
em
p
h
asis
ed
th
e
s
h
o
r
tco
m
i
n
g
s
o
f
co
n
v
en
tio
n
al
s
tatis
tical
m
o
d
els,
s
u
ch
as
au
to
r
eg
r
ess
iv
e
in
teg
r
ated
m
o
v
in
g
av
er
a
g
e
(
AR
I
MA
)
an
d
ex
p
o
n
en
tial
s
m
o
o
th
in
g
.
T
h
ese
m
o
d
els
o
f
ten
s
tr
u
g
g
le
to
ca
p
tu
r
e
n
o
n
lin
ea
r
p
atter
n
s
an
d
lo
n
g
-
r
a
n
g
e
c
o
r
r
elatio
n
s
in
c
o
m
p
lex
f
o
r
ec
asti
n
g
task
s
.
As
a
r
esu
lt,
th
e
r
esear
ch
co
m
m
u
n
ity
h
as
tr
an
s
itio
n
ed
to
d
ee
p
lear
n
i
n
g
tec
h
n
iq
u
es,
p
ar
tic
u
lar
ly
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs),
to
ad
d
r
ess
th
ese
is
s
u
e
s
an
d
ac
h
iev
e
e
n
h
an
ce
d
ac
cu
r
a
cy
in
tim
e
-
s
er
ies f
o
r
ec
asti
n
g
[
4
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Un
id
ir
ec
tio
n
a
l a
n
d
b
id
ir
ec
tio
n
a
l LS
TM
fo
r
n
o
n
lin
ea
r
s
yn
th
etic
time
s
erie
s
… (
N
u
r
Ha
iz
u
m
A
b
d
R
a
h
ma
n
)
1697
Am
o
n
g
R
NN
v
ar
ian
ts
,
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
n
et
wo
r
k
s
h
av
e
em
e
r
g
ed
as a
r
o
b
u
s
t so
lu
tio
n
f
o
r
m
o
d
ellin
g
s
eq
u
en
tial
d
ata,
d
u
e
t
o
th
e
L
STM
ch
ar
ac
ter
is
t
ic
o
f
a
g
ated
ar
ch
itectu
r
e
th
at
f
ac
ilit
ates
m
em
o
r
y
r
eten
tio
n
a
n
d
s
elec
tiv
e
ex
clu
s
io
n
o
f
i
n
f
o
r
m
atio
n
o
v
er
ex
t
en
d
ed
tim
e
p
er
io
d
s
[
5
]
.
L
ST
M
m
o
d
els
ex
ce
l
at
ca
p
tu
r
in
g
s
h
o
r
t
-
te
r
m
v
ar
iatio
n
s
an
d
lo
n
g
-
ter
m
d
e
p
en
d
e
n
cies,
m
ak
in
g
th
e
L
STM
h
ig
h
ly
s
u
itab
le
f
o
r
co
m
p
licated
d
ata
f
o
r
ec
asti
n
g
ap
p
licatio
n
s
.
T
wo
p
r
in
cip
al
L
STM
ar
ch
itectu
r
es
h
av
e
g
ain
ed
p
r
o
m
i
n
en
ce
,
wh
ic
h
ar
e
u
n
i
d
ir
ec
tio
n
al
L
STM
,
wh
ich
p
r
o
ce
s
s
es
in
p
u
t
s
eq
u
en
ce
s
in
a
s
in
g
le
f
o
r
war
d
d
ir
ec
tio
n
,
an
d
b
i
d
ir
ec
tio
n
al
L
STM
(
B
iLST
M)
,
wh
ich
a
n
aly
s
es
d
ata
in
b
o
th
f
o
r
war
d
an
d
b
ac
k
w
ar
d
d
ir
ec
tio
n
s
[
6
]
.
T
h
is
b
id
ir
ec
tio
n
al
m
ec
h
an
is
m
allo
ws
B
iLST
M
m
o
d
els
to
lev
er
ag
e
co
n
tex
tu
al
in
f
o
r
m
atio
n
f
r
o
m
b
o
th
p
ast
an
d
f
u
tu
r
e
s
tates,
p
o
ten
tially
en
h
a
n
cin
g
f
o
r
ec
asti
n
g
p
er
f
o
r
m
an
c
e
in
ap
p
licatio
n
s
wh
er
e
co
m
p
r
eh
en
s
iv
e
s
eq
u
e
n
ce
co
n
tex
t is ess
en
tial.
I
n
r
ec
e
n
t
y
ea
r
s
,
th
e
liter
atu
r
e
o
n
L
STM
-
b
ased
f
o
r
ec
asti
n
g
h
as
r
ap
id
ly
e
v
o
lv
e
d
to
tack
le
in
cr
ea
s
in
g
ly
co
m
p
licated
tem
p
o
r
al
p
atter
n
s
ac
r
o
s
s
s
ev
er
al
ap
p
licatio
n
ar
ea
s
.
Simu
latio
n
s
tu
d
ies
co
n
s
is
ten
tl
y
s
h
o
w
th
at
B
iLST
M
m
o
d
els
o
u
tp
er
f
o
r
m
u
n
id
ir
ec
tio
n
al
m
o
d
els
in
ca
p
t
u
r
in
g
in
tr
icate
tem
p
o
r
al
d
e
p
e
n
d
en
cies,
esp
ec
ially
in
s
h
o
r
t
-
an
d
m
ed
iu
m
-
ter
m
f
o
r
ec
asti
n
g
co
n
tex
ts
s
u
ch
as
tr
af
f
ic
f
lo
w,
s
to
ck
m
ar
k
et
p
r
e
d
ictio
n
,
an
d
h
y
d
r
o
l
o
g
ical
m
o
d
ellin
g
.
T
h
e
b
id
ir
ec
tio
n
al
ar
ch
itectu
r
e
allo
ws
th
e
m
o
d
el
to
an
aly
s
e
in
p
u
t
s
eq
u
e
n
ce
s
in
b
o
th
d
ir
ec
tio
n
s
,
u
tili
s
in
g
in
f
o
r
m
atio
n
f
r
o
m
b
o
th
p
ast
an
d
f
u
tu
r
e
s
tates
to
im
p
r
o
v
e
f
o
r
ec
ast
ac
cu
r
ac
y
.
Mu
lti
-
lay
er
B
iLST
M
m
o
d
els
h
av
e
d
em
o
n
s
tr
ated
co
n
s
id
er
a
b
le
im
p
r
o
v
em
e
n
ts
in
f
o
r
ec
ast
ac
cu
r
ac
y
r
elativ
e
to
s
tack
ed
o
r
h
y
b
r
id
Un
i
-
L
STM
m
o
d
els,
p
ar
ticu
lar
ly
f
o
r
ex
ten
d
ed
f
o
r
ec
ast h
o
r
izo
n
s
[
7
]
-
[
9
]
.
I
n
co
m
p
ar
ativ
e
a
n
aly
s
is
,
th
e
p
er
f
o
r
m
a
n
ce
ev
alu
ati
o
n
r
elies
o
n
a
v
a
r
iety
o
f
er
r
o
r
m
etr
ics,
s
u
ch
as
r
o
o
t
m
ea
n
s
q
u
ar
ed
er
r
o
r
(
R
MSE
)
,
m
ea
n
ab
s
o
lu
te
er
r
o
r
(
MA
E
)
,
m
ea
n
ab
s
o
lu
te
p
er
ce
n
tag
e
e
r
r
o
r
(
MA
PE)
,
a
n
d
co
ef
f
icien
t
o
f
d
eter
m
in
atio
n
(
R
2
)
.
B
esid
es
th
e
in
f
er
en
tial
s
t
atis
tical
te
s
ts
,
s
u
ch
as
th
e
W
il
co
x
o
n
Sig
n
ed
-
R
an
k
T
est,
is
also
u
s
ed
[
1
0
]
.
B
iLST
M
m
o
d
els,
wh
ich
p
r
o
d
u
ce
l
o
wer
er
r
o
r
r
ates
an
d
g
r
ea
ter
co
r
r
elatio
n
s
co
r
es
ac
r
o
s
s
m
an
y
ap
p
licatio
n
s
an
d
f
o
r
ec
ast
h
o
r
izo
n
s
,
ar
e
f
r
eq
u
en
tly
f
a
v
o
u
r
ed
b
y
em
p
ir
ical
r
esu
lts
.
T
h
ese
im
p
r
o
v
em
e
n
ts
,
h
o
wev
er
,
co
m
e
at
th
e
co
s
t
o
f
lo
n
g
e
r
tr
ai
n
in
g
tim
es
a
n
d
g
r
ea
ter
co
m
p
u
t
atio
n
al
co
m
p
lex
ity
,
p
r
o
m
p
tin
g
r
esear
ch
e
r
s
to
ex
p
lo
r
e
h
ar
d
war
e
ac
ce
ler
ati
o
n
an
d
o
p
tim
is
atio
n
tech
n
iq
u
es
to
r
ed
u
ce
r
eso
u
r
ce
r
eq
u
ir
em
e
n
ts
.
W
h
ile
en
s
em
b
le
ap
p
r
o
ac
h
es
th
at
in
teg
r
ate
L
STM
,
B
iLST
M,
an
d
o
th
er
d
e
ep
lear
n
in
g
m
o
d
els
o
f
f
er
p
r
o
m
is
in
g
av
en
u
es
f
o
r
en
h
an
ce
d
r
esil
ien
ce
,
ap
p
l
y
in
g
cr
o
s
s
-
v
alid
atio
n
,
m
u
lti
-
h
o
r
izo
n
f
o
r
ec
asti
n
g
,
an
d
d
o
m
ain
-
s
p
ec
if
ic
m
ea
s
u
r
es f
u
r
t
h
er
im
p
r
o
v
es th
e
r
eliab
ilit
y
o
f
th
ese
co
m
p
ar
ativ
e
e
v
alu
atio
n
s
[
1
1
]
-
[
1
4
]
.
I
n
ad
d
itio
n
,
alth
o
u
g
h
B
iLST
M
m
o
d
els
o
f
f
er
clea
r
b
en
ef
its
in
ac
cu
r
ac
y
an
d
ca
p
tu
r
in
g
co
n
tex
t,
th
eir
p
r
ac
tical
ap
p
licatio
n
is
o
f
ten
lim
ited
b
y
co
m
p
u
tatio
n
al
tr
ad
e
-
o
f
f
s
.
R
esear
ch
in
d
icate
s
th
a
t
B
iLST
M
m
o
d
els
r
eq
u
ir
e
m
o
r
e
t
r
ain
in
g
tim
e
an
d
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
an
d
m
ay
c
o
n
v
e
r
g
e
m
o
r
e
s
lo
wly
th
an
L
STM
m
o
d
els.
T
h
is
h
as
s
p
ar
k
ed
in
ter
est
i
n
o
p
tim
is
atio
n
tech
n
iq
u
es,
s
u
ch
as
h
y
p
er
p
ar
am
eter
tu
n
in
g
an
d
h
ar
d
war
e
ac
ce
le
r
atio
n
,
to
b
alan
ce
p
r
ed
i
ctiv
e
p
er
f
o
r
m
an
ce
with
ef
f
ici
en
cy
.
Ad
d
itio
n
ally
,
th
e
in
ter
p
r
etab
ilit
y
o
f
d
ee
p
lear
n
in
g
m
o
d
els r
em
ain
s
a
s
ig
n
if
ican
t c
h
allen
g
e,
with
atten
tio
n
m
ec
h
a
n
is
m
s
an
d
f
ea
tu
r
e
s
el
ec
tio
n
em
er
g
in
g
as
im
p
o
r
tan
t
s
tr
ateg
ies
f
o
r
im
p
r
o
v
in
g
tr
a
n
s
p
ar
en
cy
an
d
tr
u
s
t.
T
h
e
liter
atu
r
e
also
em
p
h
asis
es
th
e
im
p
o
r
tan
ce
o
f
b
etter
in
teg
r
atin
g
L
STM
m
o
d
els
with
co
m
p
lem
en
tar
y
m
e
th
o
d
s
,
s
u
ch
as
au
to
r
eg
r
ess
iv
e
in
teg
r
ated
m
o
v
in
g
av
er
ag
e
with
ex
o
g
en
o
u
s
v
ar
ia
b
le
(
AR
I
MA
X)
an
d
tr
an
s
f
o
r
m
er
-
b
ased
n
etwo
r
k
s
,
to
f
u
r
th
er
en
h
an
ce
f
o
r
ec
asti
n
g
ac
cu
r
ac
y
an
d
g
e
n
er
alis
ab
ilit
y
[
1
5
]
,
[
1
6
]
.
A
n
o
tab
le
r
esear
ch
g
ap
p
e
r
s
is
ts
in
th
e
r
eliab
le
ap
p
licatio
n
o
f
L
STM
ar
ch
itectu
r
es
f
o
r
lo
n
g
-
ter
m
f
o
r
ec
asti
n
g
h
o
r
izo
n
s
.
