T
E
L
K
O
M
N
I
K
A
T
elec
o
m
m
un
ica
t
io
n,
Co
m
pu
t
ing
,
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
18
,
No
.
5
,
Octo
b
er
2
0
2
0
,
p
p
.
2
5
2
5
~
2
5
3
2
I
SS
N:
1
6
9
3
-
6
9
3
0
,
ac
cr
ed
ited
First Gr
ad
e
b
y
Kem
en
r
is
tek
d
i
k
ti,
Dec
r
ee
No
: 2
1
/E/KPT
/2
0
1
8
DOI
: 1
0
.
1
2
9
2
8
/TE
L
KOM
NI
K
A.
v
1
8
i5
.
1
4
8
8
7
2525
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
Prec
ipi
tatio
n pre
diction usi
ng
r
ecu
rrent n
eura
l n
etw
o
rks
a
nd lon
g
sho
rt
-
t
e
rm memo
ry
M
is
hk
a
Aldity
a
P
ria
t
na
1
,
E
s
m
er
a
lda
C.
Dj
a
m
a
l
2
1
De
p
a
rt
m
e
n
t
o
f
M
e
teo
r
o
l
o
g
y
,
I
n
st
it
u
t
Te
k
n
o
lo
g
i
Ba
n
d
u
n
g
,
In
d
o
n
e
si
a
2
De
p
a
rt
m
e
n
t
o
f
In
fo
rm
a
ti
c
s
,
Un
i
v
e
rsitas
Je
n
d
e
ra
l
Ac
h
m
a
d
Ya
n
i,
In
d
o
n
e
sia
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
9
,
2019
R
ev
is
ed
Ap
r
12
,
2
0
2
0
Acc
ep
ted
May
1
,
2
0
2
0
P
re
d
ictio
n
o
f
m
e
teo
r
o
lo
g
ica
l
v
a
ria
b
les
su
c
h
a
s
p
re
c
ip
it
a
ti
o
n
,
tem
p
e
ra
tu
re
,
wi
n
d
sp
e
e
d
,
a
n
d
so
lar
ra
d
iatio
n
is
b
e
n
e
ficia
l
fo
r
h
u
m
a
n
li
fe
.
T
h
e
v
a
riab
le
o
b
se
rv
a
ti
o
n
s
d
a
ta
is
a
v
a
il
a
b
le
fro
m
ti
m
e
to
ti
m
e
fo
r
m
o
re
th
a
n
th
i
rty
y
e
a
rs,
sc
a
tt
e
re
d
e
a
c
h
o
b
se
r
v
a
ti
o
n
sta
ti
o
n
m
a
k
e
s
th
e
o
p
p
o
rtu
n
it
y
to
m
a
p
p
a
tt
e
rn
s
i
n
to
p
re
d
ictio
n
s.
Ho
we
v
e
r,
th
e
c
o
m
p
le
x
it
y
o
f
we
a
th
e
r
v
a
riab
les
is
v
e
ry
h
ig
h
,
o
n
e
o
f
wh
ich
is
i
n
flu
e
n
c
e
d
b
y
De
c
a
d
a
l
p
h
e
n
o
m
e
n
a
su
c
h
a
s
El
-
Ni
n
o
S
o
u
th
e
rn
Os
c
il
latio
n
a
n
d
IOD
.
Wea
t
h
e
r
p
r
e
d
ictio
n
s
c
a
n
b
e
re
v
iew
e
d
f
o
r
t
h
e
d
u
ra
ti
o
n
,
p
re
d
ictio
n
v
a
riab
les
,
a
n
d
o
b
se
rv
a
ti
o
n
sta
ti
o
n
s.
Th
is
re
se
a
rc
h
p
r
o
p
o
se
d
p
re
c
ip
it
a
ti
o
n
p
re
d
ictio
n
u
sin
g
re
c
u
rre
n
t
n
e
u
ra
l
n
e
two
r
k
s
a
n
d
lo
n
g
sh
o
rt
-
term
m
e
m
o
ry
.
Ex
p
e
rime
n
ts
we
re
c
a
rried
o
u
t
u
si
n
g
th
e
p
re
d
ictio
n
d
u
ra
t
io
n
fa
c
to
r
,
th
e
p
e
ri
o
d
a
s
a
fe
a
tu
re
a
n
d
t
h
e
a
m
o
u
n
t
o
f
d
a
ta
se
t
u
se
d
,
a
n
d
th
e
o
p
ti
m
iza
ti
o
n
m
o
d
e
l.
T
h
e
re
su
lt
s
sh
o
we
d
t
h
a
t
th
e
ti
m
e
-
lap
se
a
s
a
sh
o
rter
fe
a
tu
re
g
iv
e
s
g
o
o
d
a
c
c
u
ra
c
y
.
Also
,
th
e
d
u
ra
ti
o
n
o
f
we
e
k
ly
p
re
d
icti
o
n
s
p
r
o
v
i
d
e
s
m
o
r
e
a
c
c
u
ra
c
y
th
a
n
m
o
n
th
ly
,
wh
ich
is 8
5
.
7
1
%
c
o
m
p
a
re
d
to
8
3
.
3
3
%
o
f
th
e
v
a
li
d
a
t
io
n
d
a
ta.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
L
o
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
Me
teo
r
o
lo
g
y
d
ata
Pre
ci
p
itatio
n
p
r
ed
ictio
n
R
ec
u
r
r
en
t n
eu
r
al
n
etwo
r
k
s
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
:
E
s
m
er
ald
a
C
o
n
tess
a
Djam
al,
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics,
Un
iv
er
s
itas
J
en
d
er
al
Ach
m
ad
Yan
i,
T
er
u
s
an
J
en
d
er
al
Su
d
ir
m
an
C
i
m
ah
i
St.
,
I
n
d
o
n
esia.
E
m
ail:
esm
er
ald
a.
co
n
tess
a@
le
ctu
r
e.
u
n
jan
i.a
c
.
id
1.
I
NT
RO
D
UCT
I
O
N
I
n
d
o
n
esia
is
b
etwe
en
th
e
I
n
d
ia
n
Oce
an
an
d
th
e
Pacif
ic
Oce
a
n
an
d
two
co
n
tin
e
n
ts
,
esp
ec
ially
Asi
a
an
d
Au
s
tr
alia,
s
o
th
e
clim
ate
o
f
ten
ch
an
g
es
a
n
d
is
in
f
lu
en
ce
d
b
y
m
an
y
f
a
cto
r
s
.
Oth
er
p
h
e
n
o
m
e
n
a,
s
u
ch
as
E
l
-
Nin
o
So
u
th
er
n
Oscill
atio
n
(
E
NSO)
,
ad
d
to
th
e
co
m
p
lex
i
ty
o
f
t
h
e
en
v
ir
o
n
m
e
n
t
[
1
]
.
Pre
cip
it
atio
n
is
a
p
ar
t
o
f
th
e
clim
ate
th
at
h
as
v
er
y
co
m
p
licated
.
I
ts
ch
ar
ac
ter
is
tic
in
ev
er
y
ar
ea
is
in
d
ee
d
d
if
f
e
r
en
t
f
r
o
m
o
n
e
an
o
th
e
r
.
I
t
is
af
f
ec
ted
b
y
m
a
n
y
f
ac
to
r
s
lik
e
g
eo
g
r
a
p
h
ic,
to
p
o
g
r
a
p
h
ic,
an
d
m
o
n
o
g
r
ap
h
ic.
I
s
lan
d
s
tr
u
ctu
r
e
an
d
o
r
ien
tatio
n
h
av
e
n
o
t
b
ee
n
ca
lcu
l
ated
.
As
a
r
esu
lt,
th
e
p
r
ec
ip
itatio
n
d
is
tr
ib
u
tio
n
p
atter
n
s
ar
e
n
o
t
p
r
e
v
alen
t
in
ea
ch
ar
ea
with
o
th
er
s
in
a
b
r
o
ad
s
co
p
e.
B
esid
es
th
at,
E
NSO
in
f
lu
en
ce
s
alm
o
s
t
7
0
%
o
f
th
e
ch
an
g
e
o
f
p
r
ec
ip
itatio
n
in
I
n
d
o
n
esia
in
th
e
Ma
r
itime
C
o
n
tin
en
t
[
2
]
.
I
n
m
ea
n
wh
i
le,
o
th
e
r
p
a
r
ts
o
f
th
e
wo
r
ld
m
ar
k
ed
b
y
th
e
d
is
p
lace
m
en
t
o
f
"wa
r
m
p
o
o
l"
a
n
d
clo
u
d
f
o
r
m
atio
n
o
r
d
in
ar
y
o
cc
u
r
r
ed
in
I
n
d
o
n
esia
s
ea
g
o
ea
s
ter
ly
to
t
h
e
m
id
d
le
o
f
th
e
Pacif
ic
Oce
an
.
T
h
is
p
h
en
o
m
en
o
n
ca
n
d
eg
r
ad
e
t
h
e
p
r
ec
i
p
itatio
n
in
s
e
v
er
al
p
ar
ts
in
th
e
Pacif
ic
ar
e
a
s
o
th
at
it
r
u
n
s
in
to
d
r
o
u
g
h
t
[
3
]
.
W
ith
its
d
y
n
am
is
m
,
p
r
ec
ip
itatio
n
co
u
ld
n
o
t
b
e
d
is
co
v
er
ed
p
r
ec
is
ely
.
Pre
cip
itatio
n
is
an
ess
en
tial
p
h
en
o
m
e
n
o
n
i
n
th
e
clim
ate
s
y
s
tem
,
w
h
ich
is
ch
ao
tic
a
n
d
g
r
e
atly
af
f
ec
t
s
all
asp
ec
ts
o
f
li
fe
,
s
u
ch
as
t
h
e
a
v
ailab
ilit
y
o
f
wate
r
r
eso
u
r
ce
s
,
u
r
b
an
p
lan
n
i
n
g
,
a
g
r
icu
ltu
r
e
,
an
d
f
in
an
cial
s
ec
u
r
ity
.
Ov
er
th
e
p
ast
f
ew
y
ea
r
s
,
r
esea
r
ch
o
n
p
r
ec
ip
itatio
n
p
r
ed
ictio
n
is
o
v
er
g
r
o
win
g
.
M
an
y
m
o
d
els
h
av
e
b
ee
n
co
u
l
d
im
p
r
o
v
e
th
e
ac
c
u
r
ac
y
o
f
p
r
ec
ip
itatio
n
p
r
e
d
ictio
n
[
4
]
.
Me
an
wh
ile,
th
e
life
cy
cle
th
at
f
o
r
m
s
s
p
ec
if
ic
p
atter
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
18
,
No
.
5
,
Octo
b
e
r
2
0
2
0
:
2
5
2
5
-
2532
2526
d
ep
en
d
s
o
n
m
o
n
e
y
an
d
tim
e
[
5
]
.
I
n
d
ee
d
,
d
aily
r
ain
f
all
is
r
an
d
o
m
.
B
u
t
in
th
e
lo
n
g
r
u
n
,
h
a
v
e
a
p
ar
ticu
lar
m
o
d
el
th
at
ca
n
b
e
id
en
tifie
d
.
I
t'
s
ju
s
t
also
p
ay
in
g
atten
tio
n
to
m
an
y
f
ac
to
r
s
th
at
ar
e
u
n
ce
r
tain
,
s
u
c
h
as
clim
ate
zo
n
es,
s
u
n
s
p
o
ts
,
tid
es,
atm
o
s
p
h
er
ic
ci
r
cu
latio
n
,
an
d
o
th
e
r
h
u
m
an
ac
t
iv
ity
f
ac
to
r
s
[6
,
7]
.
