I
nte
rna
t
io
na
l J
o
urna
l o
f
Adv
a
nces in Applie
d Science
s
(
I
J
AAS)
Vo
l.
15
,
No
.
1
,
Ma
r
ch
20
26
,
p
p
.
4
1
6
~
4
2
6
I
SS
N:
2252
-
8
8
1
4
,
DOI
:
1
0
.
1
1
5
9
1
/ijaas
.
v15.
i
1
.
pp
416
-
4
2
6
416
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
a
s
.
ia
esco
r
e.
co
m
Ada
ptive sen
time
nt
a
na
ly
sis
f
o
r sto
ck ma
rkets
using
deep
lea
rning
T
a
lent
M
a
wer
e,
Selv
a
ra
j
Ra
j
a
la
k
s
hm
i,
Venu
M
a
dh
a
v
K
utha
di,
O
t
lha
pil
e
Dina
k
eny
a
ne
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
i
n
g
a
n
d
I
n
f
o
r
m
a
t
i
c
s,
S
c
h
o
o
l
o
f
N
a
t
u
r
a
l
a
n
d
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s,
B
o
t
sw
a
n
a
I
n
t
e
r
n
a
t
i
o
n
a
l
U
n
i
v
e
r
si
t
y
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
P
a
l
a
p
y
e
,
B
o
t
sw
a
n
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
6
,
2025
R
ev
is
ed
No
v
22
,
2
0
2
5
Acc
ep
ted
Feb
17
,
2
0
2
6
S
to
c
k
m
a
rk
e
ts
a
re
h
ig
h
ly
v
o
latil
e
,
m
a
k
in
g
p
rice
p
re
d
ictio
n
v
e
ry
d
iff
icu
lt
.
On
e
o
f
t
h
e
fa
c
to
rs
i
n
flu
e
n
c
in
g
th
e
v
o
latil
it
y
o
f
fi
n
a
n
c
ial
m
a
rk
e
ts
is
ra
p
id
l
y
c
h
a
n
g
i
n
g
n
e
ws
se
n
ti
m
e
n
t
.
T
h
is
st
u
d
y
p
re
se
n
ts
a
n
o
v
e
l
a
d
a
p
ti
v
e
d
e
e
p
lea
rn
i
n
g
(DL)
fra
m
e
wo
rk
fo
r
se
n
ti
m
e
n
t
a
n
a
ly
sis
with
c
o
n
c
e
p
t
d
rift
c
a
p
a
b
i
li
ti
e
s.
Th
e
p
ro
p
o
se
d
m
o
d
e
l
c
o
m
b
i
n
e
s
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
(CNN
)
,
b
id
irec
ti
o
n
a
l
l
o
n
g
sh
o
r
t
-
term
m
e
m
o
ry
(
BiL
S
T
M
)
,
a
n
d
a
tt
e
n
ti
o
n
m
e
c
h
a
n
ism
s
in
it
s p
ro
c
e
ss
in
g
a
rc
h
it
e
c
tu
re
.
T
h
e
m
o
d
e
l
i
n
p
u
ts
p
re
p
ro
c
e
ss
e
d
n
e
ws
h
e
a
d
li
n
e
s
in
to
b
o
t
h
t
h
e
CNN
a
n
d
BiL
S
T
M
-
A
tt
e
n
ti
o
n
n
e
two
r
k
s to
e
x
trac
t
lo
c
a
l
fe
a
tu
re
s,
m
o
d
e
l
c
o
n
tex
tu
a
l
d
e
p
e
n
d
e
n
c
ies
,
a
n
d
p
r
io
rit
ize
s
imp
o
rtan
t
se
n
ti
m
e
n
t
c
u
e
s
in
it
s
p
re
d
ictio
n
m
e
c
h
a
n
ism
.
We
u
se
F
a
stTex
t
a
n
d
Wo
r
d
2
Ve
c
fo
r
wo
rd
e
m
b
e
d
d
in
g
s,
w
h
il
e
in
c
re
m
e
n
tal
lea
rn
in
g
i
s
u
se
d
to
m
a
n
a
g
e
c
o
n
c
e
p
t
d
rif
t.
On
e
k
e
y
a
d
v
a
n
ta
g
e
o
f
h
a
n
d
li
n
g
c
o
n
c
e
p
t
d
r
ift
is
th
a
t
t
h
e
m
o
d
e
l
c
a
n
c
o
n
ti
n
u
o
u
sly
lea
rn
n
e
w
p
a
tt
e
rn
s
in
d
a
ta
stre
a
m
s
with
o
u
t
n
e
e
d
in
g
to
fu
ll
y
re
train
th
e
m
o
d
e
l.
Th
e
m
o
d
e
l
is
v
a
li
d
a
ted
o
n
a
c
u
ra
ted
d
a
tas
e
t
fro
m
v
a
rio
u
s
so
u
rc
e
s
w
it
h
su
p
e
rio
r
p
e
rfo
rm
a
n
c
e
a
c
ro
ss
a
ll
m
e
tri
c
s,
li
k
e
a
c
c
u
ra
c
y
(0
.
9
7
5
3
)
a
n
d
a
n
F
1
-
sc
o
re
(0
.
9
8
).
It
si
g
n
ifi
c
a
n
tl
y
o
u
t
p
e
rfo
rm
s b
e
n
c
h
m
a
rk
s li
k
e
d
isti
ll
e
d
b
id
irec
ti
o
n
a
l
e
n
c
o
d
e
r
re
p
re
se
n
t
a
ti
o
n
s
fro
m
tra
n
sfo
rm
e
rs
(
Distil
BERT
)
,
LS
TM
,
a
n
d
v
a
len
c
e
a
wa
re
d
ictio
n
a
ry
a
n
d
se
n
ti
m
e
n
t
re
a
so
n
e
r
(
V
AD
E
R
)
.
A
ru
n
o
f
ten
it
e
ra
ti
o
n
s
v
a
li
d
a
te
d
t
h
a
t
th
e
re
a
l
-
ti
m
e
p
ip
e
li
n
e
d
id
n
o
t
e
x
c
e
e
d
2
0
0
ms
in
p
ro
c
e
ss
in
g
a
n
d
c
las
sify
i
n
g
h
e
a
d
li
n
e
s.
T
h
is
sig
n
ifi
e
s
th
e
p
ra
c
ti
c
a
l
v
iab
il
i
ty
o
f
o
u
r
m
o
d
e
l
in
fi
n
tec
h
a
p
p
li
c
a
ti
o
n
s
su
c
h
a
s
a
lg
o
r
it
h
m
ic
tr
a
d
in
g
a
n
d
risk
m
a
n
a
g
e
m
e
n
t.
K
ey
w
o
r
d
s
:
Atten
tio
n
m
ec
h
an
is
m
Dee
p
lear
n
in
g
Fin
T
ec
h
Natu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
R
ea
l
-
tim
e
d
ata
p
ip
elin
e
Sen
tim
en
t a
n
aly
s
is
Sto
ck
m
ar
k
et
f
o
r
ec
asti
n
g
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
:
T
alen
t M
awe
r
e
Dep
ar
tm
en
t o
f
C
o
m
p
u
tin
g
an
d
I
n
f
o
r
m
atics,
Sch
o
o
l o
f
Natu
r
a
l a
n
d
Ap
p
lied
Scien
ce
s
B
o
ts
wan
a
I
n
ter
n
atio
n
al
Un
iv
e
r
s
ity
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
Palap
y
e,
B
o
ts
wan
a
E
m
ail:
talen
tm
awe
r
e@
g
m
ail.
co
m
o
r
m
t2
3
0
1
8
9
7
5
@
b
iu
s
t.a
c.
b
w
1.
I
NT
RO
D
UCT
I
O
N
Pre
d
ictin
g
s
to
ck
m
ar
k
et
tr
en
d
s
is
im
p
o
r
tan
t
f
o
r
in
v
esto
r
s
an
d
p
o
licy
m
ak
er
s
.
H
o
wev
er
,
it
is
s
till
d
if
f
icu
lt
b
ec
a
u
s
e
o
f
th
e
c
o
m
p
lex
m
ix
o
f
n
u
m
e
r
ical
d
ata
a
n
d
o
th
e
r
f
ac
to
r
s
lik
e
n
ews
s
en
t
im
en
t
an
d
in
v
esto
r
p
s
y
ch
o
lo
g
y
th
at
co
n
tr
ib
u
te
to
s
to
ck
p
r
ice
m
o
v
em
en
ts
[
1
]
,
[
2
]
.
Pas
t
s
tu
d
ies
d
o
cu
m
en
t
tw
o
m
ain
tech
n
iq
u
es,
f
u
n
d
am
e
n
tal
an
aly
s
is
an
d
te
ch
n
ical
an
aly
s
is
,
u
s
ed
f
o
r
s
t
o
ck
m
a
r
k
et
p
r
ed
ictio
n
.
T
ec
h
n
ical
an
aly
s
is
u
s
es
ar
tific
ial
in
tellig
en
ce
o
r
m
ac
h
in
e
lear
n
in
g
(
AI
/ML
)
o
n
h
is
to
r
ical
d
ata.
Fu
n
d
am
en
tal
an
a
ly
s
is
,
o
n
th
e
o
th
e
r
h
an
d
,
ev
al
u
ates
f
in
an
cial
s
tate
m
en
ts
to
p
r
ed
ict
f
u
tu
r
e
s
to
ck
p
r
ices.
B
o
th
,
h
o
wev
er
,
o
f
ten
m
is
s
th
e
im
p
o
r
tan
t
ef
f
ec
t o
f
m
a
r
k
et
s
en
tim
en
t f
r
o
m
n
ews a
n
d
s
o
cial
m
ed
ia
[
3
]
,
[
4
]
.
D
e
e
p
l
e
a
r
n
i
n
g
(
DL
)
t
e
c
h
n
i
q
u
es
c
a
n
h
e
l
p
o
v
e
r
c
o
m
e
t
h
is
li
m
i
t
ati
o
n
b
y
a
u
t
o
m
a
t
i
c
al
l
y
i
d
e
n
ti
f
y
i
n
g
p
a
t
t
e
r
n
s
an
d
u
s
i
n
g
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
s
s
i
n
g
(
N
L
P
)
f
o
r
u
n
s
t
r
u
c
t
u
r
e
d
d
a
t
a
[
5
]
,
[
6
]
.
S
e
v
e
r
al
s
t
u
d
i
es
,
e
.
g
.
,
[
4
]
,
[
7
]
h
a
v
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
A
d
a
p
tive
s
en
timen
t a
n
a
lysi
s
fo
r
s
to
ck
ma
r
ke
ts
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ta
len
t Ma
w
ere
)
417
d
e
m
o
n
s
t
r
a
t
e
d
t
h
e
e
f
f
i
c
a
c
y
o
f
D
L
m
e
t
h
o
d
s
t
h
a
t
i
n
t
e
g
r
a
t
e
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
is
i
n
s
t
o
c
k
m
a
r
k
e
t
p
r
e
d
i
c
t
i
o
n
.
I
n
v
e
s
t
o
r
s
'
s
e
n
t
i
m
e
n
t
s
c
h
a
n
g
e
r
a
p
i
d
l
y
d
u
e
t
o
m
a
n
y
f
a
c
t
o
r
s
,
a
n
d
a
m
o
d
e
l
t
r
a
i
n
e
d
o
n
o
l
d
e
r
d
a
ta
m
a
y
f
a
i
l
t
o
p
r
e
d
i
c
t
a
c
c
u
r
a
t
e
l
y
i
f
s
e
n
t
i
m
e
n
t
s
s
h
i
f
t
r
ap
i
d
l
y
.
T
h
i
s
is
a
n
o
n
g
o
i
n
g
c
h
a
l
le
n
g
e
w
h
e
r
e
m
o
d
e
l
s
la
c
k
r
e
a
l
-
tim
e
a
d
a
p
t
a
b
i
l
it
y
,
a
r
e
v
u
l
n
e
r
a
b
l
e
t
o
c
o
n
c
e
p
t
d
r
i
f
t
,
a
n
d
a
r
e
g
e
n
e
r
a
l
l
y
u
n
c
l
e
a
r
o
n
h
o
w
th
e
d
e
c
i
s
i
o
n
-
m
a
k
i
n
g
p
r
o
c
e
s
s
wo
r
k
s
[
8
]
,
[
9
]
.
T
o
s
p
ec
if
ically
m
o
d
el
f
in
a
n
cial
d
ata,
tr
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
lik
e
f
in
an
cial
b
i
d
ir
ec
tio
n
al
en
co
d
er
r
ep
r
esen
tatio
n
s
f
r
o
m
tr
a
n
s
f
o
r
m
er
s
(
Fin
B
E
R
T
)
[
1
0
]
h
av
e
b
e
en
d
ev
el
o
p
ed
.
Su
ch
m
o
d
els
o
f
f
er
d
o
m
ain
-
s
p
ec
if
ic
p
r
o
wess
b
u
t
h
a
v
e
b
ee
n
h
ea
v
ily
cr
iticized
f
o
r
th
eir
h
ea
v
y
co
m
p
u
tatio
n
al
r
e
q
u
ir
em
e
n
ts
an
d
ar
e
th
u
s
n
o
t
s
u
itab
le
f
o
r
h
ig
h
-
f
r
eq
u
e
n
cy
s
tr
ea
m
in
g
ap
p
licatio
n
s
.
A
cr
itical
g
ap
ex
is
ts
,
th
er
ef
o
r
e,
f
o
r
f
r
am
ewo
r
k
s
th
at
b
alan
ce
h
ig
h
ac
cu
r
ac
y
,
c
o
m
p
u
tatio
n
al
ef
f
ici
en
cy
,
an
d
r
ea
l
-
tim
e
a
d
ap
tatio
n
to
b
o
th
ev
o
l
v
in
g
m
a
r
k
et
lan
g
u
ag
e
an
d
ev
o
l
v
in
g
m
ar
k
et
s
en
tim
en
ts
.
I
n
r
esp
o
n
s
e
to
t
h
is
,
o
u
r
p
ap
er
p
r
o
p
o
s
es
a
n
ew
in
tellig
en
t
f
r
am
ewo
r
k
f
o
r
ad
a
p
tiv
e
s
en
tim
en
t
an
aly
s
is
in
s
to
ck
m
ar
k
ets
u
s
in
g
ad
a
p
tiv
e
DL
tech
n
iq
u
es.
Ou
r
wo
r
k
m
ak
es
th
r
ee
im
p
o
r
tan
t
co
n
tr
ib
u
tio
n
s
:
i)
we
d
esig
n
a
d
o
m
ain
-
s
p
ec
if
ic
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
-
b
id
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
C
NN
-
B
iLST
M
)
m
o
d
el
with
an
atten
tio
n
m
ec
h
an
i
s
m
.
T
h
is
s
etu
p
ef
f
ec
tiv
ely
ex
t
r
ac
ts
lo
ca
l
s
em
an
tic
p
atter
n
s
an
d
m
o
d
els
lo
n
g
-
r
an
g
e
co
n
tex
tu
al
d
ep
en
d
en
cies
i
n
f
in
a
n
cial
tex
t,
alwa
y
s
em
p
h
asizin
g
th
e
m
o
s
t
s
alien
t
s
en
tim
en
t
cu
es.
T
h
ese
ad
d
r
ess
lim
itatio
n
s
f
o
u
n
d
in
s
tan
d
alo
n
e
n
e
u
r
al
n
etwo
r
k
s
;
i
i)
we
in
tr
o
d
u
ce
a
n
ad
ap
tiv
e
r
ea
l
-
tim
e
p
ip
elin
e
wi
th
co
n
ce
p
t
d
r
if
t
d
etec
tio
n
an
d
an
in
cr
em
en
tal
lea
r
n
in
g
m
o
d
u
le
u
s
in
g
th
e
r
i
v
er
lib
r
ar
y
.