Alth
o
u
g
h
B
iLST
M
m
o
d
els
p
er
f
o
r
m
ex
ce
p
tio
n
ally
well
i
n
s
h
o
r
t
-
an
d
m
ed
i
u
m
-
ter
m
f
o
r
ec
asts
,
th
eir
p
er
f
o
r
m
a
n
ce
ten
d
s
to
d
eter
io
r
ate
o
v
e
r
lo
n
g
er
tim
e
h
o
r
izo
n
s
,
a
n
d
th
e
r
e
ar
e
n
o
estab
lis
h
ed
r
ec
o
m
m
en
d
atio
n
s
f
o
r
o
p
tim
i
s
in
g
th
ese
m
o
d
els
in
lo
n
g
-
ter
m
s
ce
n
ar
io
s
.
Mo
r
e
th
o
r
o
u
g
h
r
esear
c
h
o
n
ar
ch
itectu
r
al
tu
n
i
n
g
,
i
n
ter
p
r
e
tab
ilit
y
,
an
d
m
eth
o
d
o
lo
g
ical
r
ig
o
u
r
is
n
ee
d
ed
,
ac
c
o
r
d
in
g
to
th
e
liter
atu
r
e
,
p
ar
ticu
lar
ly
in
u
n
d
e
r
s
tu
d
ied
f
ield
s
s
u
ch
as
f
in
an
cial
r
is
k
ass
es
s
m
en
t
an
d
en
v
ir
o
n
m
e
n
tal
m
o
n
ito
r
in
g
.
T
o
ad
d
r
ess
th
e
is
s
u
es
o
f
s
ca
lab
ilit
y
,
tr
an
s
p
ar
en
c
y
,
an
d
lo
n
g
-
ter
m
p
r
ed
ictiv
e
d
ep
en
d
ab
ilit
y
,
f
u
tu
r
e
r
esear
ch
aim
s
to
b
u
ild
ef
f
ec
tiv
e
tr
ai
n
in
g
m
eth
o
d
o
lo
g
ies,
im
p
r
o
v
ed
h
y
b
r
id
s
tr
u
ctu
r
es,
an
d
s
tan
d
a
r
d
is
ed
e
v
alu
atio
n
f
r
am
ewo
r
k
s
[
9
]
,
[
1
7
]
,
[
1
8
]
.
W
h
ile
th
e
s
u
p
er
io
r
ity
o
f
ad
v
an
ce
d
L
STM
an
d
B
iLST
M
ar
ch
itectu
r
es
is
well
d
o
cu
m
en
ted
,
d
ir
ec
t,
h
ea
d
-
to
-
h
ea
d
c
o
m
p
a
r
is
o
n
s
b
e
twee
n
u
n
id
ir
ec
tio
n
al
an
d
Bi
L
STM
m
o
d
els
o
n
b
o
th
s
im
u
la
ted
an
d
r
ea
l
s
to
ck
p
r
ice
d
atasets
r
em
ain
lim
ited
.
R
ec
en
t
s
tu
d
ies
th
at
p
r
o
v
id
e
s
u
ch
co
m
p
ar
is
o
n
s
u
tili
s
e
s
tan
d
ar
d
is
ed
m
etr
ics,
s
u
ch
as
MA
E
,
an
d
r
ep
o
r
t
th
at
B
iLST
M
m
o
d
els
g
en
e
r
ally
ac
h
iev
e
h
ig
h
er
ac
cu
r
ac
y
[
1
9
]
.
Ho
wev
er
,
lim
ited
ex
am
in
atio
n
o
f
h
o
w
th
ese
p
er
f
o
r
m
an
ce
d
if
f
e
r
en
ce
s
d
e
p
en
d
j
o
in
tly
o
n
d
ata
len
g
th
a
n
d
k
ey
h
y
p
er
p
ar
am
eter
s
.
I
n
ad
d
itio
n
,
alth
o
u
g
h
r
ec
en
t
w
o
r
k
h
as
ex
p
lo
r
ed
g
r
id
s
ea
r
ch
an
d
a
d
ap
tiv
e
g
r
ad
ien
t
-
b
a
s
ed
s
tr
ateg
ies
f
o
r
h
y
p
er
p
ar
am
eter
o
p
tim
is
atio
n
in
d
ee
p
lear
n
in
g
m
o
d
els
[
3
]
,
[
2
0
]
,
s
y
s
tem
atic
ev
alu
atio
n
s
th
at
lin
k
th
ese
ch
o
ices
to
n
o
n
lin
ea
r
d
ata
-
g
en
er
atin
g
m
ec
h
an
is
m
s
an
d
v
a
r
y
in
g
s
am
p
le
s
izes a
r
e
s
till
in
s
u
f
f
icien
t.
T
h
is
s
tu
d
y
in
v
esti
g
ated
th
e
ef
f
ec
ts
o
f
m
o
d
el
ar
c
h
itectu
r
e,
s
am
p
le
s
ize,
an
d
h
y
p
er
p
a
r
am
et
er
s
etti
n
g
s
o
n
th
e
f
o
r
ec
asti
n
g
p
er
f
o
r
m
a
n
ce
o
f
L
STM
a
n
d
B
iLST
M
m
o
d
els
f
o
r
n
o
n
lin
ea
r
s
y
n
th
etic
tim
e
s
er
ies.
W
h
ile
ea
r
lier
s
tu
d
ies
h
av
e
ex
p
lo
r
e
d
th
e
p
er
f
o
r
m
an
ce
o
f
L
ST
M
an
d
B
iLST
M
m
o
d
els
in
v
ar
io
u
s
f
o
r
ec
asti
n
g
ap
p
licatio
n
s
,
th
ey
h
a
v
e
n
o
t
ex
p
licitly
ad
d
r
ess
ed
h
o
w
th
ei
r
p
er
f
o
r
m
an
ce
d
i
f
f
er
en
ce
s
ar
e
in
f
lu
en
ce
d
b
y
co
n
tr
o
lled
n
o
n
lin
ea
r
d
ata
-
g
en
er
atin
g
p
r
o
ce
s
s
es,
d
if
f
er
en
t
s
am
p
le
s
izes,
an
d
k
ey
h
y
p
e
r
p
ar
am
eter
s
ettin
g
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
169
6
-
1
7
0
5
1698
T
h
er
ef
o
r
e,
th
is
s
tu
d
y
aim
s
to
d
esig
n
a
u
n
if
ied
ex
p
er
im
en
t
al
f
r
am
ewo
r
k
th
at
co
m
b
in
es
n
o
n
lin
ea
r
T
AR
an
d
SET
AR
s
im
u
latio
n
s
an
d
co
n
d
u
ct
a
s
y
s
tem
atic
g
r
id
s
ea
r
ch
o
v
er
h
id
d
en
n
e
u
r
o
n
s
an
d
ep
o
ch
s
f
o
r
b
o
th
L
STM
an
d
B
iLST
M,
allo
win
g
u
s
to
q
u
an
tify
h
o
w
m
o
d
el
ar
ch
itect
u
r
e
an
d
h
y
p
er
p
a
r
am
eter
s
in
ter
ac
t
with
d
ata
len
g
th
.
T
h
is
in
teg
r
ated
ap
p
r
o
ac
h
p
r
o
v
id
es
p
r
ac
tical,
ev
id
en
ce
-
b
ased
g
u
id
elin
es
f
o
r
s
elec
tin
g
b
o
th
th
e
m
o
d
el
t
y
p
e
an
d
its
h
y
p
er
p
ar
am
ete
r
s
in
tim
e
s
er
ies f
o
r
ec
asti
n
g
ap
p
licatio
n
s
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
2
.
1
.
Unid
irec
t
io
na
l LS
T
M
T
h
e
u
n
i
d
i
r
e
cti
o
n
al
L
STM
a
r
c
h
it
ec
t
u
r
e
a
im
s
to
s
o
l
v
e
t
h
e
v
an
is
h
i
n
g
g
r
a
d
i
e
n
t
p
r
o
b
le
m
o
f
t
en
s
e
en
in
tr
a
d
i
ti
o
n
al
R
N
Ns
[
2
1
]
.
T
h
e
L
STM
u
n
it
i
n
cl
u
d
es
a
c
ell
s
tate
,
in
p
u
t
g
ate
,
o
u
t
p
u
t
g
a
t
e,
a
n
d
f
o
r
g
et
g
ate
,
wh
i
ch
w
o
r
k
to
g
et
h
e
r
to
co
n
t
r
o
l
t
h
e
f
l
o
w
of
i
n
f
o
r
m
at
io
n
.
T
h
e
c
ell
s
ta
te
a
cts
as
a
m
e
m
o
r
y
t
h
at
ca
r
r
i
es
d
at
a
ac
r
o
s
s
tim
e
s
t
e
p
s
,
w
h
il
e
t
h
e
g
a
tes
m
a
n
a
g
e
t
h
e
a
d
d
it
io
n
a
n
d
r
e
m
o
v
al
o
f
i
n
f
o
r
m
ati
o
n
.
T
h
e
f
o
r
g
e
t
g
a
t
e
d
et
e
r
m
i
n
es
w
h
i
c
h
in
f
o
r
m
at
io
n
s
h
o
u
l
d
be
r
e
m
o
v
ed
or
d
is
ca
r
d
e
d
f
r
o
m
t
h
e
p
r
e
v
i
o
u
s
c
ell
s
t
ates
.
It
e
v
a
lu
at
es
t
h
e
c
u
r
r
e
n
t
i
n
p
u
t
alo
n
g
s
i
d
e
th
e
o
u
t
p
u
t
a
n
d
m
em
o
r
y
ce
l
l
s
ta
te
f
r
o
m
t
h
e
p
r
e
v
i
o
u
s
tim
e
s
te
p
to
m
a
k
e
t
h
is
d
e
cisi
o
n
.
T
h
e
f
o
r
g
e
t
g
at
e
was
n
o
t
p
a
r
t
of
t
h
e
o
r
i
g
i
n
al
L
STM
a
r
c
h
it
ec
t
u
r
e
b
u
t
was
a
d
d
e
d
l
ate
r
to
all
o
w
th
e
n
et
wo
r
k
to
r
ese
t
its
s
tat
e.
Nex
t
g
at
e,
in
p
u
t
g
ate
d
ec
i
d
es
wh
a
t
n
ew
in
f
o
r
m
a
ti
o
n
s
h
o
u
l
d
be
r
e
tai
n
ed
a
n
d
a
d
d
e
d
to
t
h
e
n
etw
o
r
k
'
s
ce
ll
s
t
at
es.
It
d
e
te
r
m
in
es
t
h
e
a
cti
v
at
io
n
v
a
lu
es
a
n
d
h
el
p
s
s
ele
ct
th
e
c
a
n
d
i
d
at
e
v
a
lu
es
t
h
a
t
c
o
u
l
d
p
o
te
n
t
ia
lly
be
i
n
c
o
r
p
o
r
a
te
d
in
t
o
t
h
e
m
em
o
r
y
ce
ll
.
Fi
n
al
ly
,
t
h
e
o
u
t
p
u
t
g
at
e
ca
l
c
u
la
tes
t
h
e
f
in
al
o
u
t
p
u
t
of
t
h
e
L
S
T
M
b
l
o
c
k
f
o
r
t
h
e
cu
r
r
e
n
t
ite
r
at
io
n
.
It
d
o
es
t
h
is
by
co
m
b
in
i
n
g
th
e
c
u
r
r
e
n
t
i
n
p
u
t
an
d
t
h
e
p
r
e
v
i
o
u
s
u
n
it
o
u
t
p
u
t
w
it
h
t
h
e
n
ew
ly
u
p
d
ate
d
c
ell
v
al
u
e.
T
h
is
d
esi
g
n
e
n
a
b
l
es
L
S
T
Ms
to
r
et
ai
n
r
ele
v
an
t
i
n
f
o
r
m
a
tio
n
o
v
e
r
l
o
n
g
s
eq
u
e
n
ce
s
,
m
ak
in
g
t
h
e
m
es
p
e
cia
ll
y
ef
f
e
cti
v
e
f
o
r
ti
m
e
s
e
r
ies
f
o
r
ec
a
s
tin
g
tas
k
s
.
Fi
g
u
r
e
1
s
h
o
ws
th
e
a
r
c
h
it
ec
t
u
r
e
of
t
h
e
L
S
T
M
.
Fig
u
r
e
1.
T
h
e
ar
ch
itectu
r
e
of
u
n
id
ir
ec
tio
n
al
L
STM
In
th
is
s
tu
d
y
,
th
e
L
STM
m
o
d
el
is
im
p
lem
en
ted
u
s
in
g
a
s
in
g
le
lay
er
of
L
STM
u
n
its
,
f
o
ll
o
wed
by
a
f
u
lly
co
n
n
ec
ted
la
y
er
th
at
o
u
t
p
u
ts
th
e
f
o
r
ec
asted
v
alu
es.