C
lim
ate
p
ar
am
eter
d
ata
p
r
o
v
id
ed
b
y
th
e
Me
teo
r
o
lo
g
y
,
C
lim
ato
lo
g
y
,
an
d
Geo
p
h
y
s
ics
Ag
en
c
y
(
B
MK
G)
d
aily
,
ex
ce
p
t
f
o
r
p
r
ec
ip
itatio
n
p
r
o
d
u
ce
d
ev
er
y
h
o
u
r
o
f
ea
ch
s
tatio
n
.
T
h
e
u
s
e
o
f
tim
e
a
n
d
p
r
ed
ictio
n
lo
ca
tio
n
a
r
ea
,
clo
s
ely
r
elate
d
to
th
e
u
s
e
o
f
u
s
ag
e.
B
ased
o
n
th
e
ter
r
ito
r
y
o
f
u
s
e,
s
o
m
e
s
tu
d
ies u
s
e
a
g
lo
b
al
ar
ea
[
8
]
w
h
ile
o
th
er
s
tu
d
ies
ar
e
p
r
ed
ictiv
e
o
f
s
p
ec
if
ic
r
eg
io
n
s
[
9
]
.
Me
an
wh
ile,
clim
ate
p
r
ed
ictio
n
h
as
a
ti
m
e
f
r
am
e
,
s
u
ch
as
d
aily
[
8
]
o
f
s
o
lar
f
o
r
ec
asti
n
g
,
hour
s
o
f
win
d
s
p
ee
d
[
9
]
,
h
o
u
r
s
o
f
tem
p
er
atu
r
e
[
1
0
]
,
m
u
lt
i
-
s
tep
tim
e
o
f
win
d
s
p
ee
d
[
1
1
]
,
ex
tr
em
e
clim
ate
e
v
er
y
d
ay
[
5
]
.
Oth
e
r
r
esear
c
h
esti
m
a
ted
r
ain
f
all
in
a
s
h
o
r
t
t
im
e
o
r
h
o
u
r
s
[
1
2
]
,
d
ay
s
[
1
3
]
,
wee
k
s
[
1
4
]
.
So
m
e
r
esear
ch
p
r
ef
e
r
s
p
r
ec
ip
itatio
n
p
r
ed
ictio
n
in
p
er
io
d
s
m
o
n
th
l
y
with
E
NSO
[
1
5
]
a
n
d
y
ea
r
ly
[
1
6
]
to
h
el
p
in
p
r
o
p
e
r
ag
r
icu
ltu
r
al
p
lan
n
i
n
g
.
T
h
is
s
t
u
d
y
co
m
p
a
r
e
d
th
e
esti
m
ated
wee
k
ly
an
d
m
o
n
th
ly
p
r
ec
ip
itatio
n
m
o
d
els.
W
ea
th
er
f
o
r
ec
asti
n
g
is
a
n
ew
r
esear
c
h
p
r
o
b
lem
b
ec
au
s
e
its
ap
p
licatio
n
is
ex
ten
s
iv
e
in
m
an
y
s
ec
to
r
s
,
s
u
ch
as
ag
r
icu
ltu
r
e
to
f
lig
h
t
n
av
ig
atio
n
.
T
h
e
ch
allen
g
e
o
f
cl
im
ate
p
r
ed
ictio
n
is
to
ch
o
o
s
e
th
e
r
ig
h
t
v
ar
ia
b
les
an
d
d
ata
s
ets
an
d
t
o
ch
o
o
s
e
r
ep
r
esen
tativ
e
m
o
d
els
to
b
e
ab
le
to
ex
p
lo
r
e
h
id
d
en
s
tr
u
ctu
r
a
l
p
atter
n
s
in
a
lar
g
e
d
ataset
[
1
5
]
.
Un
f
o
r
tu
n
ately
,
it
is
n
o
t
a
c
o
n
v
e
n
ien
t
ca
s
e.
Pre
cip
itatio
n
is
a
v
er
y
co
m
p
licated
ev
en
t
b
ec
au
s
e
it
h
ap
p
en
s
r
an
d
o
m
ly
a
n
d
d
ep
e
n
d
s
o
n
n
u
m
er
o
u
s
f
ac
to
r
s
lik
e
tem
p
er
atu
r
e
,
h
u
m
id
ity
,
win
d
s
p
ee
d
,
an
d
clo
u
d
p
r
ess
u
r
e.
B
esid
es th
at,
th
e
d
ep
en
d
en
t v
ar
ia
b
les wh
ich
ar
e
p
o
s
s
ib
le
to
af
f
ec
t p
r
ec
ip
itatio
n
ar
e
n
o
t c
o
n
s
tan
t e
v
en
it
ca
n
n
o
t
b
e
s
u
r
e
h
o
w
m
an
y
f
ac
to
r
s
m
ay
im
p
ac
t
th
e
r
ain
f
all
.
I
t
m
a
k
es
th
e
i
n
p
u
t
p
ar
am
eter
s
to
th
e
m
o
d
el
m
a
y
n
o
t
ad
eq
u
ate
to
p
r
ed
ict
p
r
ec
i
p
itatio
n
p
r
ec
is
ely
[
1
7
]
.
C
lim
ate
f
o
r
ec
asti
n
g
m
ad
e
atten
tio
n
to
m
an
y
r
esear
ch
er
s
of
v
ar
io
u
s
b
ac
k
g
r
o
u
n
d
s
d
u
e
to
its
ef
f
ec
t
o
n
g
lo
b
al
h
u
m
an
life
.
T
h
e
s
u
p
p
o
r
t
o
f
c
o
m
p
u
ter
tech
n
o
lo
g
y
an
d
ac
ce
s
s
ib
ilit
y
to
o
b
tain
b
ig
d
ata
o
f
wea
th
er
o
b
s
er
v
atio
n
r
ec
en
t
ly
m
ad
e
m
an
y
r
esear
ch
er
s
ar
e
en
co
u
r
a
g
ed
to
lear
n
m
o
r
e
a
b
o
u
t
th
e
p
atter
n
in
th
e
lar
g
e
d
ataset
o
f
wea
th
er
p
r
e
d
ictio
n
.
I
t
ca
n
p
r
e
d
ict
wea
th
e
r
f
o
r
ec
asti
n
g
u
s
in
g
m
ac
h
in
e
lear
n
in
g
.
T
h
e
m
et
h
o
d
m
ak
es
p
o
s
s
ib
le
lear
n
i
n
g
t
h
e
p
atter
n
o
f
p
r
ec
ip
itatio
n
with
o
t
h
er
v
a
r
iab
les
in
tim
e
s
er
ies
b
ef
o
r
e.
R
eg
r
ess
io
n
p
r
o
b
lem
s
p
r
o
v
id
e
s
o
m
e
ch
allen
g
in
g
r
esear
ch
in
th
e
f
ield
o
f
m
ac
h
in
e
lear
n
in
g
,
in
clu
d
in
g
wea
th
e
r
d
ata.
R
ain
f
a
ll
is
a
p
r
im
e
ex
am
p
le
b
ec
au
s
e
i
t
s
h
o
ws
u
n
iq
u
e
c
h
ar
ac
ter
is
tics
with
h
ig
h
v
o
latilit
y
an
d
ch
a
o
tic
p
atter
n
s
.
T
h
e
r
ef
o
r
e
th
e
m
ac
h
in
e
lea
r
n
in
g
m
eth
o
d
ca
n
o
u
tp
er
f
o
r
m
o
th
er
m
et
h
o
d
s
[4
]
.
Pre
d
ictio
n
v
ar
iab
les
o
f
clim
ate
wh
ich
m
ay
n
o
t
b
e
clea
r
ly
u
n
d
er
s
to
o
d
,
tr
a
d
itio
n
al
lin
ea
r
f
o
r
ec
asti
n
g
tech
n
iq
u
es
ar
e
ill
-
eq
u
ip
p
e
d
to
h
an
d
le,
o
f
ten
p
r
o
d
u
cin
g
u
n
s
atis
f
ac
to
r
y
r
esu
lts
.
Pre
v
io
u
s
r
esea
r
ch
u
s
in
g
s
tatis
tical
b
ias
co
r
r
ec
tio
n
o
n
th
e
o
u
tp
u
t
o
f
th
e
d
aily
clim
ate
m
o
d
el
in
E
u
r
o
p
e
[
1
8
]
im
p
r
o
v
ed
to
s
ee
t
h
e
r
elatio
n
b
etwe
en
E
C
MW
F a
n
d
th
e
Me
teo
r
o
lo
g
y
,
C
lim
ato
lo
g
y
,
an
d
Geo
p
h
y
s
ics Ag
en
cy
(
B
MK
G)
[
1
9
]
.
T
h
e
r
esear
ch
r
esu
lted
in
th
e
v
alu
e
o
f
th
e
t
r
an
s
f
er
f
u
n
ctio
n
f
o
r
m
e
d
f
r
o
m
th
e
b
ias
co
r
r
ec
tio
n
p
r
o
ce
s
s
ca
n
b
e
u
s
ed
to
im
p
r
o
v
e
th
e
d
is
tr
ib
u
tio
n
o
f
th
e
2
0
1
6
r
ai
n
f
all
p
r
e
d
ictio
n
o
n
th
e
is
lan
d
o
f
B
ali,
to
o
b
tain
a
b
etter
p
r
ed
ic
tio
n
.
I
n
m
ea
n
wh
ile,
s
o
m
e
r
esear
ch
in
cr
ea
s
in
g
ly
r
e
s
o
r
ts
to
tech
n
iq
u
es
th
at
ar
e
h
eu
r
is
tic
an
d
n
o
n
-
lin
ea
r
.
Su
ch
m
eth
o
d
s
u
s
e
n
eu
r
a
l
n
etwo
r
k
m
o
d
els
[
2
0
]
with
m
ac
h
in
e
-
lear
n
in
g
,
r
eg
r
ess
io
n
,
an
d
clu
s
ter
in
g
.
Oth
er
r
esear
ch
u
s
ed
d
y
n
am
ic
r
eg
io
n
al
co
m
b
in
ed
s
h
o
r
t
-
ter
m
r
ain
f
all
f
o
r
ec
asti
n
g
ap
p
r
o
ac
h
(
DR
C
F)
to
im
p
r
o
v
e
m
u
lt
ilay
er
p
e
r
ce
p
tr
o
n
a
n
d
PC
A.
T
h
e
s
tu
d
y
g
av
e
ac
cu
r
ac
y
7
5
-
9
2
% b
u
t d
ep
e
n
de
d
o
n
t
h
e
n
u
m
b
er
o
f
ML
P
[
2
1
]
.
W
ea
th
er
d
ata
p
r
ed
ictio
n
h
as
its
ch
ar
ac
ter
is
tics
,
wh
ich
d
ep
e
n
d
o
n
th
e
v
ar
iab
ilit
y
o
f
th
ese
v
ar
iab
les.
R
ain
f
all
is
v
er
y
v
ar
iab
le
co
m
p
ar
ed
to
s
o
lar
r
ad
iatio
n
,
win
d
s
p
ee
d
,
an
d
tem
p
er
atu
r
e.
Of
c
o
u
r
s
e,
th
is
co
n
d
itio
n
h
as
an
im
p
ac
t
o
n
p
r
e
d
ictio
n
a
cc
u
r
ac
y
,
as
p
r
ev
io
u
s
s
tu
d
ies
p
r
ed
ictin
g
wea
th
e
r
a
n
d
tem
p
er
atu
r
e
p
r
o
v
id
e
b
etter
ac
cu
r
ac
y
th
a
n
r
ea
l
o
n
es
s
o
t
h
at
th
ey
ca
n
b
e
u
s
ed
in
r
ea
l
-
tim
e
[
1
6
]
.