T
h
is
allo
ws
th
e
m
o
d
e
l
to
u
p
d
ate
c
o
n
tin
u
o
u
s
ly
b
ase
d
o
n
r
ec
en
t
d
ata.
I
t
h
elp
s
m
a
in
tain
ac
cu
r
ac
y
i
n
v
o
lati
le
m
ar
k
ets
with
o
u
t
th
e
n
ee
d
f
o
r
ex
p
e
n
s
iv
e
r
etr
ain
i
n
g
;
an
d
iii)
to
v
alid
ate
o
u
r
m
o
d
el,
we
c
o
n
d
u
ct
ex
ten
s
iv
e
b
en
c
h
m
ar
k
i
n
g
th
r
o
u
g
h
a
t
h
o
r
o
u
g
h
em
p
ir
ical
e
v
al
u
atio
n
ag
ai
n
s
t
a
v
ar
iety
o
f
b
a
s
elin
es,
in
clu
d
in
g
a
tr
an
s
f
o
r
m
er
,
d
is
till
ed
B
E
R
T
(
Dis
til
B
E
R
T
)
,
a
DL
m
o
d
el
,
L
STM
,
C
NN
-
B
iLST
M,
an
d
lex
i
co
n
-
b
ased
to
o
ls
lik
e
v
alen
ce
awa
r
e
d
ictio
n
ar
y
a
n
d
s
en
tim
en
t
r
ea
s
o
n
er
(
VADE
R
)
an
d
T
ex
tB
lo
b
.
O
u
r
r
es
u
lts
s
h
o
w
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
ac
r
o
s
s
m
u
ltip
le
m
etr
ics.
F
i
n
a
n
c
i
a
l
m
a
r
k
e
t
s
h
a
v
e
a
l
w
a
y
s
s
p
a
r
k
e
d
i
n
t
e
r
e
s
t
a
m
o
n
g
m
a
n
y
g
r
o
u
p
s
s
i
n
c
e
t
h
e
b
e
g
i
n
n
i
n
g
o
f
t
i
m
e
.
I
n
i
t
i
a
l
l
y
,
e
a
r
l
y
m
a
r
k
e
t
t
h
e
o
r
i
e
s
,
s
u
c
h
a
s
t
h
e
e
f
f
i
c
i
e
n
t
m
a
r
k
e
t
h
y
p
o
t
h
e
s
i
s
(
E
M
H
)
[
1
1
]
a
n
d
r
a
n
d
o
m
w
a
l
k
t
h
e
o
r
y
[
1
2
]
,
p
o
s
ited
th
at
p
r
ices
f
u
lly
r
ef
lect
all
av
ailab
le
in
f
o
r
m
atio
n
an
d
ar
e
in
h
er
en
tly
u
n
p
r
e
d
ictab
le.
Ho
wev
er
,
b
eh
av
io
r
al
f
in
an
ce
h
as
ch
alle
n
g
ed
th
is
v
iew.
T
h
is
g
r
o
u
p
ar
g
u
es
th
at
co
g
n
itiv
e
b
iase
s
an
d
in
v
esto
r
s
en
tim
en
t
cr
ea
te
m
ar
k
et
in
ef
f
icien
cies
th
at
ca
n
b
e
ex
p
l
o
ited
to
p
r
e
d
ict
f
u
tu
r
e
s
to
ck
p
r
ices
[
1
3
]
,
[
1
4
]
.
I
t
is
f
r
o
m
th
is
v
iew
th
at
s
en
tim
en
t a
n
aly
s
is
h
as b
ec
o
m
e
a
k
e
y
to
o
l f
o
r
u
n
d
er
s
tan
d
in
g
m
ar
k
et
tr
en
d
s
f
r
o
m
u
n
s
tr
u
ctu
r
ed
tex
t
[
1
5
]
.
S
e
n
t
i
m
e
n
t
a
n
a
l
y
s
is
i
s
t
h
e
p
r
o
c
e
s
s
o
f
a
u
t
o
m
a
t
i
c
a
l
l
y
i
d
e
n
t
i
f
y
i
n
g
a
n
d
c
a
t
e
g
o
r
i
z
i
n
g
p
e
o
p
l
e
’
s
e
m
o
t
i
o
n
s
,
o
p
i
n
i
o
n
s
,
a
n
d
o
r
a
t
t
i
t
u
d
es
b
a
s
ed
o
n
e
x
t
r
a
c
t
e
d
t
e
x
t
.
S
e
n
ti
m
e
n
t
a
n
a
l
y
s
is
h
a
s
e
v
o
l
v
e
d
f
r
o
m
l
e
x
ic
o
n
-
b
a
s
e
d
m
e
t
h
o
d
s
l
i
k
e
V
AD
E
R
a
n
d
t
r
a
d
i
t
i
o
n
al
M
L
c
l
as
s
i
f
i
e
r
s
t
o
D
L
a
r
c
h
it
e
c
tu
r
e
s
.
R
ec
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s
(
R
N
Ns
)
,
e
s
p
e
ci
a
l
l
y
t
h
e
L
S
T
M
n
e
t
w
o
r
k
s
,
a
r
e
g
r
e
a
t
a
t
m
o
d
e
l
i
n
g
l
o
n
g
-
t
e
r
m
d
e
p
e
n
d
e
n
c
i
e
s
i
n
s
e
q
u
e
n
ti
a
l
t
e
x
t
[
1
6
]
.
M
e
a
n
w
h
i
l
e
,
C
N
Ns
a
r
e
g
o
o
d
a
t
d
e
t
e
c
t
i
n
g
l
o
c
al
n
-
g
r
a
m
p
a
t
t
e
r
n
s
t
h
a
t
m
a
y
c
ar
r
y
s
e
n
t
i
m
e
n
t
p
at
t
e
r
n
s
[
1
7
]
.
H
y
b
r
i
d
m
o
d
e
l
s
l
i
k
e
C
N
N
-
L
S
T
M
h
a
v
e
b
e
e
n
p
r
o
p
o
s
e
d
t
o
c
o
m
b
i
n
e
b
o
t
h
s
t
r
e
n
g
t
h
s
[
1
8
]
.
I
n
a
p
r
e
v
i
o
u
s
s
t
u
d
y
,
Wu
e
t
a
l
.
[
2
]
i
n
t
e
g
r
a
t
e
d
s
e
n
t
i
m
e
n
t
f
r
o
m
n
e
w
s
w
i
t
h
t
e
c
h
n
i
c
a
l
i
n
d
i
c
a
t
o
r
s
u
s
i
n
g
a
n
L
S
T
M
w
i
t
h
a
t
t
e
n
ti
o
n
,
b
u
t
t
h
e
i
r
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
is
r
e
l
i
ed
s
o
l
e
l
y
o
n
W
o
r
d
2
V
e
c
.
R
e
c
e
n
t
s
t
u
d
i
es
b
y
[
1
9
]
a
n
d
[
2
0
]
h
i
g
h
l
i
g
h
t
t
h
e
g
r
o
w
i
n
g
u
s
e
o
f
i
n
c
r
em
e
n
t
a
l
l
ea
r
n
i
n
g
f
o
r
f
i
n
a
n
c
i
a
l
d
a
t
a
s
t
r
e
a
m
s
,
y
e
t
i
t
s
i
n
t
e
g
r
a
t
i
o
n
w
i
t
h
a
h
y
b
r
i
d
C
N
N
-
B
i
L
S
T
M
-
A
tt
e
n
t
i
o
n
a
r
c
h
i
te
c
t
u
r
e
f
o
r
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
is
r
e
m
a
i
n
s
u
n
d
e
r
e
x
p
l
o
r
e
d
.
R
ec
en
tly
,
tr
an
s
f
o
r
m
er
-
b
ase
d
m
o
d
els
lik
e
B
E
R
T
[
2
1
]
an
d
its
f
in
an
cial
d
o
m
ain
ad
ap
tatio
n
,
Fin
B
E
R
T
[
1
0
]
h
a
s
ac
h
iev
ed
n
e
w
h
ig
h
s
i
n
NL
P
task
s
d
u
e
to
its
d
ee
p
u
n
d
er
s
tan
d
in
g
o
f
co
n
tex
t.
Desp
ite
th
eir
s
tr
en
g
th
s
,
tr
an
s
f
o
r
m
er
s
r
eq
u
i
r
e
a
lo
t
o
f
c
o
m
p
u
tin
g
p
o
wer
,
wh
ich
m
ak
es
r
e
al
-
tim
e
u
s
e
d
if
f
icu
lt.
Fu
r
th
er
m
o
r
e
,
m
o
s
t
ex
is
tin
g
m
o
d
els
ar
e
s
tatic
an
d
s
tr
u
g
g
le
with
co
n
ce
p
t
d
r
if
t.
C
o
n
ce
p
t
d
r
if
t
r
ef
er
s
to
th
e
p
h
en
o
m
en
o
n
wh
er
e
th
e
s
tatis
ti
ca
l p
r
o
p
e
r
ties
o
f
th
e
tar
g
et
v
ar
iab
le,
m
ar
k
et
s
en
tim
en
t,
c
h
an
g
e
o
v
er
tim
e
[
2
1
]
.
Ou
r
m
o
d
el
a
d
d
r
ess
es
th
ese
g
ap
s
b
y
p
r
o
p
o
s
in
g
a
co
m
p
u
tatio
n
ally
ef
f
icien
t
h
y
b
r
id
m
o
d
el
th
at
in
co
r
p
o
r
ates
a
d
ed
icate
d
co
n
c
ep
t
d
r
if
t
ad
ap
tatio
n
m
o
d
u
le,
b
r
id
g
in
g
th
e
p
er
f
o
r
m
an
ce
o
f
DL
with
th
e
ag
ilit
y
r
eq
u
ir
ed
f
o
r
r
ea
l
-
tim
e
f
in
tech
ap
p
licatio
n
s
.
T
ab
le
1
s
u
m
m
ar
i
zes
a
s
y
n
th
esis
o
f
s
o
m
e
ex
tan
t liter
atu
r
e
r
elate
d
to
th
e
s
tu
d
y
.
H
ig
h
lig
h
ti
n
g
t
h
e
r
e
s
ea
r
ch
f
o
c
u
s
,
m
eth
o
d
o
lo
g
y
u
s
ed
,
k
e
y
c
o
n
tr
ib
u
tio
n
s
,
an
d
t
h
e
id
en
tifie
d
g
ap
s
i
n
th
e
p
ar
ticu
lar
s
tu
d
y
.
2.
M
E
T
H
O
D
2
.
1
.
Sy
s
t
e
m
a
rc
hite
ct
ure
a
n
d da
t
a
pip
eline
Desig
n
ed
s
p
ec
if
ically
f
o
r
u
s
e
in
f
in
an
cial
p
r
e
d
ictio
n
s
y
s
tem
s
,
o
u
r
f
r
am
ewo
r
k
,
wh
ich
w
o
r
k
s
in
b
o
th
o
f
f
lin
e
tr
ain
in
g
an
d
o
n
lin
e
in
f
er
en
ce
,
is
s
h
o
wn
in
Fig
u
r
e
1
.
T
h
e
s
y
s
tem
b
eg
in
s
with
a
co
m
p
r
eh
e
n
s
iv
e
d
ata
ac
q
u
is
itio
n
an
d
p
r
o
ce
s
s
in
g
p
i
p
elin
e
,
wh
ich
we
d
is
cu
s
s
as f
o
llo
ws
.
i)
Data
ac
q
u
is
itio
n
p
r
o
ce
s
s
,
f
o
r
tr
ain
in
g
an
d
test
in
g
o
u
r
m
o
d
el,
in
v
o
lv
ed
cr
ea
tin
g
a
cu
s
to
m
Py
th
o
n
p
i
p
elin
e
u
s
in
g
B
ea
u
tifu
lSo
u
p
an
d
Scr
ap
y
,
wh
e
r
e
we
co
llected
o
v
e
r
1
.
2
m
illi
o
n
f
i
n
an
cial
n
ews
ar
ticles
f
r
o
m
s
o
u
r
ce
s
lik
e
B
lo
o
m
b
er
g
,
R
eu
t
er
s
,
an
d
th
e
W
all
Stre
et
J
o
u
r
n
al
f
o
r
th
e
p
e
r
io
d
b
etwe
en
2
0
1
6
an
d
2
0
2
3
.
T
h
e
r
ea
l
-
tim
e
d
ata
ac
q
u
is
itio
n
p
r
o
ce
s
s
f
o
r
in
c
r
em
en
tal
le
ar
n
in
g
in
v
o
lv
ed
c
o
n
n
ec
tin
g
v
ia
ap
p
licatio
n
p
r
o
g
r
a
m
m
in
g
i
n
ter
f
ac
e
(
API
)
to
a
d
y
n
am
ic
p
ip
elin
e
,
New
s
API
,
wh
ich
s
tr
ea
m
s
ar
ticles
e
v
er
y
5
m
in
u
tes
d
u
r
in
g
tr
ad
in
g
h
o
u
r
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
2
0
2
6
:
416
-
4
2
6
418
ii)
Data
p
r
ep
r
o
ce
s
s
in
g
in
clu
d
e
d
ca
r
ef
u
l
tex
t
clea
n
in
g
p
r
o
ce
s
s
es
s
u
ch
as
r
em
o
v
in
g
n
o
n
-
alp
h
an
u
m
e
r
ic
ch
ar
ac
ter
s
,
tex
t
n
o
r
m
aliza
tio
n
,
lo
wer
ca
s
in
g
tex
t,
ex
p
an
d
in
g
co
n
tr
ac
tio
n
s
,
lem
m
atiza
tio
n
,
an
d
f
ilter
in
g
o
u
t n
o
is
e.
Sen
ten
ce
s
ar
e
p
ad
d
e
d
o
r
tr
u
n
ca
ted
t
o
a
len
g
t
h
o
f
1
2
8
to
k
e
n
s
.
iii)
T
ex
t
em
b
ed
d
in
g
is
cr
itical
in
NL
P
task
s
.
W
e
em
p
lo
y
ed
a
h
y
b
r
id
e
m
b
ed
d
in
g
s
tr
ateg
y
s
im
ilar
to
[
1
]
,
b
u
t
with
m
in
o
r
d
is
tin
ctio
n
s
s
u
ch
as
th
e
u
s
e
o
f
W
o
r
d
2
Vec
[
2
2
]
an
d
Fas
tTe
x
t
[
2
3
]
in
s
tead
o
f
u
s
in
g
Dis
til
B
E
R
T
.
W
o
r
d
2
Vec
[
2
]
was
ef
f
ec
tiv
e
in
ca
p
tu
r
i
n
g
b
r
o
a
d
s
em
an
tic
r
elatio
n
s
h
ip
s
.
Fas
tT
ex
t
[
3
]
h
elp
e
d
with
s
u
b
wo
r
d
in
f
o
r
m
atio
n
,
wh
ich
is
im
p
o
r
tan
t
f
o
r
d
ea
lin
g
with
r
ar
e
f
in
a
n
cial
ter
m
s
.
T
h
e
300
-
d
im
en
s
io
n
al
v
ec
to
r
s
f
r
o
m
ea
ch
ap
p
r
o
ac
h
wer
e
co
m
b
in
ed
to
cr
ea
te
a
6
0
0
-
d
i
m
en
s
io
n
al
wo
r
d
r
ep
r
esen
tatio
n
,
w
h
ich
will
b
e
f
ed
to
th
e
C
NN
-
B
I
L
STM
-
Atten
tio
n
n
etwo
r
k
to
p
r
o
v
id
e
a
r
ich
f
ea
tu
r
e
f
o
u
n
d
atio
n
f
o
r
th
e
m
o
d
el'
s
s
en
tim
en
t a
n
aly
s
is
t
ask
[
2
4
]
.