T
h
e
m
o
d
el
is
tr
ain
ed
u
s
in
g
a
m
ea
n
s
q
u
ar
e
d
er
r
o
r
(
MS
E)
lo
s
s
f
u
n
ctio
n
,
w
h
ich
q
u
an
tifie
s
th
e
d
if
f
er
e
n
ce
b
etwe
en
th
e
p
r
e
d
icted
an
d
ac
tu
al
v
alu
es.
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
in
v
o
lv
es
b
ac
k
p
r
o
p
a
g
a
tio
n
th
r
o
u
g
h
tim
e
(
B
PTT
)
,
w
h
er
e
th
e
g
r
ad
ien
ts
ar
e
co
m
p
u
te
d
an
d
u
p
d
ated
u
s
in
g
an
o
p
tim
is
atio
n
alg
o
r
ith
m
,
th
e
Ad
am
o
p
tim
is
er
.
2
.
2
.
B
idi
re
ct
io
na
l LS
T
M
T
h
e
B
iLST
M
ar
ch
itectu
r
e
e
x
ten
d
s
th
e
ca
p
ab
ilit
ies
of
th
e
s
tan
d
ar
d
L
STM
by
p
r
o
ce
s
s
in
g
in
p
u
t
s
eq
u
en
ce
s
in
b
o
th
f
o
r
war
d
an
d
b
ac
k
war
d
d
ir
ec
tio
n
s
[
2
2
]
.
T
h
is
d
u
al
p
r
o
ce
s
s
in
g
en
ab
les
th
e
m
o
d
el
to
ca
p
tu
r
e
co
n
tex
tu
al
in
f
o
r
m
atio
n
f
r
o
m
b
o
th
p
ast
an
d
f
u
tu
r
e
tim
e
s
tep
s
,
wh
ich
c
an
be
esp
ec
ially
u
s
ef
u
l
in
tim
e
s
er
ie
s
f
o
r
ec
asti
n
g
,
wh
e
r
e
f
u
t
u
r
e
co
n
t
ex
t
m
ay
o
f
f
er
v
alu
ab
le
in
s
ig
h
t
s
in
to
th
e
cu
r
r
en
t
s
tate.
T
h
e
B
iLST
M
co
n
s
is
ts
of
two
s
ep
ar
ate
L
STM
lay
er
s
,
w
h
er
e
one
p
r
o
ce
s
s
es
th
e
in
p
u
t
s
eq
u
en
ce
f
r
o
m
s
tar
t
to
f
in
is
h
(
f
o
r
war
d
L
STM
)
,
an
d
an
o
th
er
p
r
o
ce
s
s
es
it
f
r
o
m
en
d
to
s
tar
t
(
b
ac
k
war
d
L
STM
)
.
T
h
e
o
u
tp
u
ts
f
r
o
m
b
o
th
lay
e
r
s
ar
e
th
en
co
n
ca
ten
ated
to
m
ain
tain
tem
p
o
r
al
r
elatio
n
s
h
ip
s
.
Fig
u
r
e
2
illu
s
tr
ates
th
e
ar
ch
itectu
r
e
of
t
h
e
B
iLST
M.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Un
id
ir
ec
tio
n
a
l a
n
d
b
id
ir
ec
tio
n
a
l LS
TM
fo
r
n
o
n
lin
ea
r
s
yn
th
etic
time
s
erie
s
… (
N
u
r
Ha
iz
u
m
A
b
d
R
a
h
ma
n
)
1699
Fig
u
r
e
2.
T
h
e
ar
c
h
itectu
r
e
of
Bi
L
STM
Similar
to
th
e
L
STM
m
o
d
el,
th
e
B
iLST
M
is
tr
ain
ed
u
s
in
g
a
MSE
lo
s
s
f
u
n
ctio
n
a
n
d
o
p
tim
is
ed
th
r
o
u
g
h
b
ac
k
p
r
o
p
ag
atio
n
.
T
h
e
ar
ch
itectu
r
e
m
ay
i
n
clu
d
e
ad
d
itio
n
al
lay
er
s
,
s
u
ch
as
d
r
o
p
o
u
t
lay
er
s
,
to
p
r
ev
e
n
t
o
v
er
f
itti
n
g
an
d
im
p
r
o
v
e
g
e
n
er
alis
atio
n
.
T
h
e
ch
o
ice
of
h
y
p
er
p
ar
am
eter
s
u
s
ed
in
th
e
B
iLST
M
in
th
is
p
ap
er
is
th
e
s
am
e
as
th
o
s
e
u
s
ed
in
th
e
L
STM
m
o
d
el.
2
.
3
.
F
o
re
ca
s
t
a
cc
ura
c
y
m
ea
s
ure
Ma
n
y
m
etr
ics
ca
n
be
u
s
ed
to
ass
ess
th
e
ac
cu
r
ac
y
of
p
r
ed
i
ctio
n
m
o
d
els.
MSE
is
o
f
te
n
u
tili
s
ed
in
s
tu
d
ies
to
m
ea
s
u
r
e
ac
cu
r
ac
y
.
MSE
m
ea
s
u
r
es
th
e
d
if
f
er
en
ce
s
b
etwe
en
th
e
ac
tu
al
v
alu
es
a
n
d
p
r
e
d
icted
v
alu
es.
MSE
m
ea
s
u
r
es
th
e
er
r
o
r
s
’
v
a
r
iab
ilit
y
.
Hen
ce
,
th
e
s
m
aller
t
h
e
MSE
,
th
e
b
etter
t
h
e
p
r
e
d
i
ctio
n
p
er
f
o
r
m
an
ce
.
On
e
d
r
awb
ac
k
of
MSE
is
th
at
it
is
s
en
s
itiv
e
to
d
ata
tr
an
s
f
o
r
m
atio
n
s
or
c
h
an
g
es
in
sc
a
le,
an
d
it
p
en
alis
es
ex
tr
em
e
er
r
o
r
s
.
T
h
er
ef
o
r
e,
a
n
o
th
er
m
ea
s
u
r
e
u
s
ed
in
th
is
s
tu
d
y
is
th
e
MA
E
.
MA
E
m
e
asu
r
es
th
e
av
er
a
g
e
m
ag
n
itu
d
e
of
th
e
e
r
r
o
r
s
in
th
e
p
r
ed
icted
v
al
u
es.
T
h
e
f
o
r
m
u
la
s
to
co
m
p
u
te
MSE
an
d
MA
E
a
r
e
as
(
1
)
,
(
2
)
:
=
1
∑
(
−
̂
)
2
=
1
(
1
)
=
1
∑
|
−
̂
|
=
1
(
2
)
wh
er
e
is
th
e
ac
tu
al
v
alu
e
at
tim
estep
,
an
d
̂
is
th
e
p
r
ed
icted
v
alu
e
at
tim
estep
.
2
.
4
.
Sim
ula
t
i
o
n study
T
wo
r
eg
im
es
-
s
witch
in
g
n
o
n
-
l
in
ea
r
m
o
d
els,
th
e
th
r
esh
o
ld
au
to
r
eg
r
ess
iv
e
(
T
AR
)
an
d
s
e
lf
-
ex
citin
g
th
r
esh
o
ld
au
to
r
eg
r
ess
iv
e
(
SE
T
AR
)
m
o
d
els,
ar
e
u
s
ed
in
t
h
is
s
im
u
latio
n
s
tu
d
y
to
g
e
n
e
r
ate
non
-
lin
ea
r
tim
e
s
er
ies
d
ata.
T
h
ese
clas
s
ical
r
e
g
im
e
-
s
witch
in
g
m
o
d
els
ar
e
wi
d
ely
u
s
ed
to
m
o
d
el
f
in
an
cial
s
er
ies
with
d
if
f
er
en
t
d
y
n
am
ics
in
“lo
w”
an
d
“h
i
g
h
”
r
eg
im
es.
Sto
ck
p
r
ices
o
f
ten
ex
h
ib
it
n
o
n
lin
ea
r
b
e
h
av
io
u
r
,
asy
m
m
etr
ic
ad
ju
s
tm
en
t
an
d
r
e
g
im
e,
w
h
ich
can
be
ca
p
tu
r
ed
by
allo
win
g
th
e
au
to
r
eg
r
ess
iv
e
p
a
r
am
eter
s
to
c
h
an
g
e
wh
en
a
th
r
esh
o
ld
is
cr
o
s
s
ed
[
2
3
]
.
Sp
ec
if
ically
,
th
is
s
tu
d
y
co
n
s
id
er
a
two
-
r
eg
im
e
T
AR
(
2
;1
,
1
)
m
o
d
el
as
(
3
)
:
=
{
−
0
.
2
−
1
+
,
∣
−
1
∣
≤
1
,
1
.
0
−
1
+
,
∣
−
1
∣
>
1
,
(
3
)
an
d
a
two
-
r
eg
im
e
SET
AR
(
2
;1
,
1
)
m
o
d
el
as
(
4
)
:
=
{
0
.
19
+
0
.
92
−
1
+
,
∣
−
1
∣
≤
2
.
5
,
0
.
78
+
0
.
72
−
1
+
,
∣
−
1
∣
>
2
.
5
,
(
4
)
wh
er
e
∼
(
0
,
1
)
.
T
h
e
au
to
r
eg
r
ess
iv
e
co
ef
f
icien
ts
ar
e
ch
o
s
en
in
th
e
i
n
t
er
v
al
[
−
1
,
1
]
to
in
d
u
ce
n
o
n
lin
ea
r
b
u
t
s
tatio
n
ar
y
d
y
n
am
ics th
at
r
esem
b
le
s
to
ck
p
r
ice
b
eh
av
i
o
u
r
[
2
3
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
169
6
-
1
7
0
5
1700
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
E
ac
h
n
o
n
-
lin
ea
r
m
o
d
el
s
im
u
lated
th
e
d
ataset
with
two
d
i
f
f
er
en
t
s
am
p
le
s
izes,
1
0
0
0
a
n
d
3
0
0
0
,
r
esp
ec
tiv
ely
.
Fo
r
th
e
n
etwo
r
k
h
y
p
e
r
p
ar
am
ete
r
s
,
h
id
d
e
n
n
eu
r
o
n
s
an
d
ep
o
ch
s
,
th
r
ee
d
if
f
er
en
t
n
u
m
b
er
s
wer
e
u
s
ed
f
o
r
ea
ch
m
o
d
el.
C
o
n
s
eq
u
en
tly
,
th
e
to
tal
n
u
m
b
er
o
f
co
m
b
in
atio
n
s
f
o
r
s
im
u
lated
d
ata
an
d
h
y
p
e
r
p
ar
am
eter
co
n
f
ig
u
r
atio
n
s
is
1
8
f
o
r
b
o
th
t
h
e
L
STM
a
n
d
B
iLST
M
m
o
d
e
ls
.
T
h
e
T
AR
an
d
SET
AR
s
im
u
lated
s
er
ies
wer
e
r
ep
licated
3
0
tim
es
u
s
in
g
in
d
ep
en
d
en
t
r
a
n
d
o
m
d
r
aws
f
r
o
m
th
e
d
ef
au
lt
p
s
eu
d
o
-
r
an
d
o
m
n
u
m
b
er
g
e
n
er
ato
r
in
R
,
an
d
th
e
r
e
p
o
r
ted
MSE
an
d
MA
E
ar
e
av
e
r
ag
es
o
v
er
th
ese
3
0
r
e
p
licatio
n
s
.
T
h
e
er
r
o
r
ter
m
is
ass
u
m
ed
to
b
e
in
d
ep
en
d
en
t
an
d
id
en
tically
d
is
tr
ib
u
ted
with
(
0
,
1
)
.
T
h
e
d
atasets
ar
e
s
p
lit
at
8
0
:2
0
,
with
8
0
%
u
s
ed
f
o
r
tr
ain
in
g
an
d
2
0
%
f
o
r
test
in
g
.
Hen
ce
,
th
e
test
in
g
d
atasets
f
o
r
ea
c
h
s
am
p
le
s
ize
will
b
e
2
0
0
an
d
6
0
0
.
T
h
e
tr
ain
in
g
d
ataset
is
u
s
ed
to
tr
ain
an
d
r
u
n
th
r
o
u
g
h
L
STM
an
d
B
iLST
M
m
o
d
els.
T
h
e
tr
ain
ed
m
o
d
els
wer
e
th
en
u
s
ed
to
p
r
ed
ict
s
to
ck
p
r
ices,
an
d
th
eir
p
r
ed
ictio
n
ac
c
u
r
ac
y
was
ev
alu
ated
.
T
ab
le
1
s
u
m
m
ar
is
es
th
e
h
y
p
er
p
a
r
am
eter
v
alu
es u
s
ed
f
o
r
L
STM
an
d
B
iLST
M
in
th
is
s
tu
d
y
.
T
ab
le
1.