Oth
er
r
esear
ch
d
ev
elo
p
ed
t
o
p
r
ed
ict
tem
p
er
atu
r
e
an
d
h
u
m
id
ity
[
2
2
]
.
Dee
p
lear
n
in
g
is
n
ew
co
m
p
u
tin
g
in
d
ata
m
in
in
g
a
n
d
m
ac
h
in
e
lear
n
in
g
[
6
]
.
A
n
eu
r
al
n
etwo
r
k
with
d
ee
p
ar
ch
itectu
r
es
h
as
b
ec
o
m
e
a
k
in
d
o
f
p
o
wer
f
u
l
to
o
l
t
o
r
etr
iev
e
th
e
h
ig
h
-
lev
el
ab
s
tr
ac
t
f
ea
tu
r
es
o
f
b
ig
d
ata.
So
m
e
m
eth
o
d
s
th
at
ar
e
o
f
ten
u
s
ed
i
n
d
ee
p
lear
n
i
n
g
ar
e
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
,
wh
ich
co
n
v
o
l
u
tes
with
a
f
ix
e
d
s
ize
k
er
n
el.
W
ea
th
er
d
ata
ca
n
b
e
v
iewe
d
as
im
ag
e
r
y
,
s
o
it
ca
n
b
e
r
eso
lv
ed
with
C
NN
[
2
3
]
.
Ho
wev
er
,
f
o
r
a
lim
ited
im
ag
e
s
eg
m
en
t,
it
in
d
ee
d
will
co
llid
e
with
m
em
o
r
y
l
im
itatio
n
s
,
s
o
o
th
er
m
eth
o
d
s
ar
e
n
ee
d
ed
.
On
e
wa
y
th
at
ca
n
b
e
u
s
ed
is
L
STM
f
o
r
r
ai
n
f
all
p
r
e
d
ictio
n
[
2
4
]
.
Usi
n
g
C
NN
ca
n
b
e
m
o
d
if
ied
in
o
n
e
d
im
en
s
io
n
,
f
o
r
ex
am
p
le,
f
o
r
s
u
n
lig
h
t p
r
ed
ict
io
n
[
2
5
]
,
an
d
p
r
ed
ict
p
r
ec
i
p
itatio
n
[
2
6
]
.
Me
an
wh
ile,
f
o
r
tim
e
s
er
ies
o
f
t
en
u
s
e
R
NN.
T
h
e
u
n
iq
u
e
n
ess
o
f
th
e
R
NNs
is
th
e
f
ee
d
b
ac
k
c
o
n
n
ec
tio
n
,
wh
ich
co
n
v
ey
s
in
ter
f
er
e
n
ce
in
f
o
r
m
atio
n
at
t
h
e
p
r
e
v
io
u
s
in
p
u
t
th
at
will
b
e
ac
co
m
m
o
d
ated
to
t
h
e
f
o
llo
win
g
f
ac
ts
.
T
h
e
R
NN
ca
n
s
tu
d
y
s
eq
u
en
tia
l
o
r
tim
e
-
v
ar
y
in
g
p
atter
n
s
s
o
th
at
it
is
to
o
ls
in
m
o
d
elin
g
in
tr
icate
wea
th
er
d
ata
p
atter
n
s
with
ac
cu
r
ate
m
u
lti
-
s
tep
esti
m
ates
[
2
7
]
.
T
h
is
r
esear
ch
p
r
ed
icts
r
ain
f
all
in
th
e
B
an
d
u
n
g
ar
ea
,
wh
ich
is
th
e
ce
n
ter
o
f
th
e
b
asin
,
wh
ich
h
as
a
h
eig
h
t
o
f
7
9
1
m
ab
o
v
e
s
ea
lev
el
(
ASL)
.
T
h
e
h
ig
h
est
p
o
in
t
is
in
th
e
No
r
th
with
a
n
altitu
d
e
o
f
1
0
5
0
m
a
b
o
v
e
s
ea
lev
el,
an
d
th
e
lo
west
p
o
in
t
is
in
th
e
s
o
u
th
with
a
n
elev
atio
n
o
f
6
7
5
m
ab
o
v
e
s
ea
lev
el.
T
h
e
ar
ea
s
u
r
r
o
u
n
d
e
d
b
y
m
o
u
n
tain
s
f
o
r
m
s
th
e
city
o
f
B
an
d
u
n
g
in
to
a
k
in
d
o
f
b
asin
(
B
an
d
u
n
g
B
asin
)
.
T
h
e
s
u
r
r
o
u
n
d
in
g
m
o
u
n
tain
clim
ate
s
ig
n
i
f
ican
tly
a
f
f
ec
ts
th
e
B
an
d
u
n
g
city
clim
ate
.
Ho
wev
er
,
in
r
ec
en
t
y
ea
r
s
,
th
e
tem
p
er
at
u
r
e
h
as
b
ee
n
in
cr
ea
s
in
g
,
a
n
d
th
e
r
ain
y
s
ea
s
o
n
b
ec
o
m
es
m
o
r
e
p
r
o
lo
n
g
ed
th
an
u
s
u
al.
I
n
p
as
t
y
ea
r
s
,
th
e
r
ain
y
s
ea
s
o
n
is
m
o
r
e
in
ten
s
iv
e
h
ap
p
en
i
n
g
in
B
an
d
u
n
g
.
Natu
r
ally
,
B
an
d
u
n
g
is
q
u
ite
a
co
o
l
ar
ea
.
Du
r
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
P
r
ec
ip
ita
tio
n
p
r
ed
ictio
n
u
s
in
g
r
ec
u
r
r
en
t n
eu
r
a
l n
etw
o
r
ks a
n
d
…
(
Mis
h
ka
A
ld
itya
P
r
ia
tn
a
)
2527
th
e
y
ea
r
2
0
1
2
,
r
ec
o
r
d
ed
th
at
th
e
h
ig
h
est
tem
p
er
atu
r
e
in
B
an
d
u
n
g
r
ea
ch
e
d
3
0
.
9
◦
C
,
wh
ich
o
c
cu
r
r
ed
in
Sep
tem
b
er
,
an
d
th
e
lo
west tem
p
er
at
u
r
e
in
B
an
d
u
n
g
in
2
0
1
2
was 1
7
.
4
◦
C
t
h
at
h
ap
p
e
n
ed
in
J
u
ly
.
T
h
is
p
ap
er
p
r
o
p
o
s
ed
a
p
r
ec
ip
itatio
n
p
r
ed
ictio
n
m
o
d
el
o
f
th
e
B
an
d
u
n
g
r
eg
i
o
n
u
s
in
g
p
r
e
cip
itatio
n
,
h
u
m
id
ity
,
tem
p
e
r
atu
r
e,
s
o
lar
r
a
d
iatio
n
o
f
3
6
-
y
ea
r
s
b
e
f
o
r
e
.
T
h
e
m
o
d
el
u
s
in
g
R
NNs
with
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
L
STM
)
to
p
r
e
d
ict
p
r
ec
ip
itati
o
n
.
T
h
e
u
n
i
q
u
e
n
ess
o
f
th
e
R
NNs
is
th
e
f
ee
d
b
ac
k
co
n
n
ec
ti
o
n
,
wh
ich
co
n
v
ey
s
in
ter
f
er
en
ce
in
f
o
r
m
atio
n
at
th
e
p
r
ev
io
u
s
in
p
u
t
th
at
will
b
e
ac
co
m
m
o
d
ate
d
to
th
e
n
ex
t
d
ata
.
Ma
ch
in
e
lear
n
in
g
u
s
ed
th
is
m
o
d
el
Netwo
r
k
b
ase
d
o
n
co
n
s
ec
u
tiv
e
tim
e
with
p
r
io
r
clim
ate
d
ata
,
s
o
p
r
o
d
u
ce
d
r
ain
f
all
p
r
ed
ictio
n
.
T
h
e
in
p
u
t
v
a
r
iab
les
o
f
m
ac
h
in
e
lear
n
in
g
ar
e
m
in
im
u
m
tem
p
er
atu
r
e,
m
ax
im
u
m
tem
p
er
atu
r
e,
av
e
r
ag
e
tem
p
er
atu
r
e,
r
elativ
e
h
u
m
id
it
y
,
d
u
r
atio
n
o
f
s
u
n
r
ad
iatio
n
,
av
er
ag
e
win
d
s
p
ee
d
,
m
ax
im
u
m
win
d
s
p
ee
d
,
an
d
p
r
ec
ip
itatio
n
to
p
r
ed
ict
r
ai
n
f
all
in
a
ce
r
tain
p
e
r
io
d
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
2
.
1
.
D
a
t
a
s
et
W
ea
th
er
d
ata
p
r
o
v
id
ed
b
y
th
e
I
n
d
o
n
esian
Ag
en
cy
f
o
r
Me
teo
r
o
lo
g
ical,
C
lim
ato
lo
g
ical
,
an
d
(
B
MK
G)
f
r
o
m
1
9
8
1
to
2
0
1
7
.
T
h
is
r
esea
r
ch
u
s
in
g
th
e
B
an
d
u
n
g
city
r
e
g
io
n
in
th
e
an
aly
s
is
.
Data
s
ets
p
r
o
v
id
e
d
c
o
n
s
is
ts
o
f
eig
h
t
v
a
r
iab
les
(
m
in
im
u
m
te
m
p
er
atu
r
e,
m
ax
i
m
u
m
tem
p
er
atu
r
e,
a
v
er
ag
e
tem
p
er
at
u
r
e,
r
elativ
e
h
u
m
id
ity
,
d
u
r
atio
n
o
f
s
u
n
r
ad
iatio
n
,
a
v
er
ag
e
win
d
s
p
ee
d
,
m
a
x
im
u
m
win
d
s
p
ee
d
,
an
d
p
r
ec
i
p
itatio
n
)
of
3
6
y
ea
r
s
(
1
9
8
1
-
2
0
1
7
)
.
T
h
is
r
esear
ch
u
s
ed
two
p
er
io
d
s
o
f
p
r
ec
ip
i
tatio
n
p
r
e
d
ictio
n
,
m
ain
ly
we
ek
ly
a
n
d
m
o
n
t
h
ly
co
n
f
ig
u
r
atio
n
s
.
T
h
e
v
alu
es
o
f
clim
ate
v
ar
iab
les
in
th
at
p
er
io
d
ar
e
th
e
d
aily
av
er
ag
es
,
m
in
i
m
u
m
o
r
m
ax
im
u
m
.
T
h
er
e
a
r
e
f
o
u
r
s
ce
n
ar
io
s
i
n
th
e
ex
p
e
r
im
en
t,
p
ar
ticu
lar
ly
,
v
ar
i
atio
n
s
in
th
e
am
o
u
n
t
o
f
tr
ain
in
g
d
ata
(
1
0
y
ea
r
s
an
d
f
iv
e
y
ea
r
s
)
,
a
n
d
th
e
d
u
r
atio
n
o
f
th
e
p
r
e
d
ictio
n
s
(
m
o
n
th
ly
a
n
d
wee
k
ly
)
with
d
etails:
−
10
-
y
ea
r
s
: a
n
n
u
al
(
4
3
2
d
ata
s
ets,
o
v
er
lap
1
1
m
o
n
th
s
)
−
10
-
y
ea
r
s
: m
o
n
th
ly
(
1
8
7
2
d
ata
s
ets,
o
v
er
lap
5
0
w
ee
k
s
)
−
5
-
y
ea
r
s
: a
n
n
u
al
(
8
6
4
d
ata
s
ets,
o
v
er
lap
1
1
m
o
n
th
s
)
−
5
-
y
ea
r
s
: m
o
n
th
ly
(
3
7
4
4
d
ata
s
ets,
o
v
er
lap
5
0
wee
k
s
)
All o
f
th
e
m
o
d
els,
7
5
% is
u
s
ed
f
o
r
t
r
ain
in
g
d
ata,
an
d
2
5
% is
u
s
ed
f
o
r
n
o
n
-
tr
ain
in
g
o
r
test
d
ata.