T
ab
le
1
.
Su
m
m
a
r
y
o
f
r
elate
d
wo
r
k
A
u
t
h
o
r
/
s
R
e
se
a
r
c
h
f
o
c
u
s
M
e
t
h
o
d
o
l
o
g
y
K
e
y
c
o
n
t
r
i
b
u
t
i
o
n
s
G
a
p
s
R
a
n
i
a
n
d
K
u
mar
[
1
7
]
S
e
n
t
i
me
n
t
a
n
a
l
y
si
s
(
S
A
)
u
si
n
g
a
C
N
N
mo
d
e
l
.
−
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
w
i
t
h
v
a
r
y
i
n
g
c
o
n
f
i
g
u
r
a
t
i
o
n
s.
−
C
o
m
p
a
r
i
so
n
w
i
t
h
M
L
b
a
s
e
l
i
n
e
s
(
n
a
i
v
e
B
a
y
e
s (
NB
)
,
s
u
p
p
o
r
t
v
e
c
t
o
r
ma
c
h
i
n
e
(
S
V
M
)
,
k
-
n
e
a
r
e
st
n
e
i
g
h
b
o
r
s (
k
-
NN
)
).
−
M
o
d
e
l
a
c
h
i
e
v
e
d
9
5
%
a
c
c
u
r
a
c
y
.
−
S
h
o
w
e
d
o
p
t
i
ma
l
C
N
N
c
o
n
f
i
g
u
r
a
t
i
o
n
s
D
e
e
p
e
r
C
N
N
s (
3
+
l
a
y
e
r
s)
i
n
c
r
e
a
s
e
d
t
r
a
i
n
i
n
g
t
i
me
w
i
t
h
o
u
t
si
g
n
i
f
i
c
a
n
t
a
c
c
u
r
a
c
y
g
a
i
n
s
.
C
h
e
n
e
t
a
l
.
[
3
5
]
R
e
v
i
e
w
o
f
S
A
a
p
p
l
i
c
a
t
i
o
n
s
i
n
t
h
e
st
o
c
k
mar
k
e
t
.
R
e
v
i
e
w
e
d
2
2
3
a
r
t
i
c
l
e
s fr
o
m
W
e
b
o
f
S
c
i
e
n
c
e
I
d
e
n
t
i
f
i
e
d
k
e
y
t
r
e
n
d
s
:
D
L
,
n
e
w
s s
e
n
t
i
me
n
t
,
i
n
v
e
st
o
r
b
e
h
a
v
i
o
r
.
Ex
c
l
u
d
e
d
n
o
n
-
En
g
l
i
s
h
st
u
d
i
e
s
a
n
d
n
o
n
-
W
o
s
-
i
n
d
e
x
e
d
p
a
p
e
r
s
.
M
e
h
t
a
b
a
n
d
S
e
n
[
3
6
]
S
t
o
c
k
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
u
si
n
g
LSTM
+
S
A
U
sed
LSTM
,
T
w
i
t
t
e
r
S
A
,
sel
f
-
o
r
g
a
n
i
z
i
n
g
f
u
z
z
y
n
e
u
r
a
l
n
e
t
w
o
r
k
(
S
O
F
N
N
)
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
.
S
e
n
t
i
me
n
t
-
a
u
g
m
e
n
t
e
d
S
O
F
N
N
o
u
t
p
e
r
f
o
r
ms
LSTM
a
n
d
t
r
a
d
i
t
i
o
n
a
l
mo
d
e
l
s
.
R
e
a
l
-
t
i
me
a
p
p
l
i
c
a
b
i
l
i
t
y
n
o
t
t
e
st
e
d
.
S
i
d
o
g
i
e
t
a
l
.
[
3
7
]
S
t
o
c
k
p
r
e
d
i
c
t
i
o
n
u
si
n
g
F
i
n
B
ER
T
a
n
d
LS
TM
.
−
LSTM
w
i
t
h
F
i
n
B
E
R
T
f
o
r
S
A
.
−
I
n
t
r
a
d
a
y
p
r
i
c
e
d
a
t
a
w
i
t
h
25
-
m
i
n
u
t
e
l
a
g
.
F
i
n
B
ER
T
i
mp
r
o
v
e
s
p
r
e
d
i
c
t
i
o
n
a
c
c
u
r
a
c
y
o
v
e
r
g
e
n
e
r
a
l
B
ER
T
.
F
o
c
u
se
d
o
n
i
n
t
r
a
d
a
y
d
a
t
a
;
n
o
t
v
a
l
i
d
a
t
e
d
f
o
r
l
o
n
g
-
t
e
r
m
p
r
e
d
i
c
t
i
o
n
s
.
T
u
l
e
t
a
l
.
[
3
8
]
R
e
v
i
e
w
o
f
D
L
t
e
c
h
n
i
q
u
e
s f
o
r
S
A
.
Li
t
e
r
a
t
u
r
e
r
e
v
i
e
w
o
f
C
N
N
,
R
N
N
,
D
B
N
,
a
n
d
h
y
b
r
i
d
m
o
d
e
l
s i
n
S
A
.
C
o
m
p
r
e
h
e
n
si
v
e
o
v
e
r
v
i
e
w
o
f
D
L
a
p
p
l
i
c
a
t
i
o
n
s i
n
S
A
.
N
o
n
e
w
e
mp
i
r
i
c
a
l
met
h
o
d
s
p
r
o
p
o
se
d
.
S
i
n
g
h
e
t
a
l
.
[
2
0
]
S
A
i
s
u
s
i
n
g
t
r
a
d
i
t
i
o
n
a
l
M
L
c
l
a
ss
i
f
i
e
r
s.
−
U
sed
N
B
,
J
4
8
,
B
F
T
r
e
e
,
O
n
e
R
;
f
e
a
t
u
r
e
s
e
l
e
c
t
i
o
n
(
D
F
,
M
I
,
I
G
)
.
−
W
EK
A
e
v
a
l
u
a
t
i
o
n
.
N
B
w
a
s f
a
st
e
st
,
O
n
e
R
mo
s
t
p
r
e
c
i
se
w
i
t
h
a
c
c
9
2
.
3
4
%
.
La
c
k
s
a
d
v
a
n
c
e
d
p
r
e
p
r
o
c
e
ssi
n
g
l
i
k
e
w
o
r
d
e
m
b
e
d
d
i
n
g
s.
S
o
u
m
a
e
t
a
l
.
[
3
9
]
S
A
f
o
r
e
c
a
s
t
i
n
g
u
si
n
g
R
N
N
-
LSTM
U
sed
G
l
o
V
e
w
o
r
d
v
e
c
t
o
r
s
;
t
r
a
i
n
e
d
o
n
T
h
o
mso
n
R
e
u
t
e
r
s
n
e
w
s
a
n
a
l
y
t
i
c
s
(
TR
N
A
)
a
n
d
D
o
w
J
o
n
e
s
i
n
d
u
st
r
i
a
l
a
v
e
r
a
g
e
(
D
JI
A
)
st
o
c
k
d
a
t
a
.
A
c
c
i
mp
r
o
v
e
d
w
i
t
h
h
i
e
r
a
r
c
h
i
c
a
l
t
r
a
i
n
i
n
g
.
R
e
l
i
a
n
c
e
o
n
st
o
c
k
p
r
i
c
e
f
l
u
c
t
u
a
t
i
o
n
s
f
o
r
sen
t
i
m
e
n
t
l
a
b
e
l
l
i
n
g
a
n
d
n
o
m
u
l
t
i
l
i
n
g
u
a
l
su
p
p
o
r
t
.
Fig
u
r
e
1
.
Sen
tim
en
t a
n
aly
s
is
m
o
d
el
ar
ch
itectu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
A
d
a
p
tive
s
en
timen
t a
n
a
lysi
s
fo
r
s
to
ck
ma
r
ke
ts
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ta
len
t Ma
w
ere
)
419
2
.
2
.
M
o
del
a
rc
hite
ct
ure
T
h
e
c
e
n
t
r
a
l
c
o
m
p
o
n
e
n
t
o
f
o
u
r
f
r
a
m
e
w
o
r
k
i
s
t
h
e
C
NN
-
B
i
L
S
T
M
-
A
t
t
e
n
ti
o
n
n
e
t
w
o
r
k
,
w
h
i
c
h
i
s
s
u
p
p
o
r
t
e
d
b
y
f
o
u
r
l
a
y
e
r
s
.
A
m
u
l
t
i
-
s
c
al
e
C
N
N
f
o
r
l
o
c
a
l
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
f
o
r
m
s
t
h
e
i
n
i
t
i
a
l
l
a
y
e
r
.
I
n
t
h
i
s
la
y
e
r
,
t
h
e
e
m
b
e
d
d
e
d
i
n
p
u
t
s
e
q
u
e
n
c
e
g
o
es
t
h
r
o
u
g
h
f
i
v
e
p
a
r
a
l
l
el
1
D
c
o
n
v
o
l
u
t
i
o
n
al
l
a
y
e
r
s
w
it
h
k
e
r
n
e
l
s
i
z
es
o
f
1
,
2
,
3
,
4
,
a
n
d
5
,
f
a
c
i
l
it
a
t
i
n
g
t
h
e
m
o
d
e
l
t
o
f
i
n
d
s
e
n
t
i
m
e
n
t
c
u
e
s
at
d
i
f
f
e
r
e
n
t
l
e
v
els
,
f
r
o
m
u
n
i
g
r
a
m
s
t
o
5
-
g
r
a
m
s
.
T
w
o
h
u
n
d
r
e
d
f
i
l
t
e
r
s
w
i
l
l
b
e
u
s
e
d
a
t
e
a
c
h
l
a
y
er
,
f
o
l
l
o
w
e
d
b
y
a
r
e
c
t
i
f
i
e
d
l
i
n
e
a
r
u
n
i
t
(
R
e
L
U
)
a
c
ti
v
a
t
i
o
n
a
n
d
g
l
o
b
a
l
m
a
x
-
p
o
o
l
i
n
g
,
r
e
s
u
l
t
i
n
g
i
n
a
2
0
0
-
d
i
m
e
n
s
i
o
n
al
v
e
c
t
o
r
p
e
r
k
e
r
n
e
l
a
n
d
a
c
o
n
c
at
e
n
a
t
e
d
1
,
000
-
d
i
m
e
n
s
i
o
n
a
l
f
e
atu
r
e
v
e
c
t
o
r
f
o
r
i
n
p
u
t
t
o
t
h
e
n
e
x
t
l
a
y
e
r
.
I
n
t
h
e
s
e
co
n
d
l
a
y
er
,
a
Bi
L
S
T
M
l
ay
e
r
,
wh
i
c
h
wo
r
k
s
in
p
a
r
a
l
l
e
l
to
th
e
C
N
N
l
ay
er
an
d
i
s
d
e
s
ig
n
ed
f
o
r
s
e
q
u
en
t
i
a
l
c
o
n
t
e
x
t
m
o
d
e
l
in
g
,
a
l
s
o
t
ak
e
s
i
n
p
u
t
f
r
o
m
th
e
o
r
i
g
in
a
l
e
m
b
e
d
d
e
d
s
eq
u
en
c
e.
W
i
t
h
in
th
e
l
a
y
e
r
,
t
e
x
t
i
s
an
a
l
y
ze
d
i
n
b
o
t
h
f
o
r
wa
r
d
a
n
d
b
ac
k
w
ar
d
d
ir
e
c
t
io
n
s
t
o
ca
p
tu
r
e
l
o
n
g
-
r
an
g
e
d
e
p
e
n
d
e
n
c
i
e
s
a
n
d
c
o
n
t
e
x
t,
r
e
s
u
l
t
i
n
g
i
n
a
s
eq
u
en
c
e
o
f
h
id
d
en
s
t
a
t
e
s
.
T
h
e
t
h
i
r
d
l
ay
e
r
,
w
o
r
k
in
g
in
s
e
q
u
e
n
ce
w
i
t
h
th
e
B
i
L
S
T
M
l
a
y
er
,
c
o
n
s
i
s
t
s
o
f
a
n
a
t
t
en
t
i
o
n
m
e
c
h
a
n
i
s
m
f
o
r
s
a
l
i
en
c
e
w
e
i
g
h
t
in
g
.
H
i
d
d
en
s
t
a
t
e
s
o
u
t
p
u
t
f
r
o
m
th
e
B
i
L
S
T
M
l
ay
er
a
r
e
in
g
e
s
t
ed
,
an
d
a
w
e
ig
h
t
ed
s
u
m
c
o
n
t
ex
t
v
e
c
t
o
r
i
s
c
a
lc
u
l
at
e
d
.
A
cc
o
r
d
in
g
to
V
a
s
w
an
i
e
t
a
l
.
[
2
5
]
,
a
t
t
en
t
i
o
n
m
e
c
h
an
i
s
m
s
h
e
lp
a
m
o
d
e
l
t
o
f
o
cu
s
o
n
t
h
e
m
o
s
t
i
m
p
o
r
t
a
n
t
s
e
n
t
im
e
n
t
-
r
e
l
a
te
d
w
o
r
d
s
w
h
i
l
e
i
g
n
o
r
i
n
g
ir
r
e
l
ev
a
n
t
i
n
f
o
r
m
a
t
i
o
n
,
th
er
eb
y
im
p
r
o
v
in
g
a
n
e
u
r
a
l
n
e
t
wo
r
k
’
s
p
er
f
o
r
m
a
n
ce
.
T
h
e
o
u
tp
u
t
o
f
t
h
i
s
l
ay
er
i
s
a
128
-
d
im
e
n
s
io
n
a
l
a
t
t
en
t
i
o
n
-
w
ei
g
h
t
ed
v
e
c
t
o
r
.
B
o
th
in
p
u
ts
f
r
o
m
th
e
C
NN
a
n
d
th
e
C
NN
-
B
iLST
M
-
Atten
ti
o
n
lay
e
r
ar
e
co
m
b
in
ed
an
d
f
ed
in
to
a
class
if
icatio
n
lay
er
.
T
h
is
co
m
b
in
ed
r
e
p
r
esen
tatio
n
g
o
es
th
r
o
u
g
h
a
d
r
o
p
o
u
t
lay
er
at
a
r
ate
o
f
0
.
5
f
o
r
r
eg
u
lar
izatio
n
.
A
1
2
8
-
u
n
it
d
e
n
s
e
lay
er
with
R
eL
U
ac
tiv
atio
n
an
d
a
s
o
f
tm
ax
lay
er
f
o
r
a
th
r
ee
-
class
s
en
tim
en
t
class
if
icatio
n
,
p
o
s
itiv
e,
n
eg
ativ
e,
o
r
n
eu
tr
al.
2
.
3
.
Ada
ptiv
e
le
a
rni
ng
m
ec
ha
nis
m
Mo
s
t
f
in
an
cial
m
ar
k
ets
ex
h
ib
it
in
h
er
en
t
n
o
n
-
s
tatio
n
ar
ity
[
2
6
]
,
[
2
7
]
n
ec
ess
itatin
g
an
ad
ap
tiv
e
f
r
am
ewo
r
k
th
at
co
n
tin
u
o
u
s
ly
u
p
d
ates
to
e
v
o
lv
in
g
s
en
tim
en
t
p
atter
n
s
.