Hy
p
er
p
ar
a
m
eter
g
r
i
d
s
ea
r
ch
v
alu
es
of
L
STM
a
n
d
B
iLST
M
H
y
p
e
r
p
a
r
a
m
e
t
e
r
s
V
a
l
u
e
s
H
i
d
d
e
n
n
e
u
r
o
n
3
2
,
6
4
,
1
2
8
H
i
d
d
e
n
l
a
y
e
r
1
O
u
t
p
u
t
n
e
u
r
o
n
1
Ep
o
c
h
2
0
,
5
0
,
1
0
0
B
a
t
c
h
si
z
e
32
O
p
t
i
mi
z
e
r
A
d
a
p
t
i
v
e
m
o
me
n
t
e
s
t
i
m
a
t
i
o
n
(
A
d
a
m)
Lo
ss
f
u
n
c
t
i
o
n
M
S
E
Gr
id
s
ea
r
ch
was
s
elec
ted
f
o
r
h
y
p
er
p
a
r
am
eter
o
p
tim
is
atio
n
b
e
ca
u
s
e
th
is
s
tu
d
y
in
v
o
lv
ed
a
lim
ited
an
d
p
r
ed
ef
in
e
d
s
et
o
f
h
y
p
er
p
ar
a
m
eter
v
alu
es.
T
h
is
a
p
p
r
o
ac
h
al
lo
ws
ea
ch
co
m
b
i
n
atio
n
o
f
h
i
d
d
en
n
eu
r
o
n
s
an
d
ep
o
ch
s
to
b
e
e
v
alu
ated
s
y
s
te
m
atica
lly
f
o
r
b
o
th
L
STM
an
d
B
iLST
M
m
o
d
els.
T
h
e
m
ai
n
ad
v
a
n
tag
e
o
f
g
r
id
s
ea
r
ch
in
th
is
s
tu
d
y
is
it
s
tr
an
s
p
ar
en
cy
an
d
r
ep
r
o
d
u
ci
b
ilit
y
,
as
it
en
ab
les
a
f
air
co
m
p
ar
is
o
n
o
f
m
o
d
e
l
p
er
f
o
r
m
an
ce
ac
r
o
s
s
th
e
s
am
e
h
y
p
er
p
ar
am
eter
s
ettin
g
s
.
Alth
o
u
g
h
m
o
r
e
a
d
ap
tiv
e
m
eth
o
d
s
,
s
u
ch
as
B
ay
esian
o
p
tim
izat
io
n
a
n
d
m
etah
eu
r
is
tic
alg
o
r
ith
m
s
,
m
ay
b
e
m
o
r
e
ef
f
icien
t
f
o
r
lar
g
er
a
n
d
m
o
r
e
co
m
p
le
x
s
ea
r
ch
s
p
ac
es,
g
r
id
s
ea
r
ch
an
d
its
v
ar
ian
ts
r
em
ain
r
elev
a
n
t
f
o
r
L
STM
tu
n
in
g
wh
e
n
th
e
h
y
p
er
p
ar
am
eter
s
p
ac
e
is
s
m
all
an
d
well
-
d
ef
in
ed
[
2
4
]
.
T
h
er
ef
o
r
e,
g
r
id
s
ea
r
ch
is
ap
p
r
o
p
r
iate
in
th
is
co
n
tr
o
lled
s
im
u
latio
n
s
ettin
g
b
ec
au
s
e
th
e
s
tu
d
y
aim
s
to
e
x
a
m
in
e
th
e
ef
f
ec
t
o
f
ea
c
h
s
elec
ted
co
n
f
i
g
u
r
atio
n
o
n
f
o
r
ec
asti
n
g
ac
cu
r
ac
y
.
Data
tr
an
s
f
o
r
m
atio
n
was
p
er
f
o
r
m
ed
t
o
en
h
an
ce
t
h
e
lear
n
i
n
g
o
f
th
e
L
STM
a
n
d
B
iLST
M
m
o
d
els.
Min
-
m
ax
n
o
r
m
alis
atio
n
was
c
o
n
d
u
cte
d
f
ir
s
t
in
8
0
%
tr
ain
in
g
,
en
s
u
r
in
g
th
e
d
ata
r
an
g
ed
f
r
o
m
0
to
1
,
a
n
d
t
h
e
s
am
e
s
ca
lin
g
p
ar
am
eter
s
wer
e
ap
p
lied
to
tr
an
s
f
o
r
m
th
e
test
s
et.
I
n
ad
d
itio
n
,
th
e
L
STM
an
d
B
iLST
M
n
eu
r
al
n
etwo
r
k
s
f
o
r
ea
ch
h
y
p
er
p
ar
a
m
eter
co
m
b
i
n
atio
n
wer
e
r
ep
li
ca
ted
to
ac
co
u
n
t
f
o
r
th
e
s
to
c
h
asti
c
n
atu
r
e
o
f
th
e
n
eu
r
al
n
etwo
r
k
alg
o
r
ith
m
.
T
h
e
av
er
ag
e
MSE
an
d
MA
E
wer
e
th
en
co
m
p
u
ted
a
n
d
s
h
o
wn
in
T
ab
les 2
an
d
3
.
T
ab
le
2.
MSE
an
d
MA
E
of
L
STM
an
d
B
iLST
M
f
o
r
T
AR
s
im
u
lated
d
ata
M
o
d
e
l
S
a
mp
l
e
s
i
z
e
H
i
d
d
e
n
n
e
u
r
o
n
s
M
S
E
M
A
E
Ep
o
c
h
s
Ep
o
c
h
s
20
50
1
0
0
20
50
1
0
0
LSTM
1
0
0
0
32
5
.
8
5
0
0
3
.
5
9
5
5
2
.
5
1
0
2
1
.
8
9
8
5
1
.
4
8
1
1
1
.
2
1
7
7
64
5
.
0
6
1
3
2
.
9
7
6
1
2
.
0
3
0
8
1
.
7
5
9
0
1
.
3
4
9
2
1
.
0
3
5
1
1
2
8
4
.
2
5
0
7
2
.
5
3
0
5
1
.
9
0
0
4
1
.
6
1
8
4
1
.
2
5
1
4
1
.
0
7
2
6
3
0
0
0
32
3
.
7
9
9
0
2
.
2
0
5
5
1
.
7
7
9
8
1
.
5
2
5
9
1
.
1
6
7
0
1
.
0
2
6
7
64
3
.
2
5
1
4
1
.
9
0
2
1
1
.
5
2
4
4
1
.
4
1
6
8
1
.
0
7
9
9
0
.
9
7
4
0
1
2
8
2
.
8
6
3
5
1
.
6
3
8
4
1
.
4
8
7
2
1
.
3
3
1
7
1
.
0
1
3
4
0
.
9
6
5
7
B
i
LST
M
1
0
0
0
32
3
.
7
8
5
2
2
.
4
4
4
5
1
.
9
1
1
1
1
.
5
2
0
1
1
.
2
2
1
1
1
.
0
8
2
8
64
3
.
1
0
0
1
2
.
1
2
1
3
1
.
8
1
2
2
1
.
3
7
9
8
1
.
1
3
9
9
1
.
0
5
0
1
1
2
8
2
.
6
9
2
6
1
.
9
3
9
0
1
.
7
2
0
0
1
.
2
9
6
7
1
.
0
8
8
8
1
.
0
3
6
3
3
0
0
0
32
2
.
7
2
8
4
1
.
8
7
9
7
1
.
6
9
7
9
1
.
2
9
7
3
1
.
0
8
5
3
1
.
0
2
4
5
64
2
.
3
3
2
7
1
.
8
0
9
5
1
.
5
0
3
9
1
.
2
0
3
7
1
.
0
5
8
2
0
.
9
7
1
1
1
2
8
2
.
2
6
7
5
1
.
5
6
2
6
1
.
4
5
8
5
1
.
1
9
2
7
0
.
9
9
1
0
0
.
9
6
0
1
Fo
r
th
e
T
AR
s
im
u
latio
n
s
r
esu
l
t
(
r
ef
er
to
T
ab
le
2
)
,
th
e
b
est
-
p
er
f
o
r
m
in
g
L
STM
co
n
f
ig
u
r
atio
n
f
o
r
b
o
t
h
s
am
p
le
s
izes
is
th
e
m
o
d
el
with
1
2
8
h
id
d
en
n
eu
r
o
n
s
an
d
1
0
0
ep
o
ch
s
,
y
ield
in
g
MSE
o
f
1
.
9
0
0
4
an
d
MA
E
o
f
1
.
0
7
2
6
f
o
r
=
1000
an
d
MSE
o
f
1
.
4
8
7
2
an
d
MA
E
o
f
0
.
9
6
5
7
f
o
r
=
3000
as
s
h
o
wn
in
Fig
u
r
e
3
.
T
h
e
B
iLST
M
m
o
d
el
s
h
o
ws
th
e
s
am
e
o
p
tim
al
co
n
f
ig
u
r
atio
n
,
with
s
lig
h
tly
lo
wer
er
r
o
r
s
.
T
h
e
MSE
s
h
o
wed
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Un
id
ir
ec
tio
n
a
l a
n
d
b
id
ir
ec
tio
n
a
l LS
TM
fo
r
n
o
n
lin
ea
r
s
yn
th
etic
time
s
erie
s
… (
N
u
r
Ha
iz
u
m
A
b
d
R
a
h
ma
n
)
1701
v
alu
e
o
f
1
.
7
2
0
0
an
d
MA
E
o
f
1
.
0
3
6
3
f
o
r
=
1000
,
an
d
MSE
o
f
1
.
4
5
8
5
an
d
MA
E
o
f
0
.
9
6
0
1
f
o
r
=
3000
.
Fig
u
r
es
3
(
a)
a
n
d
3
(
b
)
s
h
o
w
th
at
in
cr
ea
s
in
g
th
e
n
u
m
b
er
o
f
h
i
d
d
en
n
eu
r
o
n
s
an
d
ep
o
c
h
s
g
en
e
r
ally
d
ec
r
ea
s
es
th
e
MSE
f
o
r
b
o
t
h
ar
ch
itectu
r
es.
(
a)
(
b
)
Fig
u
r
e
3.
E
f
f
ec
t
of
h
id
d
e
n
n
e
u
r
o
n
s
an
d
e
p
o
ch
s
on
(
a
)
L
STM
an
d
(
b
)
B
iLST
M
p
er
f
o
r
m
an
ce
f
o
r
T
AR
-
s
im
u
lated
d
ata
T
ab
le
3.
MSE
an
d
MA
E
of
L
STM
an
d
B
iLST
M
f
o
r
SET
AR
s
im
u
lated
d
ata
M
o
d
e
l
S
a
mp
l
e
s
i
z
e
H
i
d
d
e
n
n
e
u
r
o
n
s
M
S
E
M
A
E
Ep
o
c
h
s
Ep
o
c
h
s
20
50
1
0
0
20
50
1
0
0
LSTM
1
0
0
0
32
2
.
0
3
1
7
1
.
3
5
5
4
1
.
0
6
9
3
1
.
1
2
7
9
0
.
9
2
3
3
0
.
8
2
5
6
64
1
.
8
2
4
1
1
.
2
0
5
0
1
.
0
4
9
8
1
.
0
6
9
3
0
.
8
7
2
7
0
.
8
2
3
1
1
2
8
1
.
6
3
3
1
1
.
1
3
6
9
1
.
0
4
9
2
1
.
0
1
5
0
0
.
8
5
0
4
0
.
8
2
2
7
3
0
0
0
32
1
.
2
8
9
7
1
.
0
6
8
5
1
.
0
5
0
5
0
.
9
0
8
4
0
.
8
2
8
7
0
.
8
2
2
4
64
1
.
1
5
2
6
1
.
0
6
7
2
1
.
0
4
9
6
0
.
8
5
9
6
0
.
8
2
8
5
0
.
8
2
2
2
1
2
8
1
.
1
1
6
2
1
.
0
6
5
9
1
.
0
4
8
5
0
.
8
4
5
9
0
.
8
2
8
3
0
.
8
2
1
8
B
i
LST
M
1
0
0
0
32
1
.
4
5
2
9
1
.
1
7
0
2
1
.
0
6
4
0
0
.
9
5
8
1
0
.
8
6
0
9
0
.
8
2
3
3
64
1
.
3
2
9
4
1
.
1
3
0
7
1
.
0
4
8
5
0
.
9
1
7
7
0
.
8
4
8
6
0
.
8
1
8
8
1
2
8
1
.
2
6
6
5
1
.
1
0
7
6
1
.
0
4
8
0
0
.
8
9
7
3
0
.
8
4
0
3
0
.
8
1
8
6
3
0
0
0
32
1
.
1
6
8
8
1
.
0
6
4
8
1
.
0
5
0
3
0
.
8
6
5
3
0
.
8
2
7
7
0
.
8
2
2
3
64
1
.
1
3
4
8
1
.
0
6
4
6
1
.
0
4
7
4
0
.
8
5
2
7
0
.
8
2
7
4
0
.
8
2
0
4
1
2
8
1
.
1
0
3
7
1
.
0
6
2
2
1
.
0
4
4
1
0
.
8
4
1
6
0
.
8
2
6
4
0
.