2
.
2
.
P
re
cipita
t
i
o
n pre
dict
io
n m
o
del
T
h
e
d
esig
n
o
f
p
r
ec
i
p
itatio
n
p
r
e
d
ictio
n
u
s
in
g
R
NN
an
d
L
STM
is
s
h
o
wn
in
Fig
u
r
e
1
.
Pre
d
ictio
n
ch
o
o
s
es
o
n
e
o
f
ten
class
es
wi
th
a
s
p
e
cif
ic
in
ter
v
al.
Pas
t
r
esear
ch
u
s
e
d
m
u
ltiv
ar
iates
f
o
r
r
ain
f
all
p
r
ed
ictio
n
[
2
8
]
with
R
NN
an
d
L
STM
.
So
m
etim
es,
clim
ate
d
ata
h
as
m
is
s
in
g
o
r
lo
s
t
o
b
s
er
v
atio
n
.
T
h
er
ef
o
r
e,
th
e
d
ata
n
ee
d
to
p
r
ep
ar
e
au
t
o
m
atica
lly
b
ef
o
r
e
th
e
n
ex
t
p
r
o
ce
s
s
.
So
m
e
s
o
lu
tio
n
is
an
in
ter
p
o
latio
n
o
f
s
o
m
e
av
ailab
le
d
ata,
m
u
ltiv
ar
iate
[
2
9
]
,
o
r
p
r
ed
ict
t
h
e
m
is
s
in
g
u
s
in
g
r
ef
in
em
e
n
t
f
u
n
ctio
n
[
1
0
]
.
Me
an
wh
ile,
B
MK
G
p
r
o
v
id
es
d
aily
d
ata.
T
h
er
e
f
o
r
e,
wee
k
ly
an
d
m
o
n
th
ly
p
r
ed
ictio
n
s
n
ee
d
to
co
n
v
er
t
d
aily
o
r
wee
k
ly
d
ata.
So
m
e
s
tu
d
ies
u
s
e
d
th
e
av
er
a
g
e
v
alu
e
in
t
h
is
tim
e
f
r
am
e
s
o
th
at
it
r
ef
lects
its
p
r
o
je
ctio
n
s
[
3
0
]
.
T
h
is
r
esear
ch
u
s
ed
v
ar
io
u
s
v
a
r
iab
les,
i.e
.
,
m
in
im
u
m
tem
p
er
at
u
r
e,
m
ax
im
u
m
tem
p
e
r
atu
r
e,
av
er
a
g
e
tem
p
er
atu
r
e,
r
elativ
e
h
u
m
id
it
y
,
d
u
r
atio
n
o
f
s
u
n
r
ad
iatio
n
,
av
er
a
g
e
win
d
s
p
ee
d
,
m
ax
im
u
m
win
d
s
p
ee
d
,
an
d
p
r
ec
ip
itatio
n
,
wh
ich
h
av
e
d
if
f
er
e
n
t
u
n
its
.
T
h
er
ef
o
r
e,
all
d
atasets
b
ef
o
r
e
en
ter
in
g
th
e
R
NN
ar
e
n
o
r
m
alize
d
ea
r
lier
[
1
8
,
2
5
]
u
s
in
g
(
1
)
.
T
h
e
n
o
r
m
aliza
tio
n
tak
e
s
th
e
m
ax
im
u
m
an
d
m
in
im
u
m
v
alu
es
o
n
th
e
0
-
1
s
ca
le.
=
−
(
)
(
)
−
(
)
(
1
)
B
ased
o
n
th
e
f
o
r
ec
ast
p
er
io
d
,
it
is
d
iv
id
ed
in
to
two
m
o
d
els,
i.e
.
,
wee
k
ly
an
d
m
o
n
th
ly
.
T
h
e
r
esu
lts
o
f
th
e
s
tu
d
y
ar
e
th
en
co
m
p
ar
ed
an
d
ca
n
b
e
u
tili
ze
d
f
o
r
th
eir
in
d
iv
id
u
al
n
ee
d
s
.
W
h
en
u
s
ed
f
o
r
f
lo
o
d
d
is
aster
m
an
ag
em
en
t,
t
h
e
s
elec
tio
n
o
f
th
e
wee
k
is
m
o
r
e
ap
p
r
o
p
r
iate,
tak
in
g
in
to
ac
c
o
u
n
t
th
e
ca
r
r
y
i
n
g
ca
p
ac
ity
o
f
s
o
il
ab
s
o
r
p
tio
n
.
W
h
ile
th
e
r
an
g
e
o
f
th
e
p
lan
tin
g
s
ea
s
o
n
ca
n
u
s
e
a
m
o
n
th
ly
p
er
io
d
,
R
NN
ca
n
ad
ap
t
to
f
lex
ib
le
class
es
[
2
9
]
.
I
t
ca
n
tr
u
e
p
er
io
d
o
r
p
s
eu
d
o
p
e
r
io
d
.
I
n
th
e
m
ea
n
tim
e,
we
u
s
ed
r
ain
f
all
p
r
ed
ic
tio
n
s
with
a
ce
r
tain
r
an
g
e
o
f
1
0
p
r
e
-
d
eter
m
i
n
ed
c
lass
es,
i.e
.
<
6
0
m
m
,
6
0
-
1
2
0
,
1
2
0
-
180
,
1
8
0
-
2
4
0
,
2
4
0
-
3
0
0
,
3
0
0
-
3
6
0
,
3
6
0
-
4
2
0
,
420
-
4
8
0
,
4
8
0
-
5
4
0
,
an
d
>
5
4
0
m
m
as
s
h
o
wn
in
Fig
u
r
e
1
.
T
h
en
g
o
to
th
e
s
ec
o
n
d
s
tep
with
R
NN.
T
h
e
d
r
o
p
o
u
t
lay
er
s
u
s
ed
t
o
m
in
im
ize
th
e
n
u
m
b
er
o
f
in
p
u
t
n
e
u
r
o
n
s
0
.
5
p
r
o
b
ab
ilit
y
th
at
in
p
u
t
n
e
u
r
o
n
to
th
e
n
ex
t
s
tep
is
4
8
0
.
T
h
e
th
ir
d
s
tep
is
L
STM
lay
er
2
,
with
th
e
in
p
u
t
d
r
o
p
o
u
t
lay
e
r
u
s
in
g
(
2
)
-
(
7
)
.
T
h
e
f
o
u
r
th
s
tep
is
th
e
d
en
s
e
lay
er
u
s
in
g
th
e
s
ig
m
o
id
f
u
n
ctio
n
,
wh
er
e
th
e
f
in
al
r
esu
lt
f
r
o
m
th
e
p
r
e
v
io
u
s
is
en
ter
ed
in
to
(
1
)
to
p
r
o
d
u
ce
a
n
ew
weig
h
t.
2.
3
.
Rec
urre
nt
neura
l net
wo
rk
s
Dee
p
lear
n
in
g
tech
n
iq
u
es
h
a
v
e
b
ee
n
s
u
cc
ess
f
u
lly
ap
p
lied
t
o
s
o
lv
e
m
an
y
p
r
o
b
lem
s
in
clim
ate
an
d
g
eo
s
cien
ce
u
s
in
g
m
ass
iv
e
-
s
ca
led
o
b
s
er
v
ed
an
d
m
o
d
ele
d
d
a
ta
[
3
1
]
.
On
e
m
eth
o
d
o
f
d
ee
p
lear
n
in
g
is
R
NN
.
Pre
v
io
u
s
r
esea
r
ch
p
r
o
p
o
s
ed
t
r
ain
in
g
th
r
ee
m
o
d
els
o
f
d
ee
p
-
lear
n
in
g
:
R
NN,
co
n
d
itio
n
al
r
e
s
tr
icted
b
o
ltzm
an
n
m
ac
h
in
e
(
C
R
B
M)
,
an
d
C
NN
u
s
in
g
E
NSO
an
d
W
ea
th
er
d
ataset
[1
5]
.
T
h
e
b
est
ac
cu
r
ac
y
was
R
NN
u
n
til
8
4
%
of
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
18
,
No
.
5
,
Octo
b
e
r
2
0
2
0
:
2
5
2
5
-
2532
2528
p
r
ec
ip
itatio
n
f
o
r
ec
asti
n
g
u
s
i
n
g
a
d
ee
p
b
elief
n
etwo
r
k
ca
lled
DB
NPF
[
6
]
.
T
h
e
r
e
s
ea
r
ch
estab
lis
h
ed
th
e
ch
ar
ac
ter
is
tics
o
f
en
v
ir
o
n
m
en
tal
f
ac
to
r
s
an
d
f
u
tu
r
e
p
r
e
cip
itatio
n
.
Oth
er
s
tu
d
ies
p
r
o
p
o
s
ed
to
p
r
ed
ict
th
e
ex
tr
em
e
wea
th
er
u
s
in
g
Ar
tific
ial
Neu
r
al
Netwo
r
k
an
d
Su
p
p
o
r
t V
ec
to
r
Ma
ch
in
es
[
3
2
]
.
R
NN
wo
r
k
s
r
esem
b
le
th
e
wo
r
k
in
g
s
o
f
t
h
e
b
r
ai
n
,
s
u
ch
as
s
en
d
in
g
an
d
r
ec
eiv
i
n
g
in
f
o
r
m
atio
n
.
Similar
ly
,
th
e
wo
r
k
o
f
th
e
R
NN
b
r
ain
ca
n
also
s
en
d
an
d
r
ec
eiv
e
d
a
ta
f
r
o
m
a
n
eu
r
o
n
to
a
n
o
th
e
r
n
eu
r
o
n
.
I
n
g
en
er
al,
th
e
h
u
m
an
b
r
ain
is
u
s
ed
to
m
ak
e
d
ec
is
io
n
s
,
an
d
in
th
e
p
r
o
ce
s
s
o
f
m
ak
in
g
a
r
atio
n
al
d
ec
is
io
n
,
o
f
ten
tak
e
s
in
to
ac
co
u
n
t
th
e
p
ast.
R
NN
is
es
s
en
tially
an
ar
tific
ial
n
eu
r
al
n
e
two
r
k
th
at
u
s
es
r
ec
u
r
r
en
ce
b
y
u
tili
zin
g
p
ast
d
ata.
R
NN
ca
n
p
r
e
d
ict
a
s
i
tu
atio
n
in
th
e
f
u
tu
r
e
[
1
5
]
,
an
d
ca
n
also
c
lass
if
y
[
3
3
]
.
R
NN
h
as
a
n
a
r
ch
i
tectu
r
e
th
at
ca
n
b
e
u
s
ed
f
o
r
d
ata
i
n
th
e
f
o
r
m
o
f
s
eq
u
en
ce
o
r
lis
t,
as
s
h
o
wn
in
Fig
u
r
e
2
.
R
NN
is
a
m
o
d
if
icatio
n
o
f
th
e
Feed
f
o
r
war
d
Neu
r
al
Netwo
r
k
with
th
e
c
h
ar
ac
ter
is
tic
o
f
u
s
in
g
f
ee
d
b
ac
k
f
r
o
m
o
u
t
p
u
t to
in
p
u
t.
W
e
a
t
h
e
r
T
r
a
i
n
i
n
g
D
a
t
a
W
e
i
g
h
t
T
r
a
i
n
i
n
g
u
s
i
n
g
R
e
c
u
r
r
e
n
t
N
e
u
r
a
l
N
e
t
w
o
r
k
s
L
S
T
M
L
a
y
e
r
1
(
R
e
l
u
)
D
r
o
p
o
u
t
L
a
y
e
r
(
0
.