T
o
im
p
lem
en
t
th
is
,
we
in
co
r
p
o
r
ated
a
n
in
tellig
en
t
th
r
ee
-
c
o
m
p
o
n
en
t
ad
ap
tiv
e
m
ec
h
an
is
m
.
Firstl
y
,
we
im
p
lem
en
t
a
f
ee
d
b
ac
k
lo
o
p
ar
ch
itectu
r
e
o
p
er
atin
g
o
n
a
clo
s
ed
-
l
o
o
p
co
n
tr
o
l
p
r
in
cip
le,
wh
er
e
o
u
t
p
u
ts
f
r
o
m
th
e
C
NN
-
B
iLST
M
-
Atten
tio
n
n
eu
r
al
n
etwo
r
k
s
er
v
e
as
th
e
in
p
u
t
s
ig
n
al
f
o
r
c
o
n
tin
u
o
u
s
ad
a
p
tatio
n
.
W
e
u
s
e
th
e
r
iv
e
r
lib
r
a
r
y
'
s
ad
ap
tiv
e
win
d
o
win
g
(
ADWIN)
alg
o
r
ith
m
[
2
8
]
to
c
h
ec
k
th
e
e
n
tr
o
p
y
o
f
th
e
p
r
ed
ic
tio
n
d
is
tr
ib
u
tio
n
i
n
r
ea
l
tim
e.
T
h
is
m
ak
es
it
ea
s
ier
to
s
et
u
p
a
d
ir
ec
t
f
ee
d
b
ac
k
lo
o
p
b
etwe
en
m
o
d
el
p
er
f
o
r
m
an
ce
an
d
p
ar
am
eter
u
p
d
ates.
T
h
e
s
y
s
tem
m
ak
e
s
tar
g
eted
ch
an
g
es
wh
en
th
e
en
tr
o
p
y
g
o
es
o
v
e
r
a
ce
r
tain
lev
el,
lik
e
(
H
>0
.
8
5
)
,
wh
ich
p
r
o
b
a
b
ly
m
ea
n
s
th
at
th
e
co
n
f
id
en
ce
in
th
e
p
r
ed
ictio
n
is
g
o
in
g
d
o
wn
.
Gu
o
et
a
l.
[
2
9
]
f
o
u
n
d
t
h
at
m
is
ca
lib
r
atio
n
is
a
co
m
m
o
n
p
r
o
b
l
em
with
n
eu
r
al
n
etwo
r
k
s
.
W
e
s
u
g
g
est
a
me
th
o
d
t
h
at
u
s
es
d
y
n
am
ic
p
ar
am
eter
s
ca
lin
g
t
o
m
a
k
e
ca
lib
r
atio
n
m
o
r
e
m
a
r
k
et
-
awa
r
e.
B
y
lo
wer
in
g
th
e
en
tr
o
p
y
t
h
r
esh
o
ld
a
n
d
r
aisi
n
g
th
e
lear
n
in
g
r
ates
in
tellig
en
tly
,
s
ca
lin
g
lets
a
m
o
d
el
b
ec
o
m
e
m
o
r
e
s
en
s
itiv
e
wh
en
th
e
m
ar
k
et
is
v
er
y
v
o
l
atile.
T
h
is
m
ak
es
s
u
r
e
t
h
a
t
th
e
m
o
d
el
r
ea
cts
q
u
ick
l
y
to
ev
en
ts
th
at
m
o
v
e
th
e
m
ar
k
et.
I
n
s
tab
le
p
er
io
d
s
,
th
e
s
y
s
tem
k
ee
p
s
th
e
n
o
r
m
al
s
ettin
g
s
,
wh
ich
s
to
p
s
it
f
r
o
m
o
v
er
f
itti
n
g
to
s
h
o
r
t
-
ter
m
n
o
is
e.
B
y
a
p
p
ly
in
g
t
h
is
co
n
f
i
g
u
r
atio
n
,
we
m
a
k
e
s
u
r
e
th
at
th
e
f
r
am
ewo
r
k
f
o
c
u
s
es
o
n
ad
ap
ti
n
g
d
u
r
in
g
tim
es
o
f
s
ig
n
if
ican
t m
ar
k
et
ch
an
g
e.
Z
ar
g
h
an
i
an
d
Ab
ed
i
[
3
0
]
n
o
te
d
th
at
s
lid
in
g
win
d
o
w
tech
n
iq
u
es
ar
e
v
er
y
im
p
o
r
tan
t
f
o
r
wo
r
k
in
g
with
s
tr
ea
m
s
o
f
d
ata.
T
h
ey
d
id
,
h
o
wev
er
,
p
o
in
t
o
u
t
th
at
f
ix
ed
-
s
ize
win
d
o
ws
h
av
e
tr
o
u
b
le
ad
ap
t
in
g
to
ch
an
g
es
th
at
h
ap
p
en
q
u
ick
ly
,
lik
e
b
u
r
s
ty
p
a
tter
n
s
o
r
co
n
ce
p
t
d
r
if
t.
Ou
r
th
i
r
d
s
tr
ateg
y
is
to
u
s
e
an
in
tellig
en
t
win
d
o
w
s
izin
g
tech
n
iq
u
e
b
ased
o
n
m
ar
k
et
c
o
n
d
itio
n
s
to
f
ix
t
h
is
.
W
e
s
u
g
g
est
a
4
8
-
h
o
u
r
r
o
llin
g
win
d
o
w
f
o
r
in
cr
em
e
n
tal
u
p
d
ates
th
at
s
tr
ik
es
th
e
b
est
b
alan
ce
b
etwe
en
b
ein
g
q
u
ic
k
t
o
r
esp
o
n
d
a
n
d
b
ein
g
s
tatis
tica
lly
r
eliab
le.
W
h
e
n
s
p
ik
es
in
v
o
latilit
y
ar
e
d
etec
t
ed
,
th
e
s
y
s
tem
au
to
m
atica
lly
s
h
o
r
ten
s
th
is
win
d
o
w
to
2
4
h
o
u
r
s
to
g
iv
e
m
o
r
e
weig
h
t
to
r
ec
en
t,
h
ig
h
-
im
p
ac
t
d
ata.
T
o
k
ee
p
th
e
m
o
d
el
s
tab
le,
th
e
win
d
o
w
s
ize
is
in
cr
e
ase
d
d
u
r
in
g
s
tab
le
tim
es.
W
h
en
d
r
if
t
is
f
o
u
n
d
,
t
h
e
f
r
am
ewo
r
k
u
s
es
a
Ho
ef
f
d
in
g
T
r
ee
class
if
ier
[
3
1
]
tr
ain
e
d
o
n
m
in
i
-
b
atch
es
(
2
5
6
s
am
p
les)
f
r
o
m
th
e
o
p
ti
m
ized
tim
e
win
d
o
w
t
o
m
ak
e
tar
g
eted
u
p
d
ates.
T
h
is
m
eth
o
d
m
ak
es
it
ea
s
y
t
o
q
u
ick
ly
ad
a
p
t
to
n
ew
m
ar
k
et
tr
en
d
s
wh
ile
s
till
b
ein
g
ef
f
icien
t
with
co
m
p
u
ter
s
,
s
in
ce
o
n
ly
th
e
tem
p
o
r
al
atten
tio
n
p
ar
am
eter
s
ar
e
ch
a
n
g
ed
,
an
d
th
e
co
r
e
s
en
tim
e
n
t
class
if
icatio
n
lay
er
s
s
tay
th
e
s
am
e
to
av
o
id
ca
tast
r
o
p
h
ic
f
o
r
g
ettin
g
.
2
.
4
.
Str
a
t
eg
y
f
o
r
f
ine
-
t
un
ing
I
n
ad
d
itio
n
to
t
h
e
r
ea
l
-
tim
e
ad
ap
tatio
n
s
tr
ateg
y
,
we
d
is
c
u
s
s
ed
ab
o
v
e,
we
s
u
g
g
est
a
th
r
ee
-
p
h
ase
ad
ap
tatio
n
a
p
p
r
o
ac
h
to
f
in
e
-
tu
n
e
o
u
r
n
etwo
r
k
.
B
ased
o
n
Ho
war
d
a
n
d
R
u
d
e
r
'
s
wo
r
k
o
n
d
is
cr
im
in
ativ
e
f
in
e
-
tu
n
in
g
[
3
2
]
,
we
s
u
g
g
est
an
ap
p
r
o
ac
h
o
f
g
r
a
d
u
ally
u
n
f
r
ee
zin
g
lay
er
s
to
h
elp
p
r
ev
en
t
ca
tast
r
o
p
h
ic
f
o
r
g
ettin
g
wh
ile
allo
win
g
e
f
f
e
ctiv
e
d
o
m
ain
tr
an
s
f
er
.
T
h
i
s
k
i
n
d
o
f
s
tr
ateg
y
c
o
u
ld
wo
r
k
well
in
f
in
an
ce
,
wh
er
e
it's
im
p
o
r
tan
t
to
ex
p
an
d
v
o
ca
b
u
lar
y
an
d
ca
lib
r
ate
s
en
tim
e
n
t
in
ten
s
ity
to
d
ea
l
with
p
r
o
b
lem
s
with
d
o
m
ain
ad
ap
tatio
n
[
1
0
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
2
0
2
6
:
416
-
4
2
6
420
An
o
th
er
m
eth
o
d
is
m
u
lti
-
task
lear
n
in
g
,
in
wh
ich
we
ad
d
o
th
er
g
o
als,
s
u
ch
as p
r
ed
ictin
g
v
o
latilit
y
an
d
d
etec
tin
g
u
r
g
e
n
cy
.
T
h
e
r
esu
l
ts
o
f
[
3
3
]
,
wh
o
s
h
o
wed
th
at
au
x
iliar
y
task
s
im
p
r
o
v
e
m
o
d
el
r
o
b
u
s
tn
ess
an
d
g
en
er
aliza
tio
n
.
Fin
ally
,
we
r
e
co
m
m
en
d
a
co
n
tin
u
o
u
s
lear
n
i
n
g
m
eth
o
d
th
at
ca
n
d
etec
t
co
n
ce
p
t
d
r
i
f
t,
wh
ich
is
in
lin
e
with
th
e
in
cr
e
m
en
tal
ad
ap
tatio
n
id
ea
s
p
u
t
f
o
r
t
h
b
y
Gam
a
et
a
l
.
[
3
4
]
f
o
r
e
n
v
ir
o
n
m
en
ts
th
at
ar
e
n
o
t
s
tab
le.
T
h
is
m
eth
o
d
g
u
ar
a
n
tees
co
n
s
is
ten
t
lo
n
g
-
ter
m
p
e
r
f
o
r
m
a
n
ce
wh
ile
m
ain
tain
in
g
th
e
lo
w
-
laten
cy
in
f
er
en
ce
r
eq
u
ir
ed
f
o
r
r
ea
l
-
ti
m
e
tr
ad
in
g
a
p
p
licatio
n
s
.
T
h
ese
th
r
ee
ap
p
r
o
ac
h
es
w
o
r
k
well
to
g
eth
e
r
to
b
alan
ce
ad
ap
tatio
n
ef
f
icien
cy
a
n
d
p
r
ac
t
ical
d
ep
lo
y
m
e
n
t,
wh
ich
is
wh
at
f
in
tech
AI
r
esear
ch
n
ee
d
s
.
T
ab
le
2
p
r
esen
ts
th
e
f
in
al
in
ter
n
al
s
tr
u
ctu
r
e
o
f
o
u
r
m
o
d
el,
wh
ich
will b
e
tr
ain
ed
a
n
d
test
ed
.
T
ab
le
2
.
I
n
ter
n
al
s
tr
u
ctu
r
e
o
f
t
h
e
co
m
p
lete
m
o
d
el
La
y
e
r
/
c
o
m
p
o
n
e
n
t
C
o
n
f
i
g
u
r
a
t
i
o
n
O
u
t
p
u
t
s
h
a
p
e
P
a
r
a
me
t
e
r
s
F
u
n
c
t
i
o
n
I
n
p
u
t
l
a
y
e
r
S
e
q
u
e
n
c
e
l
e
n
g
t
h
:
1
2
8
Emb
e
d
d
i
n
g
d
i
m
:
6
0
0
(
1
2
8
,
6
0
0
)
-
R
e
c
e
i
v
e
s
p
a
d
d
e
d
t
e
x
t
se
q
u
e
n
c
e
s
M
u
l
t
i
-
sca
l
e
C
N
N
1
st
l
a
y
e
r
C
o
n
v
1
D
(
k
=
1
)
F
i
l
t
e
r
s:
2
0
0
,
K
e
r
n
e
l
:
1
(
1
2
8
,
2
0
0
)
1
2
0
,
2
0
0
U
n
i
g
r
a
m
p
a
t
t
e
r
n
s
2
nd
l
a
y
e
r
C
o
n
v
1
D
(
k
=
2
)
F
i
l
t
e
r
s:
2
0
0
,
K
e
r
n
e
l
:
2
(
1
2
7
,
2
0
0
)
2
4
0
,
2
0
0
B
i
g
r
a
m
p
a
t
t
e
r
n
s
3
rd
l
a
y
e
r
C
o
n
v
1
D
(
k
=
3
)
F
i
l
t
e
r
s:
2
0
0
,
K
e
r
n
e
l
:
3
(
1
2
6
,
2
0
0
)
3
6
0
,
2
0
0
Tr
i
g
r
a
m
p
a
t
t
e
r
n
s
4
th
l
a
y
e
r
C
o
n
v
1
D
(
k
=
4
)
F
i
l
t
e
r
s:
2
0
0
,
K
e
r
n
e
l
:
4
(
1
2
5
,
2
0
0
)
4
8
0
,
2
0
0
4
-
g
r
a
m
p
a
t
t
e
r
n
s
5
th
l
a
y
e
r
C
o
n
v
1
D
(
k
=
5
)
F
i
l
t
e
r
s:
2
0
0
,
K
e
r
n
e
l
:
5
(
1
2
4
,
2
0
0
)
6
0
0
,
2
0
0
5
-
g
r
a
m
p
a
t
t
e
r
n
s
G
l
o
b
a
l
ma
x
p
o
o
l
i
n
g
A
p
p
l
i
e
d
p
e
r
b
r
a
n
c
h
(
2
0
0
,
)
e
a
c
h
-
Ex
t
r
a
c
t
s m
o
st
sal
i
e
n
t
f
e
a
t
u
r
e
s
C
o
n
c
a
t
e
n
a
t
i
o
n
M
e
r
g
e
a
l
l
C
N
N
b
r
a
n
c
h
e
s
(
1
0
0
0
,
)
-
C
o
m
b
i
n
e
s m
u
l
t
i
-
sc
a
l
e
f
e
a
t
u
r
e
s
B
i
LST
M
l
a
y
e
r
U
n
i
t
s:
1
2
8
R
e
t
u
r
n
se
q
u
e
n
c
e
s:
t
r
u
e
(
1
2
8
,
2
5
6
)
1
,
4
9
4
,
5
2
8
C
a
p
t
u
r
e
s
l
o
n
g
-
r
a
n
g
e
d
e
p
e
n
d
e
n
c
i
e
s
F
o
r
w
a
r
d
a
n
d
b
a
c
k
w
a
r
d
=
2
5
6
u
n
i
t
s
A
t
t
e
n
t
i
o
n
mec
h
a
n
i
sm
S
c
o
r
e
c
a
l
c
u
l
a
t
i
o
n
t
a
n
h
a
c
t
i
v
a
t
i
o
n
(
1
2
8
,
2
5
6
)
6
5
,
7
9
2
C
o
m
p
u
t
e
s a
t
t
e
n
t
i
o
n
sc
o
r
e
s
W
e
i
g
h
t
c
o
m
p
u
t
a
t
i
o
n
so
f
t
m
a
x
n
o
r
ma
l
i
z
a
t
i
o
n
(
1
2
8
,
1
)
-
C
o
n
v
e
r
t
s
t
o
a
p
r
o
b
a
b
i
l
i
t
y
d
i
s
t
r
i
b
u
t
i
o
n
C
o
n
t
e
x
t
v
e
c
t
o
r
W
e
i
g
h
t
e
d
s
u
m
(
2
5
6
,
)
-
F
o
c
u
ses
o
n
r
e
l
e
v
a
n
t
t
e
mp
o
r
a
l
f
e
a
t
u
r
e
s
F
e
a
t
u
r
e
f
u
si
o
n
C
o
n
c
a
t
e
n
a
t
e
C
N
N
+
A
t
t
e
n
t
i
o
n
(
1
2
5
6
,
)
-
C
o
m
b
i
n
e
s l
o
c
a
l
+
g
l
o
b
a
l
p
a
t
t
e
r
n
s
D
r
o
p
o
u
t
R
a
t
e
:
0
.