8
1
9
9
A
s
im
ilar
p
atter
n
is
o
b
s
er
v
e
d
f
o
r
t
h
e
SET
AR
s
im
u
latio
n
s
(
T
ab
le
3
)
.
Fo
r
b
o
th
s
am
p
le
s
izes,
th
e
L
STM
m
o
d
el
with
1
2
8
h
id
d
e
n
n
e
u
r
o
n
s
an
d
1
0
0
ep
o
ch
s
ag
ai
n
g
iv
es
th
e
lo
west
er
r
o
r
s
,
wit
h
MSE
=
1
.
0
4
9
2
a
n
d
MA
E
=
0
.
8
2
2
7
f
o
r
=
1000
,
an
d
MSE
=
1
.
0
4
8
5
an
d
MA
E
=
0
.
8
2
1
8
f
o
r
=
3000
.
T
h
e
co
r
r
esp
o
n
d
in
g
B
iLST
M
co
n
f
ig
u
r
atio
n
y
ield
s
s
lig
h
tly
s
m
aller
er
r
o
r
s
,
with
MS
E
=
1
.
0
4
8
0
an
d
MA
E
=
0
.
8
1
8
6
f
o
r
=
1000
,
an
d
MS
E
=
1
.
0
4
4
1
a
n
d
MA
E
=
0
.
8
1
9
9
f
o
r
=
3000
as
s
h
o
wn
in
Fig
u
r
e
4
.
Fig
u
r
es
4
(
a)
an
d
4
(
b
)
co
n
f
ir
m
th
at,
ac
r
o
s
s
b
o
th
T
AR
an
d
SET
AR
m
o
d
els,
MSE
an
d
MA
E
d
ec
r
ea
s
e
as
th
e
n
u
m
b
er
o
f
h
id
d
en
n
e
u
r
o
n
s
in
c
r
ea
s
es
f
r
o
m
3
2
to
6
4
to
1
2
8
a
n
d
as
t
h
e
n
u
m
b
e
r
o
f
ep
o
c
h
s
in
cr
ea
s
es
f
r
o
m
2
0
t
o
5
0
to
1
0
0
.
I
n
a
d
d
itio
n
,
th
e
lar
g
er
s
am
p
le
s
ize
(
=
3000
)
c
o
n
s
is
ten
tly
y
ield
s
lo
wer
er
r
o
r
s
th
an
=
1000
,
an
d
f
o
r
=
3000
th
e
p
er
f
o
r
m
a
n
ce
d
i
f
f
er
en
ce
s
b
etwe
en
n
eig
h
b
o
u
r
in
g
h
y
p
e
r
p
ar
am
eter
s
ettin
g
s
b
ec
o
m
e
r
elativ
ely
s
m
all
f
o
r
b
o
th
L
ST
M
an
d
B
iLST
M.
Acr
o
s
s
b
o
th
T
AR
an
d
SET
AR
,
in
cr
ea
s
in
g
th
e
s
am
p
le
s
ize
f
r
o
m
=
100
0
to
=
3000
r
ed
u
ce
s
MSE
an
d
MA
E
f
o
r
b
o
th
ar
ch
itectu
r
es,
co
n
s
is
ten
t
with
im
p
r
o
v
ed
p
a
r
am
eter
esti
m
atio
n
an
d
r
ed
u
ce
d
g
en
er
alis
atio
n
er
r
o
r
u
n
d
er
l
ar
g
er
tr
ain
in
g
s
ets.
Ho
wev
er
,
th
e
ex
ten
t
of
im
p
r
o
v
em
e
n
t
d
ep
en
d
s
on
th
e
d
ata
-
g
en
er
atin
g
p
r
o
ce
s
s
.
Fo
r
T
AR
s
im
u
latio
n
,
th
e
b
en
ef
its
s
tr
o
n
g
ly
f
r
o
m
in
cr
ea
s
ed
ca
p
ac
ity
an
d
tr
ain
in
g
m
o
r
e
n
eu
r
o
n
s
an
d
ep
o
ch
s
,
an
d
s
h
o
ws
a
clea
r
er
ad
v
an
tag
e
f
o
r
B
iLST
M
esp
ec
ially
at
s
m
aller
s
am
p
le
s
ize.
In
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
169
6
-
1
7
0
5
1702
SET
AR
s
im
u
latio
n
,
th
e
r
esu
lts
ten
d
to
co
n
v
er
g
e
m
o
r
e
q
u
ick
ly
an
d
f
o
r
m
a
more
s
tab
l
e
p
atter
n
with
f
ew
ch
an
g
es.
T
h
is
m
ea
n
s
th
at
o
n
ce
th
e
n
u
m
b
er
of
e
p
o
ch
s
an
d
n
e
u
r
o
n
s
r
ea
ch
es
m
o
d
er
ate
-
to
-
h
ig
h
lev
els,
d
if
f
er
en
ce
s
b
etwe
en
ar
ch
itect
u
r
es
b
ec
o
m
e
less
s
ig
n
if
ican
t,
in
d
icatin
g
a
s
m
o
o
th
e
r
,
more
ef
f
icien
t
p
r
o
ce
s
s
.
To
f
u
r
th
er
s
u
p
p
o
r
t
th
is
,
a
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
w
as
co
n
d
u
cted
to
v
er
if
y
th
at
th
e
B
iLST
M
m
o
d
el
o
u
tp
er
f
o
r
m
ed
th
e
L
STM
m
o
d
el
in
f
o
r
ec
asti
n
g
th
e
tim
e
s
er
ie
s
d
ata.
T
h
e
r
esu
lts
,
p
r
esen
ted
in
T
ab
le
4,
in
d
icate
th
at
B
iLST
M
s
ig
n
if
ica
n
tly
o
u
tp
er
f
o
r
m
s
L
STM
f
o
r
b
o
th
T
A
R
an
d
SET
AR
s
im
u
lated
d
atasets
at
s
am
p
le
s
ize
s
of
1
0
0
0
an
d
3
0
0
0
.
All
-
v
alu
es
ar
e
b
el
o
w
0
.
0
5
,
co
n
f
ir
m
in
g
t
h
at
th
e
d
if
f
er
e
n
ce
s
in
f
o
r
ec
asti
n
g
er
r
o
r
s
b
etwe
en
L
STM
an
d
B
iLST
M
ar
e
s
tatis
t
ically
s
ig
n
if
ican
t.
(
a)
(
b
)
Fig
u
r
e
4
.
E
f
f
ec
t
of
h
id
d
e
n
n
e
u
r
o
n
s
an
d
e
p
o
ch
s
on
(
a
)
L
STM
an
d
(
b
)
B
iLST
M
p
er
f
o
r
m
an
ce
f
o
r
SET
AR
-
s
im
u
lated
d
ata
T
ab
le
4.
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
r
esu
lts
b
etwe
en
L
STM
an
d
B
iLST
M
S
a
m
p
l
e
s
i
z
e
M
o
d
e
l
M
e
a
su
r
e
me
n
t
u
s
e
d
-
v
a
l
u
e
1
0
0
0
TA
R
M
S
E
0
.
0
0
2
0
M
A
E
0
.
0
0
3
9
3
0
0
0
TA
R
M
S
E
0
.
0
0
2
0
M
A
E
0
.
0
0
2
0
1
0
0
0
S
ETA
R
M
S
E
0
.
0
0
2
0
M
A
E
0
.
0
0
2
0
3
0
0
0
S
ETA
R
M
S
E
0
.
0
0
2
0
M
A
E
0
.
0
0
2
0
Fro
m
a
tu
n
in
g
p
er
s
p
ec
tiv
e,
two
p
r
ac
tical
p
atter
n
s
em
er
g
e
w
h
er
e
ep
o
c
h
s
f
r
o
m
20
to
50,
an
d
50
to
1
0
0
g
en
er
ally
r
ed
u
ce
er
r
o
r
,
but
g
a
in
s
f
r
o
m
50
to
100
ep
o
ch
s
ar
e
s
m
aller
th
an
f
r
o
m
20
to
50
in
s
ev
er
al
s
ettin
g
s
,
s
u
g
g
esti
n
g
p
o
ten
tial
o
v
er
tr
ain
i
n
g
r
is
k
or
co
n
v
er
g
e
n
ce
s
atu
r
at
io
n
.
On
th
e
o
th
er
h
an
d
,
h
id
d
e
n
n
eu
r
o
n
s
f
r
o
m
32
to
64
a
n
d
64
to
128
im
p
r
o
v
e
p
er
f
o
r
m
an
ce
co
n
s
is
ten
tly
,
with
th
e
lar
g
est
g
ai
ns
o
f
ten
o
cc
u
r
r
in
g
wh
e
n
m
o
v
in
g
f
r
o
m
32
to
6
4
,
an
d
s
m
aller
in
cr
em
en
tal
g
ain
s
f
r
o
m
64
to
128
,
p
ar
ticu
lar
ly
u
n
d
er
=
3000
.
T
h
u
s
,
th
e
r
esu
lt
s
in
d
icate
th
at
B
iLST
M
is
m
o
s
t
b
en
ef
icial
wh
en
th
e
d
ata
ar
e
lim
ited
an
d
th
e
n
o
n
lin
ea
r
r
eg
im
e
-
s
witch
in
g
b
eh
av
io
u
r
is
p
r
o
n
o
u
n
ce
d
T
AR
,
wh
ile
s
tan
d
ar
d
L
STM
ca
n
be
n
e
ar
ly
as
ef
f
ec
tiv
e
f
o
r
SET
AR
-
lik
e
s
er
ies,
esp
ec
ially
at
lar
g
er
s
am
p
le
s
iz
es
wh
er
e
er
r
o
r
p
latea
u
s
ar
e
o
b
s
er
v
ed
.
T
h
is
f
in
d
in
g
is
co
n
s
i
s
ten
t
with
p
r
ev
io
u
s
s
tu
d
ies
r
ep
o
r
tin
g
th
at
B
iLST
M
can
o
u
tp
er
f
o
r
m
s
tan
d
ar
d
L
STM
in
tim
e
s
er
ies
f
o
r
ec
asti
n
g
b
ec
au
s
e
it
lear
n
s
tem
p
o
r
al
in
f
o
r
m
atio
n
f
r
o
m
b
o
th
f
o
r
wa
r
d
a
n
d
b
ac
k
war
d
d
ir
e
ctio
n
s
[
4
]
,
[
6
]
,
[
25]
.
Ho
wev
e
r
,
th
e
p
r
esen
t
r
esu
lts
f
u
r
th
er
s
h
o
w
th
at
th
is
ad
v
a
n
tag
e
is
co
n
d
itio
n
-
d
e
p
en
d
e
n
t,
wh
er
e
B
iLST
M
p
r
o
v
id
es
clea
r
er
im
p
r
o
v
em
e
n
t
f
o
r
T
AR
s
im
u
lated
d
ata
but
o
n
ly
m
ar
g
i
n
al
im
p
r
o
v
e
m
en
t
f
o
r
s
m
o
o
th
er
SET
AR
s
im
u
lated
d
ata,
esp
ec
ially
wh
e
n
th
e
s
am
p
le
s
ize
i
s
lar
g
e
[
2
6
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Un
id
ir
ec
tio
n
a
l a
n
d
b
id
ir
ec
tio
n
a
l LS
TM
fo
r
n
o
n
lin
ea
r
s
yn
th
etic
time
s
erie
s
… (
N
u
r
Ha
iz
u
m
A
b
d
R
a
h
ma
n
)
1703
4.
CO
NCLU
SI
O
N
T
h
is
r
esear
ch
co
n
d
u
cted
e
x
p
e
r
im
en
tal
s
tu
d
ies
on
th
e
a
p
p
lic
atio
n
of
m
ac
h
in
e
lear
n
in
g
to
tim
e
s
er
ies
d
ata.
No
n
lin
ea
r
tim
e
s
er
ies
d
ata
wer
e
g
en
er
ated
f
r
o
m
th
e
T
A
R
an
d
SET
AR
m
o
d
els
to
ca
p
tu
r
e
th
e
n
o
n
lin
ea
r
ity
in
s
to
ck
p
r
ices.
Simu
lated
n
o
n
-
lin
ea
r
tim
e
s
er
ies
d
ata
f
r
o
m
each
m
o
d
el
co
n
s
is
ted
of
two
s
am
p
le
s
izes,
1000
a
n
d
3
0
0
0
.
T
h
e
ex
p
er
im
e
n
t'
s
r
e
s
u
lts
wer
e
r
ep
o
r
ted
,
an
d
th
e
ac
cu
r
ac
y
an
d
p
er
f
o
r
m
an
c
e
of
u
n
id
ir
ec
tio
n
al
L
STM
s
an
d
B
iLST
Ms
wer
e
co
m
p
ar
ed
an
d
a
n
aly
s
ed
.
Gr
id
s
ea
r
ch
was
u
s
ed
to
tu
n
e
th
e
h
y
p
er
p
ar
am
eter
s
of
L
STM
an
d
B
iLST
M
m
o
d
els
to
o
p
tim
is
e
th
e
n
e
u
r
al
n
etwo
r
k
'
s
p
er
f
o
r
m
an
ce
.