2
)
L
S
T
M
L
a
y
e
r
2
(
S
i
g
m
o
i
d
)
D
e
n
s
e
L
a
y
e
r
(
s
i
g
m
o
i
d
)
W
e
a
t
h
e
r
D
a
t
a
W
e
e
k
l
y
o
r
m
o
n
t
h
l
y
P
r
e
d
i
c
t
i
o
n
R
a
i
n
f
a
l
l
u
s
i
n
g
R
e
c
u
r
r
e
n
t
N
e
u
r
a
l
N
e
t
w
o
r
k
s
L
S
T
M
L
a
y
e
r
1
(
R
e
l
u
)
D
r
o
p
o
u
t
L
a
y
e
r
(
0
.
2
)
L
S
T
M
L
a
y
e
r
2
(
S
i
g
m
o
i
d
)
D
e
n
s
e
L
a
y
e
r
(
s
i
g
m
o
i
d
)
<
6
0
6
0
-
1
2
0
1
2
0
-
1
8
0
1
8
0
-
2
4
0
2
4
0
-
3
0
0
3
0
0
-
3
6
0
3
6
0
-
4
2
0
4
2
0
-
4
8
0
4
8
0
-
5
4
0
>
5
4
0
P
r
e
-
P
r
o
c
e
s
s
i
n
g
M
i
s
s
i
n
g
d
a
t
a
h
a
n
d
l
i
n
g
C
o
n
v
e
r
t
t
o
p
e
r
i
o
d
e
N
o
r
m
a
l
i
z
e
P
r
e
-
P
r
o
c
e
s
s
i
n
g
M
i
s
s
i
n
g
d
a
t
a
h
a
n
d
l
i
n
g
C
o
n
v
e
r
t
t
o
p
e
r
i
o
d
e
N
o
r
m
a
l
i
z
e
I
n
t
e
r
v
a
l
p
r
e
c
i
p
i
t
a
t
i
o
n
p
r
e
d
i
c
t
i
o
n
(
t
e
n
c
l
a
s
s
e
s
)
Fi
g
u
r
e
1
.
Pre
cip
itatio
n
p
r
ed
icti
o
n
m
o
d
el
y
t
h
1
U
V
W
=
x
t
y
0
h
0
U
V
W
x
0
y
1
h
1
U
V
W
x
1
y
2
h
2
U
V
W
x
2
.
.
.
y
t
-
1
h
t
U
V
W
x
t
-
1
y
t
Fig
u
r
e
2
.
R
NNs a
r
ch
itectu
r
e
I
n
th
e
Ar
ch
itectu
r
e
o
f
R
NNs,
th
er
e
ar
e
s
ev
er
al
co
n
n
ec
tio
n
s
f
r
o
m
o
n
e
n
e
u
r
o
n
to
o
n
e
o
f
th
e
n
e
x
t
n
eu
r
o
n
s
.
R
NN
is
p
r
o
ce
s
s
ed
in
a
s
eq
u
en
ce
of
tim
e
.
So
th
at
ea
ch
in
f
o
r
m
atio
n
h
as
a
r
elatio
n
s
h
ip
with
o
n
e
an
o
th
er
.
T
h
is
way
m
ak
es
th
e
R
NN
h
as
a
m
e
m
o
r
y
th
at
f
u
n
ctio
n
s
to
r
e
m
em
b
er
th
e
r
esu
lts
o
f
th
e
p
r
ev
i
o
u
s
p
r
o
ce
s
s
th
at
will
b
e
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
P
r
ec
ip
ita
tio
n
p
r
ed
ictio
n
u
s
in
g
r
ec
u
r
r
en
t n
eu
r
a
l n
etw
o
r
ks a
n
d
…
(
Mis
h
ka
A
ld
itya
P
r
ia
tn
a
)
2529
u
s
ed
in
th
e
n
ex
t
p
r
o
ce
s
s
.
Ho
w
ev
er
,
wh
e
n
th
e
R
NN
p
r
o
ce
s
s
es
q
u
ite
a
l
o
t
o
f
d
ata,
it
h
as
d
if
f
ic
u
lty
in
m
ai
n
tain
in
g
in
f
o
r
m
atio
n
f
r
o
m
th
e
p
r
ev
i
o
u
s
s
tep
s
.
T
h
e
f
ir
s
t
h
id
d
en
lay
e
r
h
as
th
e
weig
h
t
o
b
tain
ed
f
r
o
m
th
e
in
p
u
t
la
y
er
,
a
n
d
at
ea
ch
lay
er
will
r
ec
eiv
e
th
e
weig
h
t
o
f
th
e
p
r
io
r
lay
er
.
T
h
en
,
th
e
ca
lcu
latio
n
o
f
th
e
n
ex
t
h
id
d
en
lay
er
u
s
es
th
e
ap
p
r
o
p
r
iate
en
tity
in
th
e
p
r
ev
io
u
s
in
p
u
t a
n
d
h
id
d
e
n
lay
er
.
Me
an
wh
ile,
th
e
f
o
r
ec
ast f
o
r
t
h
e
o
u
tp
u
t la
y
er
u
s
es
th
e
l
ast
h
id
d
en
lay
er
.
Fo
r
t
h
e
p
r
o
ce
s
s
o
f
ca
lcu
latin
g
th
e
h
id
d
e
n
lay
er
u
s
in
g
f
u
n
ctio
n
s
th
at
ca
n
b
e
s
ee
n
in
(
2
)
an
d
th
e
ca
lcu
latio
n
o
f
o
u
t
p
u
t u
s
in
g
s
o
f
tm
ax
f
u
n
ctio
n
wh
ic
h
ca
n
b
e
s
ee
n
in
(
3
)
.
ℎ
(
)
=
ta
n
h
(
(
)
+
ℎ
(
−
1
)
+
)
(
2
)
(
)
=
(
ℎ
(
)
+
)
(
3
)
I
n
th
is
s
tu
d
y
,
th
e
d
ata
co
llectio
n
th
at
was
p
r
o
ce
s
s
ed
c
o
n
s
is
ts
o
f
v
ar
io
u
s
k
in
d
s
o
f
d
ata
s
eq
u
en
ce
s
s
u
ch
as
m
eteo
r
o
lo
g
y
v
ar
iab
les
,
s
o
i
t
is
n
ec
ess
ar
y
to
g
r
o
u
p
d
ata
i
n
th
e
i
n
p
u
t
lay
er
t
o
p
r
o
v
id
e
t
h
e
tr
ain
in
g
p
r
o
ce
s
s
.
Ho
wev
er
,
R
NN
h
as
th
e
d
i
s
ad
v
an
tag
e
o
f
s
h
o
r
t
-
ter
m
m
em
o
r
y
.
T
h
er
ef
o
r
e,
in
m
ak
in
g
p
r
e
d
ictio
n
s
u
s
in
g
q
u
ite
a
lo
t
o
f
d
ata,
R
NN
h
as
s
ev
er
al
v
ar
iatio
n
s
to
s
o
lv
e
th
e
p
r
o
b
lem
.
T
h
e
g
ate
is
th
e
g
ate
d
r
ec
u
r
r
en
t
u
n
it
(
GR
U)
,
b
ac
k
p
r
o
p
ag
atio
n
th
r
o
u
g
h
tim
e
(
B
PTT
)
,
an
d
L
STM
.
T
h
is
r
e
s
ea
r
ch
u
s
e
d
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
to
o
v
er
co
m
e
th
e
v
a
n
is
h
in
g
g
r
ad
i
en
t
[
3
4
]
.
T
h
e
R
NNs
tr
ain
in
g
p
r
o
ce
s
s
is
s
im
ilar
to
n
eu
r
al
n
etwo
r
k
tr
ai
n
in
g
u
s
in
g
a
b
a
ck
p
r
o
p
ag
atio
n
alg
o
r
ith
m
b
u
t
with
f
ew
cy
cles.
T
h
e
p
a
r
a
m
eter
th
at
s
h
ar
ed
eq
u
ally
in
ev
er
y
tim
e
s
tep
,
s
o
g
r
ad
ie
n
t
f
o
r
e
ac
h
o
u
tp
u
t,
d
o
es
n
o
t
o
n
ly
d
e
p
en
d
o
n
a
ca
lcu
latio
n
f
r
o
m
th
e
cu
r
r
en
t
tim
e
s
tep
b
u
t
also
th
e
p
r
e
v
io
u
s
o
n
e
[
3
5
]
.
T
h
is
r
esear
ch
u
s
ed
L
STM
g
ates
to
o
v
e
r
co
m
e
t
h
e
d
ep
en
d
en
ce
o
f
lo
n
g
-
ter
m
p
r
o
ce
s
s
,
wh
ich
is
o
f
ten
p
h
en
o
m
e
n
a
th
at
o
cc
u
r
r
e
d
in
s
eq
u
en
tial d
ata
p
r
o
ce
s
s
in
g
,
as sh
o
wn
in
Fig
u
r
e
3
.
L
S
T
M
L
S
T
M
x
+
x
x
t
a
n
h
t
a
n
h
σ
σ
σ
C
t
-
1
S
t
-
1
t
-
1
t
+
1
C
t
S
t
x
t
S
t
t
Fig
u
r
e
3
.
L
STM
ar
c
h
itectu
r
e
T
h
e
k
ey
f
r
o
m
L
STM
is
ce
ll
s
tate
with
ar
ch
itectu
r
e
m
ar
k
ed
b
y
a
h
o
r
izo
n
tal
lin
e
th
at
f
lo
ws
f
r
o
m
C
t
-
1
u
n
til
C
t
.
L
STM
ca
n
d
elete
o
r
a
d
d
in
f
o
r
m
atio
n
to
th
e
ce
ll
s
tate
s
et
b
y
th
e
s
tr
u
ctu
r
e,
wh
ich
is
ca
lled
th
e
g
ate.
Gate
is
a
m
eth
o
d
to
p
ass
th
e
in
f
o
r
m
atio
n
,
wh
ich
c
o
n
s
is
ts
o
f
b
in
er
s
ig
m
o
id
f
u
n
ctio
n
(
)
an
d
m
u
lt
ip
licat
io
n
o
p
er
atio
n
with
x
.
B
in
er
Sig
m
o
id
f
u
n
ctio
n
is
as
s
h
o
wn
i
n
(
4
)
.
T
h
e
f
ir
s
t
s
tep
in
L
STM
is
d
ec
id
in
g
wh
at
in
f
o
r
m
atio
n
will
b
e
d
is
p
o
s
ed
o
f
f
r
o
m
t
h
e
ce
ll st
ate
ca
lled
f
o
r
g
et
g
ate
u
s
ed
(
5
)
[
3
4
]
.
(
)
=
1
1
+
−
(
4
)
=
(
[
ℎ
−
1
,
]
+
)
(
5
)
wh
er
e
h
t
-
1
an
d
x
t
v
alu
es,
ar
e
in
0
-
1
in
ter
v
al
e
v
er
y
ce
ll.
Use o
f
1
,
wh
ich
r
ep
r
esen
ts
th
is
in
f
o
r
m
atio
n
,
is
k
ep
t,
an
d
0
,
wh
ich
ex
p
r
ess
es
th
is
in
f
o
r
m
atio
n
is
d
elete
d
.
T
h
e
s
ec
o
n
d
is
d
ec
id
in
g
r
ec
e
n
t
d
ata
s
to
r
ed
in
th
e
ce
ll.
T
h
is
s
tep
is
d
iv
id
ed
in
to
two
p
ar
ts
.