5
(
1
2
5
6
,
)
-
P
r
e
v
e
n
t
s
o
v
e
r
f
i
t
t
i
n
g
D
e
n
se
l
a
y
e
r
U
n
i
t
s:
1
2
8
,
R
e
LU
(
1
2
8
,
)
1
6
0
,
8
9
6
F
i
n
a
l
f
e
a
t
u
r
e
t
r
a
n
sf
o
r
m
a
t
i
o
n
O
u
t
p
u
t
l
a
y
e
r
U
n
i
t
s :
3
,
so
f
t
ma
x
(
3
,
)
3
8
7
3
-
c
l
a
ss s
e
n
t
i
m
e
n
t
c
l
a
ssi
f
i
c
a
t
i
o
n
2
.
5
.
E
x
perim
ent
a
l set
up
a
nd
ev
a
lua
t
io
n m
e
t
rics
W
e
s
p
lit
o
u
r
d
ata
in
to
th
r
ee
s
ets,
tr
ain
in
g
(
7
0
%),
v
alid
atio
n
(
1
5
%)
,
an
d
test
in
g
(
1
5
%),
o
r
g
an
ized
b
y
tim
e
to
av
o
id
tem
p
o
r
al
leak
a
g
e
,
an
d
tr
ain
ed
it
o
v
er
1
0
0
ep
o
ch
s
with
an
Ad
am
W
o
p
tim
izer
to
m
in
im
ize
th
e
co
m
p
o
s
ite
lo
s
s
f
u
n
ctio
n
.
T
o
e
v
alu
ate
th
e
p
er
f
o
r
m
a
n
ce
o
f
o
u
r
m
o
d
el,
we
co
n
d
u
cted
two
t
y
p
es
o
f
co
m
p
a
r
is
o
n
s
,
in
clu
d
in
g
an
a
b
latio
n
s
tu
d
y
to
ass
es
s
th
e
atten
tio
n
n
etwo
r
k
s
’
co
n
tr
ib
u
tio
n
to
o
u
r
m
o
d
el'
s
p
er
f
o
r
m
an
ce
an
d
a
b
en
ch
m
ar
k
c
o
m
p
ar
is
o
n
ag
ai
n
s
t
f
iv
e
o
th
e
r
b
aselin
es.
T
h
ese
in
clu
d
e
Dis
tilB
E
R
T
,
a
co
m
p
ac
t
tr
an
s
f
o
r
m
er
;
L
STM
;
a
C
NN
-
B
iL
STM
wit
h
o
u
t
atten
tio
n
;
VADE
R
;
Fas
tTe
x
t;
an
d
T
e
x
tB
lo
b
.
W
e
s
elec
ted
b
en
ch
m
ar
k
m
o
d
els
b
ased
o
n
th
eir
p
er
f
o
r
m
an
ce
r
elev
a
n
ce
an
d
th
eir
a
b
ilit
y
to
b
e
d
e
p
lo
y
e
d
in
r
ea
l
-
tim
e.
W
h
ile
m
o
d
els
lik
e
Fin
B
E
R
T
p
r
o
v
id
e
g
r
ea
t
a
cc
u
r
ac
y
,
th
ey
ar
e
c
o
m
p
u
ta
ti
o
n
ally
co
m
p
lex
a
n
d
less
s
u
itab
le
f
o
r
th
e
h
ig
h
-
th
r
o
u
g
h
p
u
t,
lo
w
-
laten
cy
p
ip
elin
e
ce
n
t
r
al
to
o
u
r
s
tu
d
y
;
we
d
id
n
o
t
co
n
s
id
er
th
em
in
o
u
r
ev
alu
atio
n
.
R
ath
er
,
we
f
o
cu
s
o
n
m
o
d
els
th
at
ac
h
iev
e
h
ig
h
ac
cu
r
ac
y
with
co
m
p
u
tatio
n
al
ef
f
icien
cy
.
W
e
ev
alu
at
e
p
er
f
o
r
m
an
ce
u
s
in
g
s
tan
d
ar
d
c
lass
if
icatio
n
m
etr
ics
,
in
clu
d
in
g
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all
,
F1
-
s
co
r
e,
an
d
ar
ea
u
n
d
er
t
h
e
cu
r
v
e
(
AUC).
A
co
n
f
u
s
io
n
m
atr
ix
is
u
s
ed
f
o
r
e
r
r
o
r
an
aly
s
is
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
Abla
t
io
n study
a
nd
hy
perpa
ra
m
et
er
o
ptim
iza
t
i
o
n
As
d
is
cu
s
s
ed
ea
r
lier
,
th
e
in
iti
al
ev
alu
atio
n
p
lan
in
v
o
l
v
es
u
n
d
er
tak
in
g
an
ab
latio
n
s
tu
d
y
to
o
b
s
er
v
e
o
v
er
all
co
n
tr
ib
u
tio
n
o
f
ea
ch
c
o
m
p
o
n
en
t.
Sen
tim
en
t
an
aly
s
is
u
s
in
g
a
C
NN
y
ield
e
d
a
v
alid
atio
n
ac
cu
r
ac
y
o
f
6
2
%
v
alid
atio
n
lo
s
s
.
W
ith
o
u
r
f
in
al
m
o
d
el
C
NN
-
B
iL
STM
-
Att
en
tio
n
,
y
ield
in
g
a
v
alid
atio
n
ac
cu
r
ac
y
an
d
lo
s
s
o
f
8
2
% a
n
d
0
.
2
5
%
,
co
n
f
ir
m
in
g
an
o
v
er
all
co
n
t
r
ib
u
tio
n
o
f
2
0
% a
f
ter
all
co
m
p
o
n
en
ts
ar
e
in
t
eg
r
ated
.
Af
ter
p
er
f
o
r
m
in
g
a
co
m
p
o
n
e
n
t
-
wis
e
ab
latio
n
s
tu
d
y
v
alid
atin
g
ea
ch
co
m
p
o
n
en
t’
s
co
n
t
r
ib
u
t
io
n
to
th
e
o
v
er
all
m
o
d
el
’
s
p
er
f
o
r
m
an
ce
,
we
th
en
ap
p
lied
Op
tu
n
a
f
o
r
h
y
p
er
p
ar
am
eter
o
p
tim
izatio
n
f
o
r
b
o
th
o
u
r
p
r
o
p
o
s
e
d
m
o
d
el
an
d
th
e
b
aselin
es.
Fig
u
r
e
2
s
h
o
ws
th
e
t
r
ain
in
g
a
n
d
v
alid
atio
n
ac
cu
r
ac
y
(
Fig
u
r
e
2
(
a)
)
a
n
d
l
o
s
s
(
Fig
u
r
e
2
(
b
)
)
p
lo
ts
f
o
r
o
u
r
m
o
d
el
af
ter
th
e
n
ew
c
o
n
f
ig
u
r
atio
n
s
.
W
e
o
b
s
er
v
e
th
at
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
r
ea
ch
e
d
a
v
alid
atio
n
ac
cu
r
ac
y
o
f
9
7
.
4
%
with
a
v
alid
atio
n
lo
s
s
s
tab
ilizin
g
u
n
d
er
0
.
1
.
W
e
attr
ib
u
te
th
e
cr
e
d
it
to
th
e
atten
tio
n
m
ec
h
an
is
m
,
wh
ich
a
cted
as a
r
eg
u
lar
izer
,
s
t
ab
ilizin
g
tr
ain
in
g
an
d
en
h
a
n
cin
g
g
e
n
e
r
aliza
tio
n
as sh
o
wn
in
T
ab
le
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
A
d
a
p
tive
s
en
timen
t a
n
a
lysi
s
fo
r
s
to
ck
ma
r
ke
ts
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ta
len
t Ma
w
ere
)
421
T
ab
le
3
.
Ab
latio
n
s
tu
d
y
v
alid
a
tio
n
lo
s
s
an
d
ac
cu
r
ac
y
v
alu
es
M
e
t
r
i
c
C
N
N
(
%)
B
i
LST
M
(
%)
S
e
n
t
i
w
a
r
e
t
r
a
d
e
(
n
o
a
t
t
e
n
t
i
o
n
)
(
%)
S
e
n
t
i
w
a
r
e
t
r
a
d
e
(
w
i
t
h
a
t
t
e
n
t
i
o
n
)
(
%)
V
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
62
70
75
82
V
a
l
i
d
a
t
i
o
n
l
o
ss
0
.
6
8
0
.
4
0
0
.
3
5
0
.
2
5
G
e
n
e
r
a
l
i
z
a
t
i
o
n
g
a
p
5
10
5
3
(
a)
(
b
)
Fig
u
r
e
2
.
Per
f
o
r
m
an
c
e
p
lo
ts
f
o
r
test
ed
m
o
d
els s
h
o
win
g
(
a)
ac
cu
r
ac
y
an
d
(
b
)
lo
s
s
3
.
2
.
M
o
del
perf
o
rma
nce
a
n
d benchm
a
rk
ing
T
a
b
le
4
s
u
m
m
a
r
i
ze
s
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
o
u
r
p
r
o
p
o
s
e
d
m
o
d
el
a
g
ai
n
s
t
t
h
e
p
r
o
p
o
s
e
d
b
ase
lin
es.
O
u
r
m
o
d
el,
C
N
N
-
B
i
L
ST
M
-
Att
en
t
io
n
(
S
e
n
ti
T
r
ad
e
r
)
,
a
ch
ie
v
e
d
t
o
p
r
esu
l
ts
,
w
it
h
g
o
o
d
s
c
o
r
es
a
cr
o
s
s
a
ll
m
et
r
i
cs
(
a
cc
u
r
ac
y
:
0
.
9
7
5
,
F
1
-
s
co
r
e
:
0
.
9
8
2
,
a
n
d
AUC:
0
.
9
7
3
)
d
e
m
o
n
s
tr
ati
n
g
o
u
r
m
o
d
el
’
s
s
u
p
e
r
i
o
r
it
y
i
n
m
o
d
elli
n
g
s
en
t
im
en
t
d
at
a
i
n
f
i
n
a
n
c
ial
m
ar
k
e
ts
.
W
e
s
u
p
p
o
r
t
t
h
e
h
y
p
o
t
h
esis
t
h
at
h
y
b
r
id
a
r
c
h
it
ec
t
u
r
e
p
e
r
f
o
r
m
s
b
ett
er
t
h
a
n
s
tan
d
al
o
n
es
s
i
n
ce
ev
e
n
Se
n
ti
T
r
a
d
e
r
,
wit
h
o
u
t
att
e
n
ti
o
n
,
a
lr
ea
d
y
o
u
t
p
er
f
o
r
m
e
d
t
h
e
s
t
a
n
d
al
o
n
e
L
S
T
M
b
y
3
.
1
%
in
ac
c
u
r
a
cy
.
Ou
r
f
i
n
d
i
n
g
s
f
u
r
t
h
er
v
ali
d
a
te
r
esu
lts
f
r
o
m
t
h
e
a
b
la
tio
n
s
tu
d
y
b
y
c
o
n
f
i
r
m
i
n
g
t
h
a
t
atte
n
t
io
n
n
etw
o
r
k
s
im
p
r
o
v
e
p
r
ed
icti
o
n
s
,
as
e
v
i
d
e
n
c
ed
b
y
t
h
e
i
m
p
r
o
v
e
d
p
e
r
f
o
r
m
a
n
c
e
w
h
en
t
h
e
at
te
n
ti
o
n
n
e
t
wo
r
k
w
as
a
d
d
e
d
.
T
h
e
atte
n
t
io
n
m
ec
h
a
n
is
m
en
h
an
ce
d
th
e
F
1
-
s
c
o
r
e
b
y
3
.
8
%
,
c
o
n
f
i
r
m
i
n
g
its
r
o
l
e
i
n
e
m
p
h
asiz
in
g
i
m
p
o
r
ta
n
t
w
o
r
d
s
a
n
d
r
e
d
u
ci
n
g
m
is
c
lass
i
f
i
ca
t
io
n
.
U
n
li
k
e
t
r
a
n
s
f
o
r
m
e
r
m
o
d
els
l
ik
e
Fi
n
B
E
R
T
,
t
h
e
p
r
o
p
o
s
e
d
m
o
d
el
ac
h
ie
v
es
t
o
p
p
e
r
f
o
r
m
a
n
c
e
w
h
i
le
k
ee
p
i
n
g
a
lig
h
t
wei
g
h
t
a
n
d
e
f
f
ic
ie
n
t
d
esi
g
n
.
O
u
r
f
r
a
m
ew
o
r
k
is
th
er
e
f
o
r
e
s
u
ita
b
l
e
e
v
e
n
f
o
r
r
e
al
-
ti
m
e
u
s
e
i
n
h
i
g
h
-
f
r
eq
u
en
c
y
t
r
a
d
i
n
g
e
n
v
ir
o
n
m
en
ts
w
h
e
r
e
l
o
w
l
ate
n
cy
is
cr
u
cia
l.
T
h
e
p
o
o
r
p
e
r
f
o
r
m
a
n
c
e
o
f
Dis
tilB
E
R
T
a
n
d
le
x
i
c
o
n
-
b
as
ed
m
o
d
e
ls
h
ig
h
l
ig
h
t
s
th
e
m
is
m
at
ch
b
e
twe
en
g
en
er
al
-
p
u
r
p
o
s
e
la
n
g
u
ag
e
m
o
d
e
ls
a
n
d
f
i
n
a
n
c
i
al
te
x
t.
T
ab
le
4
.
E
v
alu
atio
n
m
et
r
ics o
f
test
ed
m
o
d
els
M
o
d
e
l
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
s
c
o
r
e
AUC
S
e
n
t
i
w
a
r
e
t
r
a
d
e
r
(
w
i
t
h
a
t
t
e
n
t
i
o
n
)
0
.
9
7
5
3
0
.
9
6
2
6
0
.
9
7
4
6
0
.
9
8
1
6
0
.
9
7
2
8
S
e
n
t
i
w
a
r
e
t
r
a
d
e
r
(
w
i
t
h
o
u
t
a
t
t
e
n
t
i
o
n
0
.
9
4
6
7
0
.
9
5
3
9
0
.
9
4
7
3
0
.
9
6
3
3
0
.
9
5
8
9
LSTM
0
.
9
3
3
3
0
.
9
4
3
9
0
.
9
4
2
3
0
.
9
3
2
1
0
.
9
2
3
6
D
i
st
i
l
B
ER
T
0
.
4
0
1
7
0
.
7
6
1
2
0
.
4
0
1
7
0
.
2
3
5
1
0
.