T
h
e
m
ain
h
y
p
er
p
ar
am
eter
tu
n
ed
in
th
e
ex
p
er
im
e
n
t
is
th
e
n
u
m
b
er
of
h
id
d
en
n
e
u
r
o
n
s
,
with
v
alu
es
32,
6
4
,
an
d
128.
T
h
e
r
esu
l
ts
of
th
e
ex
p
er
im
e
n
t
s
h
o
wed
th
at
th
e
MSE
a
n
d
M
AE
v
alu
es
d
ec
r
ea
s
e
as
th
e
n
u
m
b
er
of
h
id
d
en
n
eu
r
o
n
s
in
c
r
ea
s
es.
T
h
e
o
p
tim
al
n
u
m
b
er
of
h
id
d
en
n
e
u
r
o
n
s
is
128,
as
it
p
r
o
v
id
es
th
e
h
i
g
h
est
p
r
ed
ictio
n
ac
cu
r
ac
y
.
Hid
d
e
n
n
e
u
r
o
n
s
ar
e
ar
tific
ial
n
eu
r
o
n
s
th
at
r
ec
eiv
e
weig
h
te
d
in
p
u
ts
an
d
g
en
er
ate
an
o
u
t
p
u
t
v
ia
an
ac
tiv
atio
n
f
u
n
ctio
n
.
Alth
o
u
g
h
a
s
m
all
n
u
m
b
er
of
h
id
d
en
n
e
u
r
o
n
s
m
ay
im
p
r
o
v
e
g
en
er
alis
atio
n
,
it
m
ay
ca
u
s
e
th
e
n
eu
r
al
n
etwo
r
k
to
lo
s
e
th
e
ca
p
ac
ity
to
s
o
lv
e
th
e
p
r
o
b
lem
.
Hen
ce
,
it
is
es
s
en
tial
to
d
eter
m
in
e
th
e
o
p
tim
al
n
u
m
b
e
r
of
h
i
d
d
en
n
e
u
r
o
n
s
by
a
n
aly
s
in
g
d
if
f
er
en
t
d
atasets
,
as
th
e
n
etw
o
r
k
ar
ch
itectu
r
e,
th
e
d
e
g
r
ee
of
n
o
is
e,
an
d
th
e
co
m
p
lex
ity
of
th
e
f
u
n
ctio
n
s
h
o
u
ld
be
co
n
s
id
er
ed
.
In
th
is
s
tu
d
y
,
r
esu
lts
s
h
o
wed
th
at
a
h
ig
h
er
n
u
m
b
er
of
h
i
d
d
en
n
eu
r
o
n
s
(
1
2
8
)
is
p
r
ef
er
r
ed
f
o
r
f
o
r
ec
asti
n
g
s
to
ck
p
r
ice
tim
e
s
er
ies
d
ata,
as
it
ac
h
iev
es
th
e
h
ig
h
est
p
r
ed
ictio
n
ac
c
u
r
ac
y
.
T
h
e
r
esu
lts
s
h
o
wed
th
at
s
am
p
le
s
ize
af
f
ec
ted
th
e
p
er
f
o
r
m
an
ce
of
L
STM
an
d
B
iLST
M
m
o
d
els.
B
o
th
m
o
d
els
p
er
f
o
r
m
e
d
b
etter
at
f
o
r
ec
asti
n
g
s
im
u
lated
d
ata
with
a
3
0
0
0
-
s
am
p
le
s
ize
th
a
n
with
a
1
0
0
0
-
s
am
p
le
s
ize.
B
iLST
M
m
o
d
els
ar
e
n
eu
r
al
n
et
wo
r
k
s
th
at
ar
e
w
ell
k
n
o
wn
f
o
r
th
eir
ab
ilit
y
to
lear
n
lo
n
g
-
ter
m
d
ep
en
d
e
n
cies
in
d
ata.
Ad
d
itio
n
ally
,
th
e
co
m
p
ar
is
o
n
b
etwe
en
th
e
L
STM
an
d
B
iLST
M
m
o
d
els
s
h
o
wed
th
at
th
e
B
iLST
M
m
o
d
el
ac
h
iev
ed
h
ig
h
er
p
r
ed
ictio
n
ac
cu
r
ac
y
a
n
d
lo
wer
er
r
o
r
v
alu
es
o
v
er
all.
T
h
is
r
esu
l
t
in
d
icate
s
th
at
lear
n
in
g
tem
p
o
r
al
in
f
o
r
m
atio
n
f
r
o
m
b
o
th
f
o
r
war
d
an
d
b
ac
k
war
d
d
ir
ec
tio
n
s
p
o
s
itiv
ely
af
f
ec
ts
th
e
p
r
ec
is
io
n
of
tim
e
s
er
ies
f
o
r
ec
asti
n
g
.
E
v
en
t
h
o
u
g
h
B
iLST
M
p
r
o
v
id
es
m
o
r
e
ac
cu
r
ate
r
esu
lts
,
it
r
eq
u
ir
es
more
c
o
m
p
u
tatio
n
al
p
o
wer
.
T
h
er
ef
o
r
e,
in
s
it
u
atio
n
s
wh
er
e
r
eso
u
r
ce
s
ar
e
lim
ited
or
tim
e
is
tig
h
t,
a
well
-
s
u
ited
u
n
i
d
ir
ec
tio
n
al
L
STM
co
u
ld
be
a
b
etter
,
m
o
r
e
p
r
ac
tical
ch
o
ice,
b
alan
cin
g
p
e
r
f
o
r
m
a
n
ce
an
d
ef
f
icien
cy
.
Ho
wev
er
,
s
in
ce
th
is
s
tu
d
y
is
b
ased
on
s
y
n
th
etic
T
AR
an
d
SET
AR
d
ata
with
s
elec
ted
h
y
p
e
r
p
a
r
am
eter
s
ettin
g
s
,
th
e
f
in
d
in
g
s
s
h
o
u
ld
be
in
ter
p
r
eted
as
e
v
id
en
ce
f
r
o
m
a
co
n
tr
o
lled
s
im
u
latio
n
s
ettin
g
r
ath
er
th
an
as
a
u
n
iv
er
s
al
co
n
clu
s
io
n
f
o
r
all
n
o
n
lin
ea
r
f
o
r
ec
asti
n
g
p
r
o
b
lem
s
.
Fu
tu
r
e
s
tu
d
ies
m
a
y
ex
te
n
d
t
h
is
wo
r
k
by
u
s
in
g
r
ea
l
-
wo
r
ld
n
o
n
lin
ea
r
tim
e
s
er
ie
s
d
ata,
b
r
o
ad
er
h
y
p
e
r
p
ar
a
m
eter
s
p
ac
es,
m
u
lti
-
s
tep
f
o
r
ec
asti
n
g
h
o
r
izo
n
s
,
co
m
p
u
tatio
n
al
tim
e
a
n
aly
s
is
,
an
d
h
y
b
r
i
d
m
o
d
ellin
g
ap
p
r
o
a
ch
es
to
p
r
o
v
i
d
e
a
m
o
r
e
c
o
m
p
lete
ev
alu
atio
n
of
f
o
r
ec
asti
n
g
ac
cu
r
ac
y
an
d
m
o
d
el
co
m
p
lex
ity
.
ACK
NO
WL
E
DG
M
E
N
T
S
W
e
ar
e
g
r
ate
f
u
l
t
o
th
e
an
o
n
y
m
o
u
s
r
ev
iewe
r
s
f
o
r
th
eir
v
alu
ab
le
f
ee
d
b
ac
k
.
W
e
ap
p
r
e
ciate
th
eir
in
p
u
t a
n
d
s
u
p
p
o
r
t.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
e
au
t
h
o
r
s
wo
u
ld
lik
e
to
th
an
k
th
e
Min
is
tr
y
o
f
Hig
h
er
E
d
u
ca
tio
n
f
o
r
p
r
o
v
id
in
g
f
in
a
n
cial
s
u
p
p
o
r
t
u
n
d
er
th
e
Fu
n
d
am
en
tal
R
esear
ch
Gr
an
ts
Sch
em
e
-
E
ar
ly
C
ar
ee
r
R
esear
ch
(
FR
GS
-
E
C
)
No
.
FR
GS
-
E
C
/1
/2
0
2
4
/STG0
6
/UMPSA/0
2
/1
(
Un
iv
er
s
ity
r
ef
er
en
ce
R
DU2
4
3
7
1
1
)
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Nu
r
Haiz
u
m
Ab
d
R
ah
m
an
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Han
i Sy
ah
id
a
Z
u
lk
i
f
li
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
169
6
-
1
7
0
5
1704
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
e
au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
ter
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
s
u
p
p
o
r
tin
g
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
wer
e
g
en
er
ated
th
r
o
u
g
h
s
im
u
latio
n
u
s
in
g
th
e
eq
u
atio
n
s
an
d
s
ettin
g
s
d
escr
ib
ed
in
th
e
p
ap
er
.
RE
F
E
R
E
NC
E
S
[
1
]
F
.
H
a
m
a
mi
a
n
d
I
.
A
.
D
a
h
l
a
n
,
“
U
n
i
v
a
r
i
a
t
e
t
i
me
ser
i
e
s
d
a
t
a
f
o
r
e
c
a
s
t
i
n
g
o
f
a
i
r
p
o
l
l
u
t
i
o
n
u
si
n
g
LSTM
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
i
n
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
A
d
v
a
n
c
e
m
e
n
t
i
n
D
a
t
a
S
c
i
e
n
c
e
,
E
-
l
e
a
r
n
i
n
g
a
n
d
I
n
f
o
rm
a
t
i
o
n
S
y
s
t
e
m
(
I
C
AD
EI
S
)
,
O
c
t
.
2
0
2
0
,
p
p
.
5
5
–
5
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
D
EI
S
4
9
8
1
1
.
2
0
2
0
.
9
2
7
7
3
9
3
.
[
2
]
A
.
I
.
S
u
n
n
y
,
M
.
M
.
S
.
M
a
sw
o
o
d
,
a
n
d
A
.
A
l
h
a
r
b
i
,
“
D
e
e
p
l
e
a
r
n
i
n
g
-
b
a
s
e
d
st
o
c
k
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
u
si
n
g
LST
M
a
n
d
bi
-
d
i
r
e
c
t
i
o
n
a
l
LSTM
m
o
d
e
l
,
”
i
n
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
N
o
v
e
l
I
n
t
e
l
l
i
g
e
n
t
a
n
d
L
e
a
d
i
n
g
Em
e
rg
i
n
g
S
c
i
e
n
c
e
s
(
N
I
L
ES
)
,
O
c
t
.
2
0
2
0
,
p
p
.
8
7
–
9
2
,
d
o
i
:
1
0
.
1
1
0
9
/
N
I
LES5
0
9
4
4
.
2
0
2
0
.
9
2
5
7
9
5
0
.
[
3
]
I
.
A
.
I
b
r
a
h
i
m,
“
I
mp
r
o
v
e
d
g
r
i
d
se
a
r
c
h
a
l
g
o
r
i
t
h
m
f
o
r
o
p
t
i
m
a
l
LST
M
p
e
r
f
o
r
ma
n
c
e
:
a
c
a
se
st
u
d
y
o
n
a
u
s
t
r
a
l
i
a
n
e
l
e
c
t
r
i
c
i
t
y
p
r
i
c
e
f
o
r
e
c
a
st
i
n
g
,
”
i
n
2
0
2
4
I
E
EE
Po
w
e
r
E
n
e
r
g
y
S
o
c
i
e
t
y
G
e
n
e
r
a
l
Me
e
t
i
n
g
(
PE
S
G
M)
,
J
u
l
.
2
0
2
4
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
1
0
9
/
P
ESG
M
5
1
9
9
4
.
2
0
2
4
.
1
0
6
8
8
8
4
9
.
[
4
]
R
.
L.
A
b
d
u
l
j
a
b
b
a
r
,
H
.
D
i
a
,
a
n
d
P
.
-
W
.
Tsa
i
,
“
U
n
i
d
i
r
e
c
t
i
o
n
a
l
a
n
d
b
i
d
i
r
e
c
t
i
o
n
a
l
LS
TM
mo
d
e
l
s
f
o
r
s
h
o
r
t
-
t
e
r
m
t
r
a
f
f
i
c
p
r
e
d
i
c
t
i
o
n
,”
J
o
u
rn
a
l
o
f
Ad
v
a
n
c
e
d
T
r
a
n
s
p
o
rt
a
t
i
o
n
,
v
o
l
.
2
0
2
1
,
2
0
2
1
,
p
p
.
1
–
1
6
,
d
o
i
:
1
0
.
1
1
5
5
/
2
0
2
1
/
5
5
8
9
0
7
5
.
[
5
]
S
.
H
o
c
h
r
e
i
t
e
r
a
n
d
J
.