First
,
I
n
p
u
t
g
ate
will
d
eter
m
in
e
v
alu
es
th
at
will
b
e
u
p
d
a
ted
u
s
in
g
(
6
)
an
d
ca
lcu
latio
n
f
o
r
r
ec
e
n
t
ce
ll
ca
n
d
id
ate
(
̂
)
wh
ich
;
=
(
[
ℎ
−
1
,
]
+
)
(
6
)
̂
=
ℎ
(
[
ℎ
−
1
,
]
+
)
(
7
)
T
h
er
e
will
b
e
ad
d
ed
to
th
e
o
ld
ce
ll
s
tate
(
C
t
)
with
a
ce
ll
(
̂
)
ca
n
d
id
ate
u
s
in
g
(
8
)
.
T
h
e
f
o
r
m
e
r
ce
ll
s
tate
m
u
ltip
licated
b
y
f
o
r
g
ettin
g
s
tate
.
A
n
d
ce
ll
ca
n
d
id
ate
m
u
ltip
licated
with
th
e
in
p
u
t
g
ate.
I
t
u
p
d
ated
th
e
ce
ll
s
tate.
t
i
C
f
C
C
t
t
t
t
~
1
*
*
+
=
−
(
8
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
18
,
No
.
5
,
Octo
b
e
r
2
0
2
0
:
2
5
2
5
-
2532
2530
T
h
e
l
a
s
t
s
t
e
p
i
s
t
h
e
o
u
t
p
u
t
g
a
t
e
u
s
e
d
t
o
d
e
t
e
r
m
i
n
e
w
h
i
c
h
o
u
t
p
u
t
w
i
l
l
b
e
p
r
o
d
u
c
e
d
b
a
s
e
d
o
n
c
e
l
l
s
t
a
t
e
f
r
o
m
t
h
e
r
e
s
u
l
t
s
o
f
(
6
)
c
a
l
c
u
l
a
t
ed
w
i
t
h
t
h
e
b
i
n
e
r
s
i
g
m
o
i
d
f
u
n
c
t
i
o
n
a
s
s
h
o
w
n
i
n
(
9
)
F
u
r
t
h
e
r
m
o
r
e
,
m
u
l
t
i
p
l
i
c
a
t
e
d
w
i
t
h
a
c
t
i
v
a
t
i
n
g
f
u
n
c
t
i
o
n
f
r
o
m
u
p
d
a
t
e
d
c
e
l
l
s
t
a
t
e
u
s
i
n
g
(
10
)
.
S
o
m
e
p
r
e
v
i
o
u
s
r
e
s
e
a
r
c
h
u
s
e
d
s
i
g
m
o
i
d
a
n
d
t
a
n
h
[
3
5
]
.
=
(
[
ℎ
−
1
,
]
+
)
(
9
)
ℎ
=
∗
ℎ
(
)
(
10
)
3.
RE
SU
L
T
S
A
ND
AN
AL
Y
SI
S
E
x
p
er
im
en
ts
f
r
o
m
p
r
ec
ip
itatio
n
p
r
e
d
ictio
n
m
o
d
els
ar
e
ca
r
r
i
ed
o
u
t
with
v
ar
iatio
n
s
in
th
e
i
n
ter
v
al
o
f
tr
ain
in
g
d
ata
(
1
0
y
ea
r
s
an
d
f
iv
e
y
ea
r
s
)
,
a
n
d
th
e
p
r
ed
icti
o
n
tim
e
(
m
o
n
th
ly
a
n
d
wee
k
l
y
)
,
with
th
e
R
NN
co
n
f
ig
u
r
atio
n
in
T
ab
le
1
.
I
n
g
ettin
g
th
e
o
p
tim
al
p
r
ed
ictio
n
,
v
ar
iatio
n
s
ar
e
p
er
f
o
r
m
ed
o
n
t
h
e
d
ataset
tim
e
an
d
p
r
ed
ictio
n
d
u
r
ati
o
n
.
T
h
e
e
x
p
er
im
en
t
also
test
ed
th
e
ac
cu
r
ac
y
o
f
t
h
e
o
p
tim
izatio
n
m
o
d
el,
th
e
ad
ap
tiv
e
m
o
m
e
n
t
esti
m
atio
n
(
Ad
am
)
m
o
d
el,
an
d
th
e
s
to
ch
asti
c
g
r
ad
ien
t
d
escen
t
(
SGD)
m
o
d
el.
Acc
u
r
ac
y
is
ca
lcu
lated
f
r
o
m
th
e
ac
cu
r
ac
y
o
f
t
h
e
o
u
tp
u
t
class
ag
ain
s
t
th
e
ac
t
u
al
class
lab
el.
T
h
is
r
esear
ch
u
s
ed
two
p
r
ed
ic
tiv
e
m
o
d
els.
First,
it u
s
ed
wea
th
er
d
ata
f
o
r
ten
y
e
ar
s
an
d
f
iv
e
y
ea
r
s
.
B
o
th
m
o
d
el
s
ar
e
test
ed
wee
k
ly
an
d
m
o
n
th
ly
.
T
h
e
ac
cu
r
ac
y
is
o
b
tain
ed
as
in
T
a
b
le
2
,
Fig
u
r
e
4
o
f
10
-
y
ea
r
s
d
at
aset,
an
d
Fig
u
r
e
5
o
f
th
e
5
-
y
ea
r
s
d
ataset
.
T
h
is
s
im
u
latio
n
d
ev
elo
p
e
d
u
s
in
g
two
o
p
tim
ize
r
m
o
d
el
(
Ad
am
an
d
SGD)
to
c
o
r
r
ec
t w
eig
h
t
,
i
n
2
0
0
ep
o
ch
.
T
ab
le
1
.
R
NN
co
n
f
ig
u
r
atio
n
C
o
n
f
i
g
u
r
a
t
i
o
n
10
-
y
e
a
r
s
5
-
y
e
a
r
s
w
e
e
k
l
y
mo
n
t
h
l
y
w
e
e
k
l
y
mo
n
t
h
l
y
D
a
t
a
s
e
t
1
8
7
2
4
3
2
3
7
4
4
8
6
4
I
n
p
u
t
4
1
6
0
9
6
0
2
0
8
0
4
8
0
H
i
d
d
e
n
4
1
6
0
64
2
0
8
0
64
D
r
o
p
o
u
t
0
.
2
0
.
2
0
.
2
0
.
2
D
e
n
se
13
13
13
13
O
u
t
p
u
t
l
a
y
e
r
10
10
10
10
T
ab
le
2
.
Acc
u
r
ac
y
o
f
Pre
cip
ita
tio
n
to
war
d
d
ataset
an
d
d
u
r
atio
n
M
o
d
e
l
A
c
c
u
r
a
c
y
(
%)
10
-
y
e
a
r
s
–
4
3
2
d
a
t
a
s
e
t
5
-
y
e
a
r
s
–
8
6
4
d
a
t
a
se
t
W
e
e
k
l
y
M
o
n
t
h
l
y
W
e
e
k
l
y
M
o
n
t
h
l
y
Tr
a
i
n
i
n
g
D
a
t
a
Te
st
D
a
t
a
Tr
a
i
n
i
n
g
D
a
t
a
Te
st
D
a
t
a
Tr
a
i
n
i
n
g
D
a
t
a
Te
st
D
a
t
a
Tr
a
i
n
i
n
g
D
a
t
a
Te
st
D
a
t
a
A
d
a
m
9
9
.
2
1
8
0
.
9
5
9
6
.
6
0
7
3
.
2
5
9
9
.
6
0
8
5
.
7
1
9
9
.
2
1
8
3
.
3
3
S
G
D
9
7
.
2
2
6
5
.
0
7
9
1
.
6
6
6
1
.
4
2
9
6
.
4
2
7
9
.
3
6
9
2
.
8
5
7
5
.
2
1
(a
)
(b
)
Fig
u
r
e
4
.
T
h
e
ac
cu
r
ac
y
o
f
1
0
y
ea
r
s
–
4
3
2
d
ata
s
et
(
a)
wee
k
l
y
(
b
)
m
o
n
th
l
y
(
a)
(
b
)
Fig
u
r
e
5
.
T
h
e
ac
cu
r
ac
y
o
f
5
y
e
ar
s
–
8
6
4
d
ata
s
et
(
a)
wee
k
ly
(
b
)
m
o
n
th
l
y
T
h
i
s
s
t
u
d
y
i
s
l
o
o
k
i
n
g
f
o
r
r
e
p
r
e
s
e
n
t
a
t
i
v
e
d
a
t
a
s
e
t
s
r
e
p
r
e
s
e
n
t
i
n
g
p
a
t
t
e
r
n
s
o
f
r
a
i
n
f
a
l
l
o
c
c
u
r
r
e
n
c
e
.
F
r
o
m
T
a
b
l
e
2
,
i
t
i
s
s
h
o
w
n
t
h
a
t
a
n
i
n
t
e
r
v
a
l
o
f
5
-
y
e
a
r
s
f
o
r
w
e
a
t
h
e
r
p
r
e
d
i
c
t
i
o
n
i
s
e
n
o
u
g
h
t
o
p
r
o
v
i
d
e
b
e
t
t
e
r
a
c
c
u
r
a
c
y
.
T
h
e
u
s
e
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
P
r
ec
ip
ita
tio
n
p
r
ed
ictio
n
u
s
in
g
r
ec
u
r
r
en
t n
eu
r
a
l n
etw
o
r
ks a
n
d
…
(
Mis
h
ka
A
ld
itya
P
r
ia
tn
a
)
2531
o
f
s
h
o
r
t
e
r
t
i
m
e
r
a
n
g
e
s
g
i
v
e
s
t
h
e
c
o
n
s
e
q
u
e
n
c
e
o
f
t
h
e
m
a
n
y
v
a
r
i
a
t
i
o
n
s
o
f
t
r
a
i
n
i
n
g
d
a
t
a
,
o
f
c
o
u
r
s
e
,
p
r
o
v
i
d
i
n
g
b
e
t
t
e
r
a
c
c
u
r
a
c
y
.
B
a
s
e
d
o
n
F
i
g
u
r
e
5
a
n
d
T
a
b
l
e
2
,
v
i
s
i
b
l
e
5
-
y
e
a
r
d
a
t
a
s
e
t
s
p
r
o
v
i
d
e
a
n
a
c
c
u
r
a
c
y
o
f
8
3
.
3
3
%
f
o
r
m
o
n
t
h
l
y
a
n
d
8
5
.
7
1
%
f
o
r
w
e
e
k
l
y
.
W
h
i
l
e
i
n
F
i
g
u
r
e
5
g
a
v
e
a
n
a
c
c
u
r
a
c
y
o
f
7
3
.
2
5
%
f
o
r
m
o
n
t
h
l
y
a
n
d
8
0
.
9
5
%
f
o
r
w
e
e
k
l
y
.
T
h
i
s
r
e
s
u
l
t
s
t
r
e
n
g
t
h
e
n
s
t
h
e
h
y
p
o
t
h
e
s
i
s
t
h
a
t
t
h
e
a
m
o
u
n
t
o
f
t
r
a
i
n
i
n
g
d
a
t
a
i
s
m
o
r
e
d
o
m
i
n
a
n
t
t
h
a
n
t
h
e
n
u
m
b
e
r
o
f
v
a
r
i
a
b
l
e
s
t
h
a
t
a
r
e
p
r
o
c
e
s
s
e
d
.
R
e
f
e
r
t
o
t
h
e
p
r
e
d
i
c
t
i
o
n
p
e
r
i
o
d
s
i
n
T
a
b
l
e
2
.
I
t
a
p
p
e
a
r
s
t
h
a
t
w
e
e
k
l
y
p
e
r
i
o
d
p
r
e
d
i
c
t
i
o
n
s
h
a
v
e
b
e
t
t
e
r
a
c
c
u
r
a
c
y
(
8
5
.