5
2
7
0
F
a
st
T
e
x
t
0
.
0
3
7
7
0
.
0
3
5
9
0
.
0
3
7
7
0
.
0
3
5
0
0
.
0
4
3
0
V
a
d
e
r
0
.
3
5
1
5
0
.
5
1
3
3
0
.
3
5
1
5
0
.
4
1
6
3
0
.
5
1
2
3
Te
x
t
B
l
o
b
0
.
1
8
4
1
0
.
4
5
7
0
0
.
1
8
4
1
0
.
2
5
1
8
0
.
3
2
4
1
3
.
3
.
Ana
ly
s
is
o
f
c
o
nfusi
o
n
m
a
t
rice
s
As
o
u
tlin
ed
in
o
u
r
ev
alu
atio
n
f
r
am
ewo
r
k
,
Fig
u
r
e
3
p
r
esen
ts
a
p
lo
t
o
f
th
e
c
o
n
f
u
s
io
n
m
atr
ices
f
o
r
Dis
til
B
E
R
T
,
L
STM
,
an
d
b
o
th
v
ar
iatio
n
s
o
f
Sen
tiTr
ad
er
.
W
e
u
s
e
co
n
f
u
s
io
n
m
atr
ices
d
u
e
to
th
eir
ab
ilit
y
to
o
f
f
er
d
etailed
in
s
ig
h
ts
in
to
h
o
w
well
ea
ch
m
o
d
e
l
p
er
f
o
r
m
s
ac
r
o
s
s
s
en
tim
en
t
ca
teg
o
r
ies.
W
e
o
b
s
er
v
e
th
at
th
e
L
STM
b
aselin
e
h
as
h
ig
h
co
n
f
u
s
io
n
b
etwe
en
class
es,
esp
ec
ially
b
etwe
en
p
o
s
itiv
e
an
d
n
e
g
ativ
e
s
en
tim
en
ts
,
m
is
class
if
y
in
g
2
0
i
n
s
tan
ce
s
b
etwe
en
th
ese
two
.
Ho
wev
er
,
th
e
Dis
tilB
E
R
T
co
n
f
u
s
io
n
is
m
o
r
e
p
r
o
n
o
u
n
ce
d
,
p
r
o
v
id
i
n
g
f
u
r
th
er
s
u
p
p
o
r
t
to
o
u
r
ea
r
lier
ass
er
tio
n
th
at
it
s
tr
u
g
g
les
to
ca
p
tu
r
e
th
e
s
u
b
tle
m
e
an
in
g
s
in
f
in
an
cial
lan
g
u
ag
e.
A
s
ig
n
if
ican
t
im
p
r
o
v
em
en
t
is
o
b
s
er
v
e
d
o
n
S
en
tiTr
ad
er
with
o
u
t
atten
tio
n
,
ac
h
iev
in
g
p
e
r
f
ec
t
class
if
icatio
n
f
o
r
th
e
n
eg
ativ
e
an
d
n
e
u
tr
al
class
es,
wh
ile
r
ed
u
cin
g
th
e
p
o
s
itiv
e
-
n
e
g
ativ
e
co
n
f
u
s
io
n
to
ju
s
t
f
o
u
r
in
s
tan
ce
s
.
No
tab
ly
,
Sen
tiTr
a
d
er
with
atten
tio
n
ac
h
iev
es
p
er
f
ec
t
class
if
ic
atio
n
ac
r
o
s
s
all
s
en
tim
en
t
ca
teg
o
r
ies
,
co
m
p
letely
r
em
o
v
in
g
cr
o
s
s
-
cl
ass
er
r
o
r
s
.
T
h
is
h
ig
h
lig
h
ts
h
o
w
well
o
u
r
ad
ap
tiv
e
f
r
am
ewo
r
k
wo
r
k
s
to
id
en
tify
s
en
tim
en
t
-
b
ea
r
in
g
p
h
r
ases
an
d
clea
r
u
p
th
e
co
n
tex
t
is
s
u
es
th
at
o
th
er
m
o
d
els
f
ac
e.
W
e
co
n
cl
u
d
e
th
at
t
h
e
s
tead
y
i
m
p
r
o
v
em
e
n
t
f
r
o
m
L
STM
to
Sen
tiTr
ad
er
with
o
u
t
atten
tio
n
,
an
d
th
en
to
Sen
tiTr
ad
er
with
atten
tio
n
,
s
u
p
p
o
r
ts
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
2
0
2
6
:
416
-
4
2
6
422
o
u
r
d
esig
n
ch
o
ices,
n
o
t
o
n
l
y
r
e
g
ar
d
in
g
im
p
o
r
tan
ce
atte
n
tio
n
m
ec
h
an
is
m
s
b
u
t
r
ea
l
-
t
im
e
ad
ap
ta
b
ilit
y
in
im
p
r
o
v
in
g
s
en
tim
en
t c
lass
if
icatio
n
in
f
in
a
n
cial
tex
t a
n
aly
s
is
.
3
.
4
.
Rea
l
-
t
im
e
perf
o
rma
nce
a
nd
pra
ct
ica
l im
pli
ca
t
io
ns
Ou
r
r
ea
l
-
tim
e
p
e
r
f
o
r
m
an
ce
e
v
alu
atio
n
d
e
m
o
n
s
tr
ates
th
e
p
i
p
elin
e'
s
o
p
er
atio
n
al
ef
f
icien
c
y
with
th
e
en
tire
p
r
o
ce
s
s
f
r
o
m
n
ews in
g
e
s
tio
n
to
s
en
tim
en
t c
lass
if
icatio
n
,
with
a
laten
cy
o
f
less
th
an
2
0
0
ms
,
as sh
o
wn
i
n
Fig
u
r
e
4
.
T
h
ese
r
esu
lts
ar
e
with
in
th
e
s
tr
ict
n
ee
d
s
o
f
h
ig
h
-
f
r
eq
u
en
cy
tr
a
d
in
g
en
v
ir
o
n
m
en
ts
.
W
e
o
b
s
er
v
ed
th
at
th
e
co
n
ce
p
t
d
r
if
t
ca
p
ab
ilit
y
s
u
cc
ess
f
u
lly
id
en
tifie
d
an
d
a
d
ju
s
ted
to
v
o
latilit
y
s
h
if
ts
d
u
r
in
g
b
ac
k
test
in
g
.
Pre
d
ictio
n
d
ec
ay
wa
s
r
ed
u
ce
d
b
y
3
2
% c
o
m
p
a
r
ed
to
a
s
tatic
m
o
d
el
,
s
ee
Fig
u
r
e
4
.
T
h
is
s
tr
en
g
th
is
im
p
o
r
tan
t f
o
r
f
in
tech
d
ep
lo
y
m
en
t
b
ec
au
s
e
it
co
n
f
ir
m
s
th
at
th
e
m
o
d
el
p
er
f
o
r
m
s
r
eliab
ly
ev
en
d
u
r
in
g
m
ar
k
et
cr
is
es.
Fig
u
r
e
4
p
r
esen
ts
a
p
er
f
o
r
m
an
ce
a
n
aly
s
is
u
s
in
g
p
ip
elin
e
laten
cy
d
is
tr
ib
u
tio
n
a
n
d
co
n
ce
p
t d
r
if
t a
d
ap
t
atio
n
m
etr
ics.
T
o
v
er
if
y
o
u
r
m
o
d
el’
s
co
n
ce
p
t
d
r
if
t
d
etec
tio
n
an
d
in
c
r
em
e
n
tal
u
p
d
ate
ca
p
ab
ilit
y
,
we
ev
alu
ated
th
e
p
r
e
an
d
p
o
s
t
C
OVI
D
-
1
9
p
er
i
o
d
.
As
illu
s
tr
ated
in
Fig
u
r
e
5
,
th
e
m
ain
f
in
d
in
g
is
th
at
a
s
t
atic
m
o
d
el
q
u
ick
l
y
b
ec
o
m
es
o
u
td
ate
d
d
u
r
in
g
a
m
ar
k
et
s
h
o
ck
,
wh
er
ea
s
o
u
r
ad
ap
tiv
e
m
o
d
el
h
an
d
led
th
is
ch
allen
g
e
th
r
o
u
g
h
co
n
tin
u
o
u
s
lear
n
i
n
g
.
W
h
ile
th
e
m
o
d
el
d
i
d
n
o
t
p
er
f
o
r
m
s
o
w
ell
in
itially
,
it
p
r
ev
e
n
ted
a
p
o
t
en
tial
4
0
-
6
0
%
d
r
o
p
in
p
er
f
o
r
m
an
ce
an
d
in
s
tead
ac
h
iev
ed
a
co
n
s
is
ten
t
3
-
8%
ad
v
an
tag
e.
T
h
is
h
ig
h
lig
h
ts
th
e
im
p
o
r
tan
t
r
o
le
o
f
ad
ap
tiv
e
f
r
a
m
ewo
r
k
s
in
r
ea
l
-
w
o
r
ld
f
in
a
n
cial
ap
p
licatio
n
s
wh
er
e
v
o
latilit
y
is
h
ig
h
.
Fig
u
r
e
3
.
C
o
n
f
u
s
io
n
m
atr
i
x
p
l
o
ts
Fig
u
r
e
4
.
R
ea
l
-
tim
e
d
ep
l
o
y
m
e
n
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
A
d
a
p
tive
s
en
timen
t a
n
a
lysi
s
fo
r
s
to
ck
ma
r
ke
ts
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ta
len
t Ma
w
ere
)
423
Fig
u
r
e
5
.
Pre
a
n
d
p
o
s
t
C
OVI
D
-
1
9
p
e
r
f
o
r
m
an
ce
Desp
ite
o
u
r
m
o
d
el'
s
r
o
b
u
s
t
p
er
f
o
r
m
an
ce
,
it
h
as
s
o
m
e
lim
itatio
n
s
th
at
n
ee
d
to
b
e
r
e
f
in
ed
.
W
e
ac
k
n
o
wled
g
e
o
u
r
f
r
am
ewo
r
k
'
s
s
o
le
r
elian
ce
o
n
tex
t
u
al
n
e
ws
h
ea
d
lin
es
f
o
r
s
en
tim
en
t
class
if
icatio
n
.
T
h
is
m
ea
n
s
th
at
it is
lim
ited
to
th
e
s
to
r
y
teller
'
s
p
o
in
t o
f
v
iew
a
n
d
n
o
t th
at
o
f
r
ea
l in
v
esto
r
s
.
T
h
e
m
o
d
el’
s
r
elian
ce
o
n
Fas
tTe
x
t
is
y
et
an
o
th
er
lim
itatio
n
th
at
m
ig
h
t
m
ak
e
th
e
m
o
d
el
s
tr
u
g
g
le
with
co
m
p
lex
f
in
an
cial
ter
m
s
an
d
ir
o
n
y
,
w
h
er
e
m
o
r
e
s
o
p
h
is
ticated
m
o
d
els lik
e
Fin
B
E
R
T
co
u
l
d
ea
s
ily
f
lo
u
r
is
h
.
Fu
tu
r
e
wo
r
k
will f
o
cu
s
o
n
:
i)
I
m
p
r
o
v
in
g
t
h
e
s
en
tim
en
t
an
al
y
s
is
ca
p
ab
ilit
y
b
y
u
s
in
g
th
e
w
h
o
le
n
ews
ar
ticle
r
ath
er
th
an
j
u
s
t
m
ak
in
g
u
s
e
o
f
h
ea
d
lin
es.
T
h
e
m
o
d
el
s
h
o
u
l
d
also
em
b
ed
o
th
er
s
o
u
r
ce
s
o
f
n
ews,
s
u
ch
as so
cial
m
ed
ia.
ii)
Ou
r
m
o
d
el
is
n
o
t
b
u
ilt
to
w
o
r
k
in
is
o
latio
n
,
a
n
d
as
s
u
ch
,
f
u
tu
r
e
wo
r
k
will
f
o
cu
s
o
n
i
n
te
g
r
atin
g
m
u
lti
-
m
o
d
al
d
ata,
in
cl
u
d
in
g
tex
tu
a
l
s
en
tim
en
t
with
tec
h
n
ical
i
n
d
icato
r
s
an
d
m
ac
r
o
ec
o
n
o
m
i
c
d
ata
i
n
o
n
e
p
r
ed
ictiv
e
m
o
d
el.
iii)
T
o
v
alid
ate
g
en
er
aliza
b
ilit
y
f
o
r
o
u
r
m
o
d
el,
we
also
s
u
g
g
est
th
at
f
u
tu
r
e
wo
r
k
ex
p
an
d
o
u
r
m
o
d
el
t
o
di
f
f
er
en
t a
s
s
et
class
es b
y
v
alid
atin
g
th
e
f
r
a
m
ewo
r
k
o
n
f
o
r
ex
,
cr
y
p
to
cu
r
r
en
cies,
an
d
co
m
m
o
d
ities
.
4.
CO
NCLU
SI
O
N
T
h
er
e
is
an
u
r
g
en
t
n
ee
d
f
o
r
th
e
estab
lis
h
m
en
t
o
f
ad
ap
tab
le
s
y
s
tem
s
f
o
r
r
ea
l
-
tim
e
f
in
an
cial
s
en
tim
en
t
an
aly
s
is
.
I
n
th
e
cu
r
r
en
t
s
tu
d
y
,
we
in
v
esti
g
ated
h
o
w
i
n
teg
r
atin
g
in
c
r
em
en
tal
lea
r
n
in
g
ap
p
r
o
ac
h
es
with
au
to
m
atic
r
etr
ain
in
g
an
d
h
y
b
r
id
ar
ch
itectu
r
es,
s
u
ch
as
th
e
C
NN
-
B
iL
STM
-
Atten
tio
n
ar
ch
itectu
r
e
with
a
co
n
ce
p
t
d
r
if
t
d
etec
tio
n
m
ec
h
an
is
m
,
ca
n
im
p
r
o
v
e
p
r
ed
ictiv
e
task
s
in
s
en
tim
en
t
an
aly
s
is
.
T
h
e
f
r
a
m
ewo
r
k
a
ch
iev
es
ex
ce
llen
t
ac
c
u
r
ac
y
a
n
d
s
tr
o
n
g
o
p
er
atio
n
al
r
esil
ien
ce
,
ac
co
r
d
in
g
t
o
th
e
em
p
ir
ica
l
r
esu
lts
.
W
e
also
n
o
ted
th
at
it
p
er
f
o
r
m
s
n
o
ticea
b
ly
b
etter
th
a
n
a
n
u
m
b
er
o
f
le
ad
in
g
b
e
n
ch
m
ar
k
s
,
estab
lis
h
in
g
a
n
ew
b
aselin
e
f
o
r
s
en
tim
en
t
-
d
r
iv
en
m
ar
k
et
an
al
y
s
is
to
o
ls
.
T
h
e
f
in
d
in
g
s
also
v
alid
ate
o
u
r
m
o
d
el'
s
u
s
ef
u
ln
ess
an
d
p
r
ac
ticality
f
o
r
f
in
tech
ap
p
licatio
n
s
d
u
e
to
its
lo
w
-
laten
cy
p
er
f
o
r
m
an
ce
an
d
f
lex
ib
ilit
y
.
T
h
e
ap
p
r
o
ac
h
ca
n
b
e
ap
p
lied
to
r
is
k
m
an
ag
em
en
t,
au
to
m
ated
tr
a
d
i
n
g
,
a
n
d
s
en
tim
e
n
t
an
aly
s
is
o
f
in
v
esto
r
s
.