S
c
h
m
i
d
h
u
b
e
r
,
“
Lo
n
g
sh
o
r
t
-
t
e
r
m
m
e
m
o
r
y
,
”
N
e
u
r
a
l
C
o
m
p
u
t
a
t
i
o
n
,
v
o
l
.
9
,
n
o
.
8
,
p
p
.
1
7
3
5
–
1
7
8
0
,
N
o
v
.
1
9
9
7
,
d
o
i
:
1
0
.
1
1
6
2
/
n
e
c
o
.
1
9
9
7
.
9
.
8
.
1
7
3
5
.
[
6
]
S
.
S
i
a
mi
-
N
a
mi
n
i
,
N
.
T
a
v
a
k
o
l
i
,
a
n
d
A
.
S
.
N
a
mi
n
,
“
T
h
e
p
e
r
f
o
r
ma
n
c
e
o
f
LS
TM
a
n
d
B
i
LST
M
i
n
f
o
r
e
c
a
st
i
n
g
t
i
me
s
e
r
i
e
s
,”
i
n
2
0
1
9
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Bi
g
D
a
t
a
(
Bi
g
D
a
t
a
)
,
L
o
s
A
n
g
e
l
e
s,
C
A
,
U
S
A
,
2
0
1
9
,
p
p
.
3
2
8
5
–
3
2
9
2
,
d
o
i
:
1
0
.
1
1
0
9
/
B
i
g
D
a
t
a
4
7
0
9
0
.
2
0
1
9
.
9
0
0
5
9
9
7
.
[
7
]
P
.
N
.
H
a
i
e
t
a
l
.
,
“
A
n
e
m
p
i
r
i
c
a
l
r
e
se
a
r
c
h
o
n
t
h
e
e
f
f
e
c
t
i
v
e
n
e
ss
o
f
d
i
f
f
e
r
e
n
t
LST
M
a
r
c
h
i
t
e
c
t
u
r
e
s
o
n
v
i
e
t
n
a
mese
st
o
c
k
m
a
r
k
e
t
,
”
i
n
A
C
M
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
Pr
o
c
e
e
d
i
n
g
S
e
r
i
e
s
,
A
ss
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
i
n
g
M
a
c
h
i
n
e
r
y
,
2
0
2
0
,
p
p
.
1
4
4
–
1
4
9
,
d
o
i
:
1
0
.
1
1
4
5
/
3
4
3
7
8
0
2
.
3
4
3
7
8
2
7
.
[
8
]
M
.
P
i
r
a
n
i
,
P
.
Th
a
k
k
a
r
,
P
.
J
i
v
r
a
n
i
,
M
.
H
.
B
o
h
a
r
a
,
a
n
d
D
.
G
a
r
g
,
“
A
c
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
si
s
o
f
A
R
I
M
A
,
G
R
U
,
LST
M
a
n
d
B
i
LST
M
o
n
f
i
n
a
n
c
i
a
l
t
i
me
ser
i
e
s
f
o
r
e
c
a
st
i
n
g
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
D
i
s
t
r
i
b
u
t
e
d
C
o
m
p
u
t
i
n
g
a
n
d
E
l
e
c
t
r
i
c
a
l
C
i
rc
u
i
t
s
a
n
d
E
l
e
c
t
r
o
n
i
c
s
(
I
C
D
C
E
C
E)
,
A
p
r
.
2
0
2
2
,
p
p
.
1
-
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
D
C
EC
E5
3
9
0
8
.
2
0
2
2
.
9
7
9
3
2
1
3
.
[
9
]
A
.
B
h
a
mb
u
,
“
S
t
o
c
k
mar
k
e
t
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
f
o
r
s
h
o
r
t
a
n
d
l
o
n
g
h
o
r
i
z
o
n
,
”
i
n
S
o
f
t
C
o
m
p
u
t
i
n
g
f
o
r
P
ro
b
l
e
m
S
o
l
v
i
n
g
,
M
.
T
h
a
k
u
r
,
S
.
A
g
n
i
h
o
t
r
i
,
B
.
S
.
R
a
j
p
u
r
o
h
i
t
,
M
.
P
a
n
t
,
K
.
D
e
e
p
,
a
n
d
A
.
K
.
N
a
g
a
r
,
E
d
s.
S
i
n
g
a
p
o
r
e
:
S
p
r
i
n
g
e
r
,
2
0
2
3
,
p
p
.
1
2
1
–
1
3
5
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
19
-
6
5
2
5
-
8
_
1
1
.
[
1
0
]
N
.
H
.
A
.
R
a
h
m
a
n
,
Q
.
P
.
Y
i
n
,
a
n
d
H
.
S
.
Zu
l
k
a
f
l
i
,
“
O
p
t
i
mi
s
i
n
g
LST
M
a
n
d
B
i
LST
M
m
o
d
e
l
s
f
o
r
t
i
me
s
e
r
i
e
s
f
o
r
e
c
a
s
t
i
n
g
t
h
r
o
u
g
h
h
y
p
e
r
p
a
r
a
me
t
e
r
t
u
n
i
n
g
,”
J
o
u
rn
a
l
o
f
Q
u
a
l
i
t
y
M
e
a
su
r
e
m
e
n
t
a
n
d
An
a
l
y
s
i
s
,
v
o
l
.
2
1
,
n
o
.
3
,
p
p
.
1
9
1
–
2
0
5
,
2
0
2
5
,
d
o
i
:
1
0
.
1
7
5
7
6
/
j
q
m
a
.
2
1
0
3
.
2
0
2
5
.
1
2
.
[
1
1
]
M
.
Jai
n
,
T.
A
l
s
k
a
i
f
,
a
n
d
S
.
D
e
v
,
“
A
r
e
d
e
e
p
l
e
a
r
n
i
n
g
mo
d
e
l
s
m
o
r
e
e
f
f
e
c
t
i
v
e
a
g
a
i
n
st
t
r
a
d
i
t
i
o
n
a
l
m
o
d
e
l
s
f
o
r
l
o
a
d
d
e
m
a
n
d
fo
r
e
c
a
st
i
ng
?
,
”
i
n
2
0
2
2
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
m
a
rt
E
n
e
r
g
y
S
y
st
e
m
s
a
n
d
T
e
c
h
n
o
l
o
g
i
e
s
(
S
E
S
T
)
,
Ei
n
d
h
o
v
e
n
,
N
e
t
h
e
r
l
a
n
d
s,
2
0
2
2
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
S
EST5
3
6
5
0
.
2
0
2
2
.
9
8
9
8
4
2
4
.
[
1
2
]
S
.
M
.
A
l
-
S
e
l
w
i
e
t
a
l
.
,
“
R
N
N
-
LSTM
:
f
r
o
m
a
p
p
l
i
c
a
t
i
o
n
s
t
o
mo
d
e
l
i
n
g
t
e
c
h
n
i
q
u
e
s
a
n
d
b
e
y
o
n
d
—
s
y
s
t
e
mat
i
c
r
e
v
i
e
w
,
”
J
o
u
rn
a
l
o
f
K
i
n
g
S
a
u
d
U
n
i
v
e
rsi
t
y
-
C
o
m
p
u
t
e
r
a
n
d
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
s
,
v
o
l
.
3
6
,
n
o
.
5
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
k
s
u
c
i
.
2
0
2
4
.
1
0
2
0
6
8
.
[
1
3
]
M
.
R
h
i
f
,
A
.
B
.
A
b
b
e
s
,
B
.
M
a
r
t
i
n
e
z
,
a
n
d
I
.
R
.
F
a
r
a
h
,
“
D
e
e
p
l
e
a
r
n
i
n
g
mo
d
e
l
s
p
e
r
f
o
r
ma
n
c
e
f
o
r
N
D
V
I
t
i
me
s
e
r
i
e
s
p
r
e
d
i
c
t
i
o
n
:
a
c
a
s
e
st
u
d
y
o
n
N
o
r
t
h
W
e
st
Tu
n
i
s
i
a
,
”
i
n
2
0
2
0
M
e
d
i
t
e
rra
n
e
a
n
a
n
d
M
i
d
d
l
e
-
E
a
st
G
e
o
sc
i
e
n
c
e
a
n
d
Re
m
o
t
e
S
e
n
s
i
n
g
S
y
m
p
o
s
i
u
m
(
M
2
G
AR
S
S
)
,
2
0
2
0
,
p
p
.
9
–
1
2
,
d
o
i
:
1
0
.
1
1
0
9
/
M
2
G
A
R
S
S
4
7
1
4
3
.
2
0
2
0
.
9
1
0
5
1
4
9
.
[
1
4
]
F
.
R
.
A
l
h
a
r
b
i
a
n
d
D
.
C
s
a
l
a
,
“
S
h
o
r
t
-
t
e
r
m
so
l
a
r
i
r
r
a
d
i
a
n
c
e
f
o
r
e
c
a
s
t
i
n
g
m
o
d
e
l
b
a
se
d
o
n
b
i
d
i
r
e
c
t
i
o
n
a
l
l
o
n
g
s
h
o
r
t
-
t
e
r
m
mem
o
r
y
d
e
e
p
l
e
a
r
n
i
n
g
,
”
i
n
2
0
2
1
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
E
l
e
c
t
r
i
c
a
l
,
C
o
m
m
u
n
i
c
a
t
i
o
n
,
a
n
d
C
o
m
p
u
t
e
r E
n
g
i
n
e
e
r
i
n
g
(
I
C
EC
C
E)
,
2
0
2
1
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
E
C
C
E
5
2
0
5
6
.
2
0
2
1
.
9
5
1
4
2
3
3
.
[
1
5
]
C
.
C
h
e
n
,
Y
.
L
i
,
K
.
Z
h
o
u
,
X
.
C
h
e
n
,
a
n
d
B
.
Ta
n
,
“
A
i
r
q
u
a
l
i
t
y
p
r
e
d
i
c
t
i
o
n
mo
d
e
l
u
t
i
l
i
z
i
n
g
Bi
-
LST
M
-
M
A
,
”
J
o
u
r
n
a
l
o
f
S
a
f
e
t
y
a
n
d
En
v
i
r
o
n
m
e
n
t
,
,
v
o
l
.
2
5
,
n
o
.
8
,
p
p
.
3
2
4
0
–
3
2
5
1
,
2
0
2
5
,
d
o
i
:
1
0
.
1
3
6
3
7
/
j
.
i
ssn
.
1
0
0
9
-
6
0
9
4
.
2
0
2
4
.
2
2
1
6
.
[
1
6
]
S
.
O
sma
n
,
N
.
S
.
G
h
e
r
i
y
a
n
i
,
S
.
S
a
l
e
m,
A
.
K
e
s
h
l
a
f
,
a
n
d
J
.
I
b
r
a
h
i
m
,
“
D
u
a
l
-
LS
TM
m
o
d
e
l
f
o
r
r
e
a
l
-
t
i
me
t
r
a
f
f
i
c
p
r
e
d
i
c
t
i
o
n
a
n
d
a
n
o
ma
l
y
d
e
t
e
c
t
i
o
n
i
n
I
o
T
e
n
v
i
r
o
n
me
n
t
,
”
i
n
S
t
u
d
i
e
s
i
n
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
A
.
O
.
A
l
b
a
j
i
,
E
d
.
,
S
p
r
i
n
g
e
r
S
c
i
e
n
c
e
a
n
d
B
u
s
i
n
e
ss
M
e
d
i
a
D
e
u
t
s
c
h
l
a
n
d
G
m
b
H
,
2
0
2
6
,
p
p
.
7
5
1
–
7
6
6
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
032
-
0
0
2
3
2
-
7
_
4
9
.
[
1
7
]
M
.
T.
P
a
w
i
t
r
a
,
H
.
F
a
k
h
r
u
r
r
o
j
a
,
a
n
d
L.
A
b
d
u
r
r
a
h
m
a
n
,
“
P
r
e
d
i
c
t
i
n
g
st
o
c
k
mar
k
e
t
u
si
n
g
C
N
N
a
n
d
B
i
LSTM
mo
d
e
l
,”
i
n
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
e
r,
C
o
n
t
r
o
l
,
I
n
f
o
rm
a
t
i
c
s
a
n
d
i
t
s
Ap
p
l
i
c
a
t
i
o
n
s
(
I
C
3
I
N
A)
,
2
0
2
4
,
p
p
.
2
6
7
–
2
7
2
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
3
I
N
A
6
4
0
8
6
.
2
0
2
4
.
1
0
7
3
2
2
4
5
.
[
1
8
]
N
.
I
.
A
.
K
a
d
e
r
,
U
.
K
.
Y
u
s
o
f
,
M
.
N
.
A
.
K
h
a
l
i
d
,
a
n
d
N
.
R
.
N
.
H
u
s
a
i
n
,
“
A
r
e
v
i
e
w
o
f
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
a
p
p
r
o
a
c
h
f
o
r
t
i
m
e
s
e
r
i
e
s
a
n
a
l
y
s
i
s
a
n
d
f
o
r
e
c
a
s
t
i
n
g
,
”
i
n
L
e
c
t
u
r
e
N
o
t
e
s
i
n
N
e
t
w
o
r
k
s
a
n
d
S
y
s
t
e
m
s
,
M
.