7
1
%
)
,
g
i
v
e
n
t
h
e
v
a
r
i
a
t
i
o
n
o
f
m
e
t
e
o
r
o
l
o
g
i
c
a
l
d
a
t
a
e
v
e
r
y
d
a
y
o
f
t
h
e
w
e
e
k
i
s
n
o
t
t
o
o
l
a
r
g
e
.
M
e
a
n
w
h
i
l
e
,
t
h
e
r
e
i
s
m
o
r
e
v
a
r
i
e
t
y
o
f
d
a
i
l
y
m
e
t
e
o
r
o
l
o
g
y
i
n
a
m
o
n
t
h
s
o
t
h
a
t
w
h
e
n
c
o
m
b
i
n
e
d
i
n
t
h
e
m
o
n
t
h
m
a
k
e
s
a
h
i
g
h
e
r
d
e
v
i
a
t
i
o
n
.
H
o
w
e
v
e
r
,
t
h
e
m
o
d
e
l
u
s
e
d
t
o
p
r
e
d
i
c
t
r
a
i
n
f
o
r
a
m
o
n
t
h
i
s
m
o
r
e
r
o
b
u
s
t
w
h
e
n
a
v
a
i
l
a
b
l
e
w
e
a
t
h
e
r
d
a
t
a
i
s
i
n
c
o
m
p
l
e
t
e
.
P
r
e
c
i
p
i
t
a
t
i
o
n
p
r
e
d
i
c
t
i
o
n
d
e
p
e
n
d
s
o
n
t
h
e
t
r
a
i
n
i
n
g
d
a
t
a
s
e
q
u
e
n
c
e
,
c
o
n
f
i
g
u
r
a
t
i
o
n
,
m
e
t
h
o
d
,
a
n
d
p
e
r
i
o
d
o
f
f
o
r
e
c
a
s
t
.
T
h
e
r
e
s
u
l
t
s
o
f
t
h
e
s
t
u
d
y
,
A
d
a
m
m
o
d
e
l
,
p
r
o
v
i
d
e
d
b
e
t
t
e
r
a
c
c
u
r
a
c
y
t
h
a
n
S
G
D
,
c
o
n
s
i
d
e
r
i
n
g
A
d
a
m
m
o
d
e
l
u
s
e
d
a
g
g
r
e
g
a
t
e
d
a
t
a
i
n
t
r
a
i
n
i
n
g
.
I
n
c
o
n
t
r
a
s
t
,
S
G
D
u
s
e
d
o
n
l
y
o
n
e
o
r
s
e
v
e
r
a
l
p
a
r
t
s
r
a
n
d
o
m
l
y
s
e
l
e
c
t
e
d
,
s
o
t
h
e
p
o
s
s
i
b
i
l
i
t
y
o
f
o
c
c
u
r
r
e
n
c
e
i
s
m
i
n
i
m
u
m
l
o
c
a
l
,
a
s
a
r
e
s
u
l
t
o
f
n
o
t
r
e
p
r
e
s
e
n
t
i
n
g
a
l
l
d
a
t
a
i
n
e
a
c
h
c
l
a
s
s
.
T
h
e
w
e
e
k
l
y
f
o
r
e
c
a
s
t
o
f
f
e
r
s
h
i
g
h
e
r
e
f
f
i
c
i
e
n
c
y
t
h
a
n
m
o
n
t
h
l
y
f
i
n
d
i
n
g
t
h
a
t
i
n
t
h
a
t
p
e
r
i
o
d
,
m
e
t
e
o
r
o
l
o
g
i
c
a
l
d
a
t
a
i
s
m
o
r
e
h
o
m
o
g
e
n
e
o
u
s
t
h
a
n
m
o
n
t
h
l
y
d
a
t
a
.
T
r
a
i
n
i
n
g
d
a
t
a
u
s
e
d
f
o
r
t
e
n
y
e
a
r
s
p
r
o
v
i
d
e
s
b
e
t
t
e
r
c
o
r
r
e
c
t
n
e
s
s
,
c
o
n
s
i
d
e
r
i
n
g
i
t
p
r
o
v
i
d
e
s
m
o
r
e
v
a
r
i
a
t
i
o
n
t
h
a
n
t
h
e
a
c
c
u
r
a
c
y
o
f
t
r
a
i
n
i
n
g
d
a
t
a
f
r
o
m
t
h
e
l
a
s
t
f
i
v
e
y
e
a
r
s
.
T
h
e
b
e
s
t
a
c
c
u
r
a
c
y
w
a
s
8
5
.
7
1
%
w
i
t
h
w
e
e
k
l
y
u
s
i
n
g
R
N
N
a
n
d
L
S
T
M
o
f
t
r
a
i
n
i
n
g
d
a
t
a
f
o
r
t
e
n
y
e
a
r
s
.
T
h
i
s
r
e
s
e
a
r
c
h
c
a
n
c
o
m
p
a
r
e
w
i
t
h
M
L
P
w
i
t
h
7
5
-
92
%
[
2
1
]
a
n
d
R
N
N
w
i
t
h
h
e
u
r
i
s
t
i
c
o
p
t
i
m
i
z
e
d
o
f
5
9
-
8
4
.
6
%
[
1
5
]
.
4.
CO
NCLU
SI
O
N
I
n
th
is
wo
r
k
,
we
s
tu
d
ie
d
h
o
w
to
u
s
e
m
u
ltip
le
v
ar
ia
b
les
o
f
wea
th
er
ca
n
r
ain
f
all
p
r
ed
ictio
n
m
o
n
th
l
y
.
W
e
p
r
o
p
o
s
ed
th
e
a
d
v
an
ta
g
e
o
f
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
to
au
to
m
atica
lly
g
av
e
ac
cu
r
ac
y
8
5
.
7
1
o
f
test
d
ata.
T
h
e
r
esear
ch
s
h
o
wed
t
h
at
u
s
i
n
g
f
iv
e
y
ea
r
s
o
f
wea
t
h
e
r
d
at
a
ca
n
p
r
ed
ict
p
r
ec
ip
itatio
n
w
ee
k
ly
an
d
m
o
n
th
ly
.
W
ea
th
er
d
ata
in
th
e
la
s
t
f
iv
e
y
ea
r
s
ca
n
p
r
ed
ict
r
ain
f
all
f
o
r
a
m
o
n
th
o
f
th
e
f
o
llo
win
g
y
ea
r
.
Ho
wev
er
,
wee
k
ly
p
r
ed
ictio
n
s
h
av
e
h
ig
h
er
ac
c
u
r
ac
y
.
T
h
e
ex
p
er
im
en
tal
r
esu
lts
also
s
h
o
w
th
at
a
lar
g
e
n
u
m
b
er
o
f
d
ata
s
ets
ca
n
im
p
r
o
v
e
ac
c
u
r
ac
y
.
I
n
t
h
e
f
u
t
u
r
e,
co
m
p
a
r
ed
th
e
p
r
o
p
o
s
ed
m
eth
o
d
s
ap
p
r
o
ac
h
with
a
p
u
b
lic
wea
th
er
f
o
r
ec
ast
ce
n
ter
r
esu
lts
an
d
d
e
m
o
n
s
tr
ate
d
th
e
ef
f
ec
tiv
e
n
ess
o
f
th
e
m
o
d
el.
So
th
at
o
u
tp
u
t
p
r
e
d
ictio
n
i
n
v
alu
e,
im
p
r
o
v
in
g
th
e
o
u
tp
u
t
class
o
f
th
is
s
tu
d
y
.
C
u
r
r
en
t
u
n
ce
r
tain
r
ain
f
all
p
r
ed
ictio
n
s
d
o
n
o
t
o
n
ly
p
ay
atten
tio
n
to
p
ast
d
ata
p
atter
n
s
b
u
t a
ls
o
n
ee
d
to
co
n
s
id
er
ex
tr
em
e
p
h
e
n
o
m
en
a
s
u
ch
as E
l N
in
o
as a
d
d
itio
n
al
f
ea
tu
r
es.
RE
F
E
R
E
NC
E
S
[1
]
R.
Hid
a
y
a
t,
M
.
Ju
n
iarti,
a
n
d
U.
M
a
ru
fa
h
,
“
Im
p
a
c
t
o
f
La
Niñ
a
a
n
d
La
Niñ
a
M
o
d
o
k
i
o
n
In
d
o
n
e
sia
Ra
in
f
a
ll
Va
riab
il
it
y
,
”
Ea
rth
a
n
d
E
n
v
iro
n
me
n
ta
l
S
c
ien
c
e
,
v
o
l.
1
4
9
,
n
o
.
1
,
p
p
.
2
1
7
-
2
2
2
,
2
0
1
7
.
[2
]
S
u
p
a
ri,
F
.
Ta
n
g
a
n
g
,
E
.
S
a
li
m
u
n
,
E.
Ald
ria
n
,
A.
S
o
p
a
h
e
lu
wa
k
a
n
,
a
n
d
L.
J
u
n
e
n
g
,
“
ENS
O
M
o
d
u
lati
o
n
o
f
S
e
a
so
n
a
l
Ra
in
fa
ll
a
n
d
Ex
trem
e
s in
In
d
o
n
e
sia
,
”
Cli
ma
te Dy
n
a
mic
s
,
v
o
l.
5
1
,
n
o
.
7
,
p
p
.
1
-
2
2
,
2
0
1
7
.
[3
]
E.
M
u
ly
a
n
a
,
“
Re
latio
n
sh
i
p
b
e
twe
e
n
ENS
O
Wi
t
h
Va
riatio
n
s
Ch
e
a
p
Ra
in
i
n
In
d
o
n
e
sia
,
”
J
u
rn
a
l
S
a
i
n
s
&
T
e
k
n
o
l
o
g
i
M
o
d
if
ika
si C
u
a
c
a
,
v
o
l
.
3
,
p
p
.
1
-
4
,
2
0
0
2
.
[4
]
S
.
Cra
m
e
r,
M
.
Ka
m
p
o
u
rid
is
,
e
t
a
l.
,
“
An
Ex
ten
si
v
e
Ev
a
lu
a
ti
o
n
o
f
S
e
v
e
n
M
a
c
h
in
e
Lea
rn
i
n
g
M
e
t
h
o
d
s
fo
r
Ra
in
fa
ll
P
re
d
ictio
n
i
n
Wea
th
e
r
De
riv
a
ti
v
e
s,”
Exp
e
rt S
y
ste
ms
wit
h
Ap
p
li
c
a
ti
o
n
s
,
v
o
l.
8
5
,
p
p
.
1
6
9
-
1
8
1
,
2
0
1
7
.
[5
]
Y.
Li
u
,
E.
Ra
c
a
h
,
J.
Co
rre
a
,
e
t
a
l.
,
“
Ap
p
li
c
a
ti
o
n
o
f
De
e
p
Co
n
v
o
l
u
ti
o
n
a
l
Ne
u
ra
l
Ne
two
rk
s
fo
r
De
tec
ti
n
g
Ex
trem
e
Wea
th
e
r
in
Cli
m
a
te Da
tas
e
ts,”
In
t
e
rn
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Ad
v
a
n
c
e
s in
Bi
g
Da
t
a
A
n
a
lytics
,
p
p
.
8
1
-
88
,
2
0
1
6
.
[6
]
P
.
Z
h
a
n
g
,
L.
Zh
a
n
g
,
e
t
a
l.