I
n
th
e
f
u
tu
r
e
,
we
h
o
p
e
to
im
p
r
o
v
e
th
e
ar
ch
itectu
r
e
b
y
co
m
b
in
in
g
m
a
cr
o
ec
o
n
o
m
ic
in
d
icato
r
s
,
tech
n
o
lo
g
ical
d
ata,
a
n
d
s
en
tim
en
t
d
ata.
E
x
p
an
d
in
g
its
s
co
p
e
to
in
co
r
p
o
r
ate
m
u
ltimo
d
al
an
d
m
u
ltil
in
g
u
al
f
in
a
n
ci
al
d
ata
co
u
ld
b
e
an
o
th
er
to
p
ic
o
f
in
v
esti
g
atio
n
,
s
tr
en
g
th
en
in
g
t
h
e
lin
k
b
etwe
en
s
o
p
h
is
ticated
NL
P a
n
d
d
y
n
a
m
ic
f
in
an
cial
m
ar
k
ets.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
2
0
2
6
:
416
-
4
2
6
424
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
e
au
th
o
r
s
s
tate
th
at
n
o
f
u
n
d
i
n
g
was r
ec
eiv
ed
f
o
r
t
h
is
p
r
o
je
ct.
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
t
r
ib
u
to
r
R
o
les
T
a
x
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
i
d
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
T
alen
t M
awe
r
e
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Selv
ar
aj
R
ajala
k
s
h
m
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
Ven
u
M
ad
h
a
v
Ku
th
ad
i
✓
✓
✓
✓
✓
✓
✓
✓
Otlh
ap
ile
Din
ak
en
y
an
e
✓
✓
✓
✓
✓
✓
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
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
I
NF
O
RM
E
D
CO
NS
E
N
T
W
e
h
av
e
o
b
tain
ed
in
f
o
r
m
ed
c
o
n
s
en
t f
r
o
m
all
in
d
iv
id
u
als in
c
lu
d
ed
in
t
h
is
s
tu
d
y
.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
e
r
esear
ch
r
elate
d
to
h
u
m
an
u
s
e
h
as
co
m
p
lied
with
al
l
th
e
r
elev
an
t
n
atio
n
al
r
e
g
u
l
atio
n
s
an
d
in
s
titu
tio
n
al
p
o
licies
f
o
llo
win
g
th
e
ten
ets
o
f
th
e
Helsin
k
i
D
ec
lar
atio
n
an
d
h
as
b
ee
n
ap
p
r
o
v
ed
b
y
th
e
au
t
h
o
r
s
'
in
s
titu
tio
n
al
r
ev
iew
b
o
ar
d
.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
th
at
s
u
p
p
o
r
t
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
ailab
l
e
f
r
o
m
th
e
co
r
r
esp
o
n
d
in
g
au
t
h
o
r
,
[
TM
]
,
u
p
o
n
r
ea
s
o
n
ab
le
r
eq
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
G
.
M
.
C
h
a
t
z
i
l
o
i
z
o
s,
D
.
G
u
n
o
p
u
l
o
s,
a
n
d
K
.
K
o
n
st
a
n
t
i
n
o
u
,
“
D
e
e
p
l
e
a
r
n
i
n
g
f
o
r
st
o
c
k
m
a
r
k
e
t
p
r
e
d
i
c
t
i
o
n
u
si
n
g
s
e
n
t
i
m
e
n
t
a
n
d
t
e
c
h
n
i
c
a
l
a
n
a
l
y
s
i
s,
”
S
N
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
5
,
n
o
.
5
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s4
2
9
7
9
-
024
-
02
651
-
5.
[
2
]
S
.
W
u
,
Y
.
L
i
u
,
Z.
Z
o
u
,
a
n
d
T.
H
.
W
e
n
g
,
“
S
_
I
_
LST
M
:
st
o
c
k
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
b
a
se
d
o
n
mu
l
t
i
p
l
e
d
a
t
a
s
o
u
r
c
e
s
a
n
d
se
n
t
i
me
n
t
a
n
a
l
y
si
s
,
”
C
o
n
n
e
c
t
i
o
n
S
c
i
e
n
c
e
,
v
o
l
.
3
4
,
n
o
.
1
,
p
p
.
4
4
–
6
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
0
/
0
9
5
4
0
0
9
1
.
2
0
2
1
.
1
9
4
0
1
0
1
.
[
3
]
K
.
W
a
n
g
,
“
M
u
l
t
i
f
a
c
t
o
r
p
r
e
d
i
c
t
i
o
n
m
o
d
e
l
f
o
r
s
t
o
c
k
mark
e
t
a
n
a
l
y
s
i
s
b
a
se
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s,”
S
c
i
e
n
t
i
f
i
c
Re
p
o
rt
s
,
v
o
l
.
1
5
,
n
o
.
1
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
5
9
8
-
0
2
5
-
8
8
7
3
4
-
6.
[
4
]
N
.
D
a
s,
B
.
S
a
d
h
u
k
h
a
n
,
S
.
S
.
B
h
a
k
t
a
,
a
n
d
S
.
C
h
a
k
r
a
b
a
r
t
i
,
“
I
n
t
e
g
r
a
t
i
n
g
E
E
M
D
a
n
d
e
n
sem
b
l
e
C
N
N
w
i
t
h
X
(
Tw
i
t
t
e
r
)
se
n
t
i
me
n
t
f
o
r
e
n
h
a
n
c
e
d
st
o
c
k
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
s,
”
S
o
c
i
a
l
N
e
t
w
o
rk
A
n
a
l
y
s
i
s
a
n
d
Mi
n
i
n
g
,
v
o
l
.
1
4
,
n
o
.
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
3
2
7
8
-
0
2
3
-
0
1
1
9
0
-
w.
[
5
]
Y
.
H
u
a
n
g
a
n
d
V
.
V
a
k
h
a
r
i
a
,
“
D
e
e
p
l
e
a
r
n
i
n
g
-
b
a
se
d
st
o
c
k
mar
k
e
t
p
r
e
d
i
c
t
i
o
n
a
n
d
i
n
v
e
s
t
me
n
t
m
o
d
e
l
f
o
r
f
i
n
a
n
c
i
a
l
m
a
n
a
g
e
m
e
n
t
,
”
J
o
u
rn
a
l
o
f
O
r
g
a
n
i
za
t
i
o
n
a
l
a
n
d
E
n
d
U
ser
C
o
m
p
u
t
i
n
g
,
v
o
l
.
3
6
,
n
o
.
1
,
2
0
2
4
,
d
o
i
:
1
0
.
4
0
1
8
/
JO
EU
C
.
3
4
0
3
8
3
.
[
6
]
H
.
Le
e
,
J.
H
.
K
i
m,
a
n
d
H
.
S
.
Ju
n
g
,
“
D
e
e
p
-
l
e
a
r
n
i
n
g
-
b
a
se
d
s
t
o
c
k
mark
e
t
p
r
e
d
i
c
t
i
o
n
i
n
c
o
r
p
o
r
a
t
i
n
g
ESG
s
e
n
t
i
m
e
n
t
a
n
d
t
e
c
h
n
i
c
a
l
i
n
d
i
c
a
t
o
r
s,
”
Sc
i
e
n
t
i
f
i
c
Re
p
o
r
t
s
,
v
o
l
.
1
4
,
n
o
.
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
3
8
/
s4
1
5
9
8
-
0
2
4
-
6
1
1
0
6
-
2.
[
7
]
N
.
D
a
s,
B
.
S
a
d
h
u
k
h
a
n
,
R
.
C
h
a
t
t
e
r
j
e
e
,
a
n
d
S
.
C
h
a
k
r
a
b
a
r
t
i
,
“
I
n
t
e
g
r
a
t
i
n
g
s
e
n
t
i
men
t
a
n
a
l
y
si
s w
i
t
h
g
r
a
p
h
n
e
u
r
a
l
n
e
t
w
o
r
k
s f
o
r
e
n
h
a
n
c
e
d
st
o
c
k
p
r
e
d
i
c
t
i
o
n
:
a
c
o
m
p
r
e
h
e
n
si
v
e
su
r
v
e
y
,
”
D
e
c
i
si
o
n
A
n
a
l
y
t
i
c
s
J
o
u
r
n
a
l
,
v
o
l
.
1
0
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
d
a
j
o
u
r
.
2
0
2
4
.
1
0
0
4
1
7
.
[
8
]
T.
T
.
K
h
o
e
i
,
H
.
O
.
S
l
i
m
a
n
e
,
a
n
d
N
.
K
a
a
b
o
u
c
h
,
“
D
e
e
p
l
e
a
r
n
i
n
g
:
s
y
st
e
ma
t
i
c
r
e
v
i
e
w
,
mo
d
e
l
s,
c
h
a
l
l
e
n
g
e
s,
a
n
d
r
e
s
e
a
r
c
h
d
i
r
e
c
t
i
o
n
s,
”
N
e
u
ra
l
C
o
m
p
u
t
i
n
g
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
5
,
n
o
.
3
1
,
p
p
.
2
3
1
0
3
–
2
3
1
2
4
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s0
0
5
2
1
-
0
2
3
-
0
8
9
5
7
-
4.
[
9
]
D
.
M
u
h
a
mm
a
d
,
I
.
A
h
m
e
d
,
K
.
N
a
v
e
e
d
,
a
n
d
M
.
B
e
n
d
e
c
h
a
c
h
e
,
“
A
n
e
x
p
l
a
i
n
a
b
l
e
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
st
o
c
k
m
a
r
k
e
t
t
r
e
n
d
p
r
e
d
i
c
t
i
o
n
,
”
H
e
l
i
y
o
n
,
v
o
l
.
1
0
,
n
o
.
2
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
h
e
l
i
y
o
n
.
2
0
2
4
.
e
4
0
0
9
5
.
[
1
0
]
D
.
A
r
a
c
i
,
“
F
i
n
B
E
R
T
:
f
i
n
a
n
c
i
a
l
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s w
i
t
h
p
r
e
-
t
r
a
i
n
e
d
l
a
n
g
u
a
g
e
mo
d
e
l
s
,
”
2
0
1
9
,
a
rX
i
v
:
1
9
0
8
.
1
0
0
6
3
.
[
1
1
]
E.
F
.
F
a
m
a
,
“
Ef
f
i
c
i
e
n
t
c
a
p
i
t
a
l
mar
k
e
t
s
:
a
r
e
v
i
e
w
o
f
t
h
e
o
r
y
a
n
d
e
mp
i
r
i
c
a
l
w
o
r
k
,
”
T
h
e
J
o
u
r
n
a
l
o
f
F
i
n
a
n
c
e
,
v
o
l
.
2
5
,
n
o
.
2
,
M
a
y
1
9
7
0
,
d
o
i
:
1
0
.
2
3
0
7
/
2
3
2
5
4
8
6
.
[
1
2
]
B.
G
.
M
a
l
k
i
e
l
,
A
ra
n
d
o
m
w
a
l
k
d
o
w
n
Wa
l
l
s
t
re
e
t
.
N
e
w
Y
o
r
k
C
i
t
y
,
U
n
i
t
e
d
S
t
a
t
e
s
:
W
.
W
.
N
o
r
t
o
n
&
C
o
m
p
a
n
y
,
2
0
1
3
.
[
1
3
]
R
.
J.
S
h
i
l
l
e
r
,
“
D
o
st
o
c
k
p
r
i
c
e
s
m
o
v
e
t
o
o
m
u
c
h
t
o
b
e
j
u
st
i
f
i
e
d
b
y
s
u
b
se
q
u
e
n
t
c
h
a
n
g
e
s
i
n
d
i
v
i
d
e
n
d
s
?
,
”
Am
e
r
i
c
a
n
E
c
o
n
o
m
i
c
R
e
v
i
e
w
,
v
o
l
.
7
1
,
n
o
.
3
,
p
p
.
4
2
1
–
3
6
,
1
9
8
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
A
d
a
p
tive
s
en
timen
t a
n
a
lysi
s
fo
r
s
to
ck
ma
r
ke
ts
u
s
in
g
d
ee
p
lea
r
n
in
g
(
Ta
len
t Ma
w
ere
)
425
[
1
4
]
D
.
K
a
h
n
e
ma
n
a
n
d
A
.
Tv
e
r
sk
y
,
“
P
r
o
sp
e
c
t
t
h
e
o
r
y
:
a
n
a
n
a
l
y
s
i
s
o
f
d
e
c
i
si
o
n
u
n
d
e
r
r
i
s
k
,
”
C
h
o
i
c
e
s
,
Va
l
u
e
s,
a
n
d
Fr
a
m
e
s
,
v
o
l
.
4
7
,
n
o
.
2
,
p
p
.
1
7
–
4
3
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
7
/
C
B
O
9
7
8
0
5
1
1
8
0
3
4
7
5
.
0
0
3
.
[
1
5
]
Q
.
X
i
a
o
a
n
d
B
.
I
h
n
a
i
n
i
,
“
S
t
o
c
k
t
r
e
n
d
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
s
e
n
t
i
me
n
t
a
n
a
l
y
s
i
s,
”
P
e
e
rJ
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
9
,
2
0
2
3
,
d
o
i
:
1
0
.
7
7
1
7
/
P
EER
J
-
C
S
.
1
2
9
3
.
[
1
6
]
P
.
D
u
b
e
y
,
P
.
D
u
b
e
y
,
a
n
d
H
.
G
e
h
a
n
i
,
“
En
h
a
n
c
i
n
g
se
n
t
i
me
n
t
a
n
a
l
y
s
i
s
t
h
r
o
u
g
h
d
e
e
p
l
a
y
e
r
i
n
t
e
g
r
a
t
i
o
n
w
i
t
h
l
o
n
g
sh
o
r
t
-
t
e
r
m
mem
o
r
y
n
e
t
w
o
r
k
s,”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
El
e
c
t
ri
c
a
l
a
n
d
C
o
m
p
u
t
e
r
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
1
5
,
n
o
.
1
,
p
p
.
9
4
9
–
9
5
7
,
2
0
2
5
,
do
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
5
i
1
.
p
p
9
4
9
-
9
5
7
.
[
1
7
]
S
.
R
a
n
i
a
n
d
P
.
K
u
mar,
“
D
e
e
p
l
e
a
r
n
i
n
g
b
a
se
d
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
u
s
i
n
g
c
o
n
v
o
l
u
t
i
o
n
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
Ar
a
b
i
a
n
J
o
u
rn
a
l
f
o
r
S
c
i
e
n
c
e
a
n
d
En
g
i
n
e
e
ri
n
g
,
v
o
l
.
4
4
,
n
o
.
4
,
p
p
.
3
3
0
5
–
3
3
1
4
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
3
3
6
9
-
0
1
8
-
3
5
0
0
-
z.
[
1
8
]
N
.
C
.
D
a
n
g
,
M
.
N
.
M
.
G
a
r
c
í
a
,
a
n
d
F
.
D
.
L.
P
r
i
e
t
a
,
“
S
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s b
a
s
e
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
:
a
c
o
m
p
a
r
a
t
i
v
e
st
u
d
y
,
”
E
l
e
c
t
r
o
n
i
c
s
,
v
o
l
.
9
,
n
o
.
3
,
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
9
0
3
0
4
8
3
.
[
1
9
]
I
.
A
l
mal
i
s
,
E
.
K
o
u
l
o
u
m
p
r
i
s
,
a
n
d
I
.
V
l
a
h
a
v
a
s,
“
S
e
c
t
o
r
-
l
e
v
e
l
s
e
n
t
i
me
n
t
a
n
a
l
y
si
s
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
,
”
K
n
o
w
l
e
d
g
e
-
B
a
se
d
S
y
s
t
e
m
s
,
v
o
l
.
2
5
8
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
k
n
o
sy
s
.
2
0
2
2
.
1
0
9
9
5
4
.