A
.
A
l
-
S
h
a
r
a
f
i
,
M
.
A
l
-
Em
r
a
n
,
K
.
S
h
a
a
l
a
n
,
a
n
d
M
.
N
.
Al
-
K
a
b
i
,
E
d
s
.
,
S
p
r
i
n
g
e
r
S
c
i
e
n
c
e
a
n
d
B
u
si
n
e
ss
M
e
d
i
a
D
e
u
t
s
c
h
l
a
n
d
G
m
b
H
,
2
0
2
3
,
p
p
.
1
2
–
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
031
-
20429
-
6_2.
[
1
9
]
N
.
Q
.
V
i
e
t
,
N
.
N
.
Q
u
a
n
g
,
N
.
K
i
n
g
,
D
.
T.
H
u
u
,
N
.
D
.
To
a
n
,
a
n
d
D
.
N
.
H
.
T
h
a
n
h
,
“
A
n
e
x
p
l
o
r
a
t
o
r
y
c
o
m
p
a
r
i
s
o
n
o
f
LST
M
a
n
d
B
i
LST
M
i
n
st
o
c
k
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
,
”
i
n
I
n
v
e
n
t
i
v
e
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
C
o
m
p
u
t
a
t
i
o
n
a
l
T
e
c
h
n
o
l
o
g
i
e
s
,
2
0
2
3
,
p
p
.
5
1
3
–
5
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
99
-
5
1
6
6
-
6
_
3
5
.
[
2
0
]
A
.
E.
O
u
a
r
d
i
,
B
.
Er
-
R
a
h
a
,
a
n
d
K
.
T
a
t
a
n
e
,
“
LS
TM
a
d
a
p
t
i
v
e
h
y
p
e
r
p
a
r
a
m
e
t
e
r
t
u
n
i
n
g
f
o
r
f
i
n
a
n
c
i
a
l
t
i
me
s
e
r
i
e
s
f
o
r
e
c
a
st
i
n
g
u
si
n
g
c
u
s
t
o
m
g
r
a
d
i
e
n
t
-
b
a
se
d
m
e
t
h
o
d
s
:
a
c
o
m
p
a
r
a
t
i
v
e
st
u
d
y
w
i
t
h
B
a
y
e
si
a
n
o
p
t
i
m
i
z
a
t
i
o
n
,
”
J
o
u
rn
a
l
o
f
T
h
e
o
ret
i
c
a
l
a
n
d
Ap
p
l
i
e
d
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
0
2
,
n
o
.
2
4
,
p
p
.
8
9
1
3
–
8
9
3
6
,
2
0
2
4
.
[
2
1
]
M
.
K
r
i
c
h
e
n
a
n
d
A
.
M
i
h
o
u
b
,
“
L
o
n
g
sh
o
r
t
-
t
e
r
m
m
e
m
o
r
y
n
e
t
w
o
r
k
s
:
a
c
o
mp
r
e
h
e
n
si
v
e
s
u
r
v
e
y
,
”
AI
,
v
o
l
.
6
,
p
.
2
1
5
,
S
e
p
.
2
0
2
5
,
d
o
i
:
1
0
.
3
3
9
0
/
a
i
6
0
9
0
2
1
5
.
[
2
2
]
Ö
.
Y
i
l
d
i
r
i
m
,
“
A
n
o
v
e
l
w
a
v
e
l
e
t
se
q
u
e
n
c
e
b
a
se
d
o
n
d
e
e
p
b
i
d
i
r
e
c
t
i
o
n
a
l
L
S
TM
n
e
t
w
o
r
k
m
o
d
e
l
f
o
r
EC
G
si
g
n
a
l
c
l
a
ss
i
f
i
c
a
t
i
o
n
,”
C
o
m
p
u
t
e
rs
i
n
B
i
o
l
o
g
y
a
n
d
Me
d
i
c
i
n
e
,
v
o
l
.
9
6
,
p
p
.
1
8
9
–
2
0
2
,
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
mp
b
i
o
m
e
d
.
2
0
1
8
.
0
3
.
0
1
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Un
id
ir
ec
tio
n
a
l a
n
d
b
id
ir
ec
tio
n
a
l LS
TM
fo
r
n
o
n
lin
ea
r
s
yn
th
etic
time
s
erie
s
… (
N
u
r
Ha
iz
u
m
A
b
d
R
a
h
ma
n
)
1705
[
2
3
]
E.
F
ı
r
a
t
,
“
S
ETA
R
(
S
e
l
f
-
e
x
c
i
t
i
n
g
t
h
r
e
s
h
o
l
d
a
u
t
o
r
e
g
r
e
ssi
v
e
)
n
o
n
-
l
i
n
e
a
r
c
u
r
r
e
n
c
y
m
o
d
e
l
l
i
n
g
i
n
EU
R
/
U
S
D
,
EU
R
/
TR
Y
a
n
d
U
S
D
/
TR
Y
p
a
r
i
t
i
e
s,
”
M
a
t
h
e
m
a
t
i
c
s
a
n
d
S
t
a
t
i
st
i
c
s
,
v
o
l
.
5
,
p
p
.
3
3
–
5
5
,
Ja
n
.
2
0
1
7
,
d
o
i
:
1
0
.
1
3
1
8
9
/
ms
.
2
0
1
7
.
0
5
0
1
0
5
.
[
2
4
]
A
.
Lo
u
ssa
i
e
f
,
R
.
K
t
a
r
i
,
a
n
d
Y
.
H
a
d
j
K
a
c
e
m
,
“
H
G
S
-
At
-
LST
M
:
a
t
t
e
n
t
i
o
n
-
b
a
s
e
d
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
m
o
d
e
l
c
o
mb
i
n
e
d
w
i
t
h
h
a
l
v
i
n
g
g
r
i
d
se
a
r
c
h
o
p
t
i
mi
z
e
r
f
o
r
h
a
r
mf
u
l
a
l
g
a
l
b
l
o
o
m
f
o
r
e
c
a
st
i
n
g
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
D
a
t
a
S
c
i
e
n
c
e
a
n
d
An
a
l
y
t
i
c
s
,
v
o
l
.
2
0
,
n
o
.
6
,
p
p
.
5
4
5
7
–
5
4
8
3
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
0
7
/
s4
1
0
6
0
-
025
-
0
0
7
8
3
-
2.
[
2
5
]
Z.
C
u
i
,
R
.
K
e
,
Z.
P
u
,
a
n
d
Y
.
W
a
n
g
,
“
S
t
a
c
k
e
d
b
i
d
i
r
e
c
t
i
o
n
a
l
a
n
d
u
n
i
d
i
r
e
c
t
i
o
n
a
l
LST
M
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
f
o
r
e
c
a
st
i
n
g
n
e
t
w
o
r
k
-
w
i
d
e
t
r
a
f
f
i
c
s
t
a
t
e
w
i
t
h
mi
ss
i
n
g
v
a
l
u
e
s
,
”
T
r
a
n
s
p
o
rt
a
t
i
o
n
R
e
se
a
rc
h
Pa
rt
C
:
Em
e
rg
i
n
g
T
e
c
h
n
o
l
o
g
i
e
s
,
v
o
l
.
1
1
8
,
p
.
1
0
2
6
7
4
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
t
r
c
.
2
0
2
0
.
1
0
2
6
7
4
.
[
2
6
]
A
.
Th
o
ma
s,
M
.
P
e
t
r
i
d
i
s
,
S
.
W
a
l
t
e
r
s,
S
.
M
.
G
h
e
y
t
a
ss
i
,
a
n
d
R
.
M
o
r
g
a
n
,
“
O
n
p
r
e
d
i
c
t
i
n
g
t
h
e
o
p
t
i
mal
n
u
m
b
e
r
o
f
h
i
d
d
e
n
n
o
d
e
s
,”
i
n
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
a
t
i
o
n
a
l
S
c
i
e
n
c
e
a
n
d
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
C
S
C
I
)
,
2
0
1
5
,
p
p
.
5
6
5
–
5
7
0
,
d
o
i
:
1
0
.
1
1
0
9
/
C
S
C
I
.
2
0
1
5
.
3
3
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Nur
H
a
izum
Abd
Ra
h
m
a
n
re
c
e
iv
e
d
h
e
r
first
d
e
g
re
e
a
n
d
P
h
.
D
.
d
e
g
re
e
fro
m
Un
iv
e
rsiti
Tek
n
o
l
o
g
i
M
a
lay
sia
,
i
n
2
0
1
0
a
n
d
2
0
1
7
,
re
sp
e
c
ti
v
e
l
y
.
S
h
e
is
c
u
rre
n
tl
y
a
se
n
io
r
lec
tu
re
r
a
t
Un
iv
e
rsiti
M
a
lay
sia
P
a
h
a
n
g
Al
-
S
u
l
tan
A
b
d
u
ll
a
h
,
wit
h
h
e
r
m
a
in
re
se
a
rc
h
i
n
tere
sts
fo
c
u
sin
g
o
n
sta
ti
stics
,
p
a
rt
icu
larl
y
i
n
ti
m
e
se
ries
a
n
a
ly
sis.
P
re
v
io
u
sly
,
sh
e
wo
rk
e
d
a
s
a
se
n
i
o
r
lec
tu
re
r
a
t
Un
iv
e
rsiti
P
u
tra
M
a
l
a
y
sia
.
S
h
e
h
a
s
a
to
tal
o
f
n
i
n
e
y
e
a
rs’
tea
c
h
in
g
e
x
p
e
rien
c
e
.
He
r
re
se
a
rc
h
p
u
b
li
c
a
ti
o
n
s
a
re
re
g
u
larl
y
p
u
b
li
s
h
e
d
in
lea
d
i
n
g
j
o
u
rn
a
ls,
a
n
d
sh
e
is
a
c
ti
v
e
l
y
in
v
o
lv
e
d
in
t
h
e
a
c
a
d
e
m
ic
c
o
m
m
u
n
it
y
,
a
tt
e
n
d
i
n
g
c
o
n
fe
re
n
c
e
s,
sy
m
p
o
si
u
m
s,
a
n
d
se
m
in
a
rs.
Be
y
o
n
d
h
e
r
u
n
iv
e
rsit
y
a
c
a
d
e
m
ic
re
sp
o
n
si
b
il
it
ies
,
s
h
e
h
o
l
d
s
a
li
fe
l
o
n
g
m
e
m
b
e
rsh
i
p
i
n
I
n
stit
u
t
S
tatisti
k
M
a
lay
sia
(IS
M
y
)
a
n
d
c
u
rre
n
tl
y
se
rv
e
s
a
s
a
c
o
u
n
c
il
m
e
m
b
e
r,
fu
lfi
l
li
n
g
t
h
e
ro
les
o
f
Ho
n
o
u
ra
ry
S
e
c
re
tary
fro
m
2
0
2
2
to
2
0
2
4
a
n
d
fro
m
2
0
2
4
to
2
0
2
6
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
h
a
izu
m
@u
m
p
sa
.
e
d
u
.
m
y
.
H
a
n
i
S
y
a
h
id
a
Zu
lk
a
fli
re
c
e
iv
e
d
h
e
r
M
a
ste
r
in
A
p
p
l
ied
S
t
a
ti
stics
a
n
d
B
.
S
c
(Ho
n
s.)
i
n
S
tatisti
c
s
fro
m
Un
i
v
e
rsiti
P
u
tra
M
a
lay
sia
i
n
y
e
a
r
2
0
1
1
a
n
d
2
0
0
9
,
re
sp
e
c
ti
v
e
ly
.
He
r
P
h
.
D
.
d
e
g
re
e
fro
m
He
rio
t
-
W
a
tt
Un
iv
e
rsity
.
S
h
e
h
a
s
b
e
e
n
w
o
rk
in
g
a
s
a
S
e
n
io
r
Lec
t
u
re
r
i
n
th
e
De
p
a
rtme
n
t
o
f
M
a
th
e
m
a
ti
c
s
a
n
d
S
tatisti
c
s
a
t
Un
iv
e
rsiti
P
u
tra
M
a
lay
s
ia
si
n
c
e
2
0
1
7
.
S
h
e
h
a
s
1
0
y
e
a
rs
o
f
tea
c
h
in
g
a
n
d
re
se
a
rc
h
e
x
p
e
rien
c
e
a
t
v
a
rio
u
s
re
p
u
te
d
in
st
it
u
t
io
n
s.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
a
p
p
l
ica
ti
o
n
s
o
f
Ba
y
e
sia
n
n
e
two
r
k
s
,
Ba
y
e
sia
n
m
o
d
e
li
n
g
,
a
n
d
i
n
fe
re
n
c
e
in
c
h
ro
n
ic
d
ise
a
se
s
,
su
c
h
a
s d
iab
e
tes
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
h
sy
a
h
id
a
@u
p
m
.
e
d
u
.
m
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.