[
2
0
]
J.
S
i
n
g
h
,
G
.
S
i
n
g
h
,
a
n
d
R
.
S
i
n
g
h
,
“
O
p
t
i
mi
z
a
t
i
o
n
o
f
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
i
s
u
s
i
n
g
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
c
l
a
ssi
f
i
e
r
s,
”
H
u
m
a
n
-
c
e
n
t
r
i
c
C
o
m
p
u
t
i
n
g
a
n
d
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
s
,
v
o
l
.
7
,
n
o
.
1
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
8
6
/
s
1
3
6
7
3
-
0
1
7
-
0
1
1
6
-
3.
[
2
1
]
J.
D
e
v
l
i
n
,
M
.
W
.
C
h
a
n
g
,
K
.
L
e
e
,
a
n
d
K
.
T
o
u
t
a
n
o
v
a
,
“
B
E
R
T
:
p
r
e
-
t
r
a
i
n
i
n
g
o
f
d
e
e
p
b
i
d
i
r
e
c
t
i
o
n
a
l
t
r
a
n
sf
o
r
me
r
s
f
o
r
l
a
n
g
u
a
g
e
u
n
d
e
r
s
t
a
n
d
i
n
g
,
”
i
n
N
AAC
L
H
L
T
2
0
1
9
-
2
0
1
9
C
o
n
f
e
r
e
n
c
e
o
f
t
h
e
N
o
rt
h
A
m
e
ri
c
a
n
C
h
a
p
t
e
r
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s:
H
u
m
a
n
L
a
n
g
u
a
g
e
T
e
c
h
n
o
l
o
g
i
e
s
-
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
C
o
n
f
e
re
n
c
e
,
2
0
1
9
,
v
o
l
.
1
,
p
p
.
4
1
7
1
–
4
1
8
6
.
[
2
2
]
T.
M
i
k
o
l
o
v
,
I
.
S
u
t
s
k
e
v
e
r
,
K
.
C
h
e
n
,
G
.
C
o
r
r
a
d
o
,
a
n
d
J
.
D
e
a
n
,
“
D
i
s
t
r
i
b
u
t
e
d
r
e
p
r
e
s
e
n
t
a
t
i
o
n
s
o
f
w
o
r
d
s
a
n
d
p
h
r
a
ses
a
n
d
t
h
e
i
r
c
o
m
p
o
si
t
i
o
n
a
l
i
t
y
”
2
0
1
3
,
a
rXi
v
:
1
3
1
0
.
4
5
4
6
.
[
2
3
]
P
.
B
o
j
a
n
o
w
s
k
i
,
E.
G
r
a
v
e
,
A
.
J
o
u
l
i
n
,
a
n
d
T
.
M
i
k
o
l
o
v
,
“
En
r
i
c
h
i
n
g
w
o
r
d
v
e
c
t
o
r
s
w
i
t
h
su
b
w
o
r
d
i
n
f
o
r
m
a
t
i
o
n
,
”
T
r
a
n
s
a
c
t
i
o
n
s
o
f
t
h
e
Asso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
,
v
o
l
.
5
,
p
p
.
1
3
5
–
1
4
6
,
2
0
1
7
,
d
o
i
:
1
0
.
1
1
6
2
/
t
a
c
l
_
a
_
0
0
0
5
1
.
[
2
4
]
P
.
K
o
u
k
a
r
a
s,
C
.
N
o
u
si
,
a
n
d
C
.
T
j
o
r
t
j
i
s,
“
S
t
o
c
k
m
a
r
k
e
t
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
m
i
c
r
o
b
l
o
g
g
i
n
g
se
n
t
i
me
n
t
a
n
a
l
y
s
i
s
a
n
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,”
T
e
l
e
c
o
m
,
v
o
l
.
3
,
n
o
.
2
,
p
p
.
3
5
8
–
3
7
8
,
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
t
e
l
e
c
o
m
3
0
2
0
0
1
9
.
[
2
5
]
A
.
V
a
sw
a
n
i
e
t
a
l
.
,
“
A
t
t
e
n
t
i
o
n
i
s
a
l
l
y
o
u
n
e
e
d
,
”
2
0
1
7
,
a
r
Xi
v
:
1
7
0
6
.
0
3
7
6
2
.
[
2
6
]
C
.
B
a
s
t
i
d
o
n
a
n
d
F
.
J
a
w
a
d
i
,
“
Tr
a
d
e
f
r
a
g
m
e
n
t
a
t
i
o
n
a
n
d
v
o
l
a
t
i
l
i
t
y
-
of
-
v
o
l
a
t
i
l
i
t
y
n
e
t
w
o
r
k
s
,
”
J
o
u
r
n
a
l
o
f
I
n
t
e
r
n
a
t
i
o
n
a
l
Fi
n
a
n
c
i
a
l
Ma
r
k
e
t
s,
I
n
st
i
t
u
t
i
o
n
s
a
n
d
M
o
n
e
y
,
v
o
l
.
9
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
n
t
f
i
n
.
2
0
2
3
.
1
0
1
9
0
8
.
[
2
7
]
X
.
L
u
,
Q
.
Ze
n
g
,
J.
Zh
o
n
g
,
a
n
d
B
.
Z
h
u
,
“
I
n
t
e
r
n
a
t
i
o
n
a
l
s
t
o
c
k
mar
k
e
t
v
o
l
a
t
i
l
i
t
y
:
a
g
l
o
b
a
l
t
a
i
l
r
i
sk
si
g
h
t
,
”
J
o
u
rn
a
l
o
f
I
n
t
e
rn
a
t
i
o
n
a
l
Fi
n
a
n
c
i
a
l
Ma
r
k
e
t
s,
I
n
st
i
t
u
t
i
o
n
s
a
n
d
M
o
n
e
y
,
v
o
l
.
9
1
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
n
t
f
i
n
.
2
0
2
3
.
1
0
1
9
0
4
.
[
2
8
]
A
.
B
i
f
e
t
a
n
d
R
.
G
a
v
a
l
d
à
,
“
Le
a
r
n
i
n
g
f
r
o
m
t
i
me
-
c
h
a
n
g
i
n
g
d
a
t
a
w
i
t
h
a
d
a
p
t
i
v
e
w
i
n
d
o
w
i
n
g
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
7
t
h
S
I
A
M
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
a
t
a
Mi
n
i
n
g
,
2
0
0
7
,
p
p
.
4
4
3
–
4
4
8
,
d
o
i
:
1
0
.
1
1
3
7
/
1
.
9
7
8
1
6
1
1
9
7
2
7
7
1
.
4
2
.
[
2
9
]
C
.
G
u
o
,
G
.
P
l
e
i
ss
,
Y
.
S
u
n
,
a
n
d
K
.
Q
.
W
e
i
n
b
e
r
g
e
r
,
“
O
n
c
a
l
i
b
r
a
t
i
o
n
o
f
mo
d
e
r
n
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
3
4
t
h
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
Ma
c
h
i
n
e
L
e
a
r
n
i
n
g
,
2
0
1
7
.
[
3
0
]
A
.
Za
r
g
h
a
n
i
a
n
d
S
.
A
b
e
d
i
,
“
D
e
s
i
g
n
i
n
g
a
d
a
p
t
i
v
e
a
l
g
o
r
i
t
h
ms
b
a
s
e
d
o
n
r
e
i
n
f
o
r
c
e
m
e
n
t
l
e
a
r
n
i
n
g
f
o
r
d
y
n
a
m
i
c
o
p
t
i
m
i
z
a
t
i
o
n
o
f
s
l
i
d
i
n
g
w
i
n
d
o
w
si
z
e
i
n
mu
l
t
i
-
d
i
me
n
s
i
o
n
a
l
d
a
t
a
st
r
e
a
ms,
”
2
0
2
5
,
a
rX
i
v
:
2
5
0
7
.
0
6
9
0
1
.
[
3
1
]
I
.
F
.
-
B
l
a
n
c
o
,
J.
D
.
C
.
-
Á
v
i
l
a
,
G
.
R
.
-
J
i
m
é
n
e
z
,
R
.
M
.
-
B
u
e
n
o
,
A
.
O
.
-
D
í
a
z
,
a
n
d
Y
.
C
.
-
M
o
t
a
,
“
O
n
l
i
n
e
a
n
d
n
o
n
-
p
a
r
a
m
e
t
r
i
c
d
r
i
f
t
d
e
t
e
c
t
i
o
n
met
h
o
d
s
b
a
s
e
d
o
n
H
o
e
f
f
d
i
n
g
’
s
b
o
u
n
d
s,”
I
EEE
T
r
a
n
sa
c
t
i
o
n
s
o
n
K
n
o
w
l
e
d
g
e
a
n
d
D
a
t
a
E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
2
7
,
n
o
.
3
,
p
p
.
8
1
0
–
8
2
3
,
M
a
r
.
2
0
1
5
,
d
o
i
:
1
0
.
1
1
0
9
/
TK
D
E.
2
0
1
4
.
2
3
4
5
3
8
2
.
[
3
2
]
J.
H
o
w
a
r
d
a
n
d
S
.
R
u
d
e
r
,
“
U
n
i
v
e
r
sa
l
l
a
n
g
u
a
g
e
m
o
d
e
l
f
i
n
e
-
t
u
n
i
n
g
f
o
r
t
e
x
t
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
i
n
AC
L
2
0
1
8
-
5
6
t
h
A
n
n
u
a
l
M
e
e
t
i
n
g
o
f
t
h
e
Asso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
,
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
C
o
n
f
e
ren
c
e
(
L
o
n
g
Pa
p
e
rs)
,
2
0
1
8
,
v
o
l
.
1
,
p
p
.
3
2
8
–
3
3
9
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
p
1
8
-
1
0
3
1
.
[
3
3
]
X
.
Li
u
,
P
.
H
e
,
W
.
C
h
e
n
,
a
n
d
J
.
G
a
o
,
“
M
u
l
t
i
-
t
a
s
k
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
u
n
d
e
r
st
a
n
d
i
n
g
,
”
i
n
AC
L
2
0
1
9
-
5
7
t
h
An
n
u
a
l
Me
e
t
i
n
g
o
f
t
h
e
Asso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s,
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
C
o
n
f
e
r
e
n
c
e
,
2
0
2
0
,
p
p
.
4
4
8
7
–
4
4
9
6
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
p
1
9
-
1
4
4
1
.
[
3
4
]
J.
G
a
m
a
,
I
.
Zl
i
o
b
a
i
t
e
,
A
.
B
i
f
e
t
,
M
.
P
e
c
h
e
n
i
z
k
i
y
,
a
n
d
A
.
B
o
u
c
h
a
c
h
i
a
,
“
A
su
r
v
e
y
o
n
c
o
n
c
e
p
t
d
r
i
f
t
a
d
a
p
t
a
t
i
o
n
,
”
AC
M
C
o
m
p
u
t
i
n
g
S
u
r
v
e
y
s
,
v
o
l
.
4
6
,
n
o
.
4
,
2
0
1
4
,
d
o
i
:
1
0
.
1
1
4
5
/
2
5
2
3
8
1
3
.
[
3
5
]
X
.
C
h
e
n
,
H
.
X
i
e
,
Z.
L
i
,
H
.
Zh
a
n
g
,
X
.
Ta
o
,
a
n
d
F
.
L
.
W
a
n
g
,
“
S
e
n
t
i
me
n
t
a
n
a
l
y
si
s
f
o
r
st
o
c
k
mar
k
e
t
r
e
s
e
a
r
c
h
:
a
b
i
b
l
i
o
m
e
t
r
i
c
st
u
d
y
,
”
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
Pro
c
e
ss
i
n
g
J
o
u
r
n
a
l
,
v
o
l
.
1
0
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
l
p
.
2
0
2
5
.
1
0
0
1
2
5
.
[
3
6
]
S
.
M
e
h
t
a
b
a
n
d
J
.
S
e
n
,
“
A
r
o
b
u
st
p
r
e
d
i
c
t
i
v
e
m
o
d
e
l
f
o
r
st
o
c
k
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ss
i
n
g
,
”
S
S
RN
El
e
c
t
ro
n
i
c
J
o
u
r
n
a
l
.
2
0
2
0
,
d
o
i
:
1
0
.
2
1
3
9
/
ssr
n
.
3
5
0
2
6
2
4
.
[
3
7
]
T.
S
i
d
o
g
i
,
R
.
M
b
u
v
h
a
,
a
n
d
T.
M
a
r
w
a
l
a
,
“
S
t
o
c
k
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
s
e
n
t
i
me
n
t
a
n
a
l
y
s
i
s,
”
i
n
C
o
n
f
e
re
n
c
e
Pr
o
c
e
e
d
i
n
g
s
-
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
S
y
s
t
e
m
s,
Ma
n
a
n
d
C
y
b
e
rn
e
t
i
c
s
,
2
0
2
1
,
p
p
.
4
6
–
51
,
d
o
i
:
1
0
.
1
1
0
9
/
S
M
C
5
2
4
2
3
.
2
0
2
1
.
9
6
5
9
2
8
3
.
[
3
8
]
Q
.
Tu
l
e
t
a
l
.
,
“
S
e
n
t
i
me
n
t
a
n
a
l
y
si
s 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:
a
r
e
v
i
e
w
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
,
n
o
.
6
,
2
0
1
7
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
1
7
.
0
8
0
6
5
7
.
[
3
9
]
W
.
S
o
u
m
a
,
I
.
V
o
d
e
n
s
k
a
,
a
n
d
H
.
A
o
y
a
ma
,
“
En
h
a
n
c
e
d
n
e
w
s
s
e
n
t
i
me
n
t
a
n
a
l
y
si
s
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
m
e
t
h
o
d
s,”
J
o
u
rn
a
l
o
f
C
o
m
p
u
t
a
t
i
o
n
a
l
S
o
c
i
a
l
S
c
i
e
n
c
e
,
v
o
l
.
2
,
n
o
.
1
,
p
p
.
3
3
–
4
6
,
F
e
b
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s4
2
0
0
1
-
019
-
0
0
0
3
5
-
x.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Ta
lent
M
a
we
r
e
is
a
lec
tu
re
r
in
th
e
De
p
a
rtme
n
t
o
f
M
a
t
h
e
m
a
ti
c
s
a
n
d
C
o
m
p
u
ter
S
c
ien
c
e
a
t
G
re
a
t
Zi
m
b
a
b
we
Un
iv
e
rsity
a
n
d
a
P
h
.
D
.
c
a
n
d
i
d
a
te
i
n
C
o
m
p
u
ti
n
g
a
n
d
In
f
o
rm
a
ti
c
s
a
t
th
e
Bo
tswa
n
a
In
ter
n
a
ti
o
n
a
l
Un
iv
e
rsity
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
l
o
g
y
(
BIUST
).
He
h
o
l
d
s
a
M
a
ste
r
o
f
M
a
n
a
g
e
m
e
n
t
I
n
fo
rm
a
ti
o
n
S
y
st
e
m
s
a
n
d
a
B
.
S
c
.
in
In
f
o
rm
a
ti
o
n
S
y
ste
m
s,
b
o
th
fro
m
M
id
lan
d
s
S
tate
Un
iv
e
rsity
,
Zi
m
b
a
b
we
.
Hi
s
re
se
a
r
c
h
in
tere
sts
in
c
lu
d
e
a
rti
ficia
l
in
telli
g
e
n
c
e
,
m
a
c
h
in
e
lea
rn
in
g
,
c
o
m
p
u
ter
se
c
u
rit
y
,
a
n
d
I
CT
fo
r
d
e
v
e
l
o
p
m
e
n
t
(ICT
4
D).
H
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
tale
n
tma
we
re
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.