,
“
A
De
e
p
-
Lea
rn
in
g
Ba
se
d
P
re
c
ip
i
tatio
n
F
o
re
c
a
stin
g
Ap
p
ro
a
c
h
Us
in
g
M
u
l
ti
p
l
e
En
v
ir
o
n
m
e
n
tal
F
a
c
to
rs,”
2
0
1
7
I
E
EE
6
th
In
ter
n
a
ti
o
n
a
l
Co
n
g
re
ss
o
n
Bi
g
Da
ta
,
Bi
g
Da
t
a
Co
n
g
re
ss
,
p
p
.
1
9
3
-
2
0
0
,
2
0
1
7
.
[7
]
N.
S
in
h
a
,
B.
P
u
r
k
a
y
a
sth
a
,
a
n
d
L.
M
a
rb
a
n
ian
g
,
“
Wea
th
e
r
P
re
d
ictio
n
b
y
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
two
rk
Dy
n
a
m
ics
,
”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
I
n
telli
g
e
n
t
En
g
i
n
e
e
rin
g
I
n
f
o
rm
a
ti
c
s
,
v
o
l
.
2
,
n
o
.
2
/3
,
p
p
.
1
7
-
8
0
,
2
0
1
4
.
[8
]
S
.
G
.
G
o
u
d
a
,
Z.
Hu
ss
e
in
,
S
.
L
u
o
,
a
n
d
Q.
Yu
a
n
,
“
M
o
d
e
l
se
lec
ti
o
n
f
o
r
a
c
c
u
ra
te
d
a
il
y
g
l
o
b
a
l
s
o
lar
ra
d
ia
ti
o
n
p
re
d
icti
o
n
in
Ch
in
a
,
”
J
o
u
rn
a
l
o
f
Clea
n
e
r P
ro
d
u
c
t
io
n
,
v
o
l.
2
2
1
,
p
p
.
1
3
2
-
1
4
4
,
2
0
1
9
.
[9
]
I.
Tan
a
k
a
a
n
d
H.
Oh
m
o
ri,
“
M
e
t
h
o
d
S
e
lec
ti
o
n
in
Diffe
re
n
t
Re
g
i
o
n
s
f
o
r
S
h
o
rt
-
Term
Wi
n
d
S
p
e
e
d
P
re
d
i
c
ti
o
n
i
n
Ja
p
a
n
,
”
S
ICE
A
n
n
u
a
l
Co
n
fer
e
n
c
e
,
v
o
l
.
2
,
p
p
.
1
8
9
-
194
,
2
0
1
5
.
[1
0
]
H.
K.
Kim
,
“
Tem
p
e
ra
tu
re
P
re
d
i
c
ti
o
n
Us
i
n
g
t
h
e
M
issi
n
g
Da
ta
R
e
fin
e
m
e
n
t
M
o
d
e
l
Ba
se
d
o
n
a
L
o
n
g
S
h
o
rt
-
Term
M
e
m
o
ry
Ne
u
ra
l
Ne
two
r
k
,
”
At
m
o
sp
h
e
re
,
v
o
l.
1
0
,
n
o
.
1
1
,
p
p
.
1
-
16
,
2
0
1
9
.
[1
1
]
F
.
Li
,
G
.
Re
n
,
a
n
d
J.
Lee
,
“
M
u
lt
i
-
ste
p
win
d
sp
e
e
d
p
re
d
icti
o
n
b
a
se
d
o
n
tu
r
b
u
le
n
c
e
in
ten
sit
y
a
n
d
h
y
b
rid
d
e
e
p
n
e
u
ra
l
n
e
two
rk
s,”
E
n
e
rg
y
Co
n
v
e
rs
io
n
a
n
d
M
a
n
a
g
e
me
n
t
,
v
o
l.
1
8
6
,
p
p
.
3
0
6
-
3
2
2
,
2
0
1
9
.
[1
2
]
S
.
M
o
o
n
,
Y.
Kim
,
Y.
He
e
,
a
n
d
B
.
M
o
o
n
,
“
Ap
p
li
c
a
ti
o
n
o
f
M
a
c
h
in
e
L
e
a
rn
in
g
t
o
a
n
Earl
y
Warn
i
n
g
S
y
ste
m
fo
r
Ve
ry
S
h
o
rt
-
Term
He
a
v
y
Ra
in
fa
ll
,
”
J
o
u
rn
a
l
o
f
Hy
d
ro
lo
g
y
,
v
o
l.
5
6
8
,
p
p
.
1
0
4
2
–
1
0
5
4
,
2
0
1
9
.
[1
3
]
S
.
Yu
a
n
,
X.
L
u
o
,
B.
M
u
,
J.
Li
,
a
n
d
G
.
Da
i,
“
P
re
d
ictio
n
o
f
N
o
rth
Atla
n
ti
c
Os
c
il
latio
n
I
n
d
e
x
wit
h
C
o
n
v
o
lu
ti
o
n
a
l
LS
TM
Ba
se
d
o
n
E
n
se
m
b
le E
m
p
iri
c
a
l
M
o
d
e
De
c
o
m
p
o
sit
io
n
,
”
At
m
o
sp
h
e
re
,
v
o
l.
1
0
,
n
o
.
2
5
2
,
p
p
.
2
-
1
3
,
2
0
1
9
.
[1
4
]
E.
P
.
P
ra
se
ty
a
a
n
d
E.
C.
Dja
m
a
l,
“
Ra
in
fa
ll
F
o
re
c
a
stin
g
fo
r
th
e
Na
t
u
r
a
l
Disa
ste
rs P
re
p
a
ra
ti
o
n
Us
in
g
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
two
rk
s,”
2
0
1
9
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
El
e
c
tri
c
a
l
E
n
g
in
e
e
ri
n
g
a
n
d
I
n
fo
rm
a
t
ics
(ICEE
I)
,
p
p
.
5
2
-
57
,
2
0
1
9
.
[1
5
]
A.
G
.
S
a
lma
n
,
B.
Ka
n
ig
o
ro
,
a
n
d
Y.
He
ry
a
d
i,
“
Wea
th
e
r
F
o
re
c
a
sti
n
g
Us
i
n
g
De
e
p
Lea
rn
in
g
Tec
h
n
i
q
u
e
s,”
in
2
0
1
5
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
A
d
v
a
n
c
e
d
C
o
mp
u
ter
S
c
ien
c
e
a
n
d
I
n
f
o
rm
a
ti
o
n
S
y
ste
ms
(ICACS
I
S
)
,
2
0
1
5
,
p
p
.
2
8
1
–
2
8
5
.
[1
6
]
M
.
Ka
n
n
a
n
,
S
.
P
ra
b
h
a
k
a
ra
n
,
a
n
d
P
.
Ra
m
a
c
h
a
n
d
ra
n
,
“
Ra
in
fa
ll
F
o
re
c
a
stin
g
Us
i
n
g
Da
ta
M
i
n
i
n
g
Tec
h
n
i
q
u
e
,
”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
E
n
g
in
e
e
rin
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l.
2
,
n
o
.
6
,
p
p
.
3
9
7
-
4
0
0
,
2
0
1
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
18
,
No
.
5
,
Octo
b
e
r
2
0
2
0
:
2
5
2
5
-
2532
2532
[1
7
]
S
.
K
.
M
o
h
a
p
a
t
r
a
,
A
.
U
p
a
d
h
y
a
y
,
a
n
d
C
.
G
o
l
a
,
“
R
a
i
n
f
a
l
l
P
r
e
d
i
c
t
i
o
n
b
a
s
e
d
o
n
1
0
0
y
e
a
r
s
o
f
M
e
t
e
o
r
o
l
o
g
i
c
a
l
D
a
t
a
,
”
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
C
o
m
p
u
t
i
n
g
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s
f
o
r
S
m
a
r
t
N
a
t
i
o
n
(
I
C
3
T
S
N
)
,
p
p
.
1
6
2
-
1
6
6
,
2
0
1
7
.
[1
8
]
C.
P
ian
i,
J.
O.
Ha
e
rter,
a
n
d
E.
Co
p
p
o
la,
“
S
tatisti
c
a
l
b
ias
c
o
rre
c
ti
o
n
fo
r
d
a
il
y
p
re
c
ip
it
a
ti
o
n
i
n
re
g
i
o
n
a
l
c
li
m
a
te m
o
d
e
ls
o
v
e
r
E
u
ro
p
e
,
”
T
h
e
o
re
ti
c
a
l
a
n
d
Ap
p
li
e
d
Cl
ima
t
o
lo
g
y
,
v
o
l.
9
9
,
n
o
.
1
–
2
,
p
p
.
1
8
7
-
1
9
2
,
2
0
1
0
.
[1
9
]
D.
Lea
ld
i,
S
.
N
u
rd
iati
,
a
n
d
A S
o
p
a
h
e
lu
wa
k
a
n
,
“
S
tatisti
c
a
l
Bias
Co
r
re
c
t
io
n
M
o
d
e
ll
in
g
fo
r
S
e
a
so
n
a
l
R
a
in
fa
ll
F
o
re
c
a
st
fo
r
th
e
Ca
se
o
f
Ba
li
Isla
n
d
,
”
J
o
u
r
n
a
l
o
f
P
h
y
sic
s
,
v
o
l.
1
0
0
8
,
n
o
.
1
,
p
p
.
1
-
1
0
,
2
0
1
8
.
[2
0
]
B.
K.
Ra
n
i
a
n
d
A.
G
o
v
a
rd
h
a
n
,
“
Ra
in
fa
ll
P
re
d
ictio
n
Us
in
g
Da
ta
M
in
in
g
Tec
h
n
iq
u
e
s
a
S
u
rv
e
y
,
”
T
h
e
S
e
c
o
n
d
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
I
n
fo
r
ma
ti
o
n
T
e
c
h
n
o
l
o
g
y
C
o
n
v
e
rg
e
n
c
e
a
n
d
S
e
rv
ice
s
,
p
p
.
2
3
-
30
,
2
0
1
3
.
[2
1
]
P
.
Zh
a
n
g
,
Y.
Jia
,
J.
G
a
o
,
W.
S
o
n
g
,
a
n
d
H.
Leu
n
g
,
“
S
h
o
rt
-
term
Ra
in
fa
ll
F
o
re
c
a
stin
g
Us
in
g
M
u
lt
i
lay
e
r
P
e
rc
e
p
tro
n
,
”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Bi
g
Da
t
a
,
v
o
l.
6
,
n
o
.
1
,
p
p
.
9
3
-
1
0
6
,
2
0
1
8
.
[2
2
]
M
.
A.
Zay
tar
a
n
d
C.
El
Am
ra
n
i
,
“
S
e
q
u
e
n
c
e
to
S
e
q
u
e
n
c
e
Wea
th
e
r
F
o
re
c
a
stin
g
wi
th
Lo
n
g
S
h
o
r
t
-
Term
M
e
m
o
ry
S
e
q
u
e
n
c
e
to
S
e
q
u
e
n
c
e
Wea
th
e
r
F
o
re
c
a
stin
g
wi
th
L
o
n
g
S
h
o
r
t
-
Term
M
e
m
o
ry
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
two
rk
s,”
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
C
o
mp
u
ter
Ap
p
l
ica
ti
o
n
s
,
v
o
l.
1
4
3
,
n
o
.
1
1
,
p
p
.
7
-
1
1
,
2
0
1
6
.
[2
3
]
B.
Zh
a
o
,
X.
Li
,
X.
L
u
,
a
n
d
Z
.
Wan
g
,
“
A
CNN
-
RNN
Arc
h
it
e
c
tu
re
fo
r
M
u
lt
i
-
Lab
e
l
Wea
th
e
r
Re
c
o
g
n
it
io
n
,
”
Ne
u
ro
c
o
mp
u
ti
n
g
,
v
o
l
.
3
2
2
,
p
p
.
4
5
-
5
7
,
2
0
1
8
.
[2
4
]
X.
S
h
i,
Z.
C
h
e
n
,
a
n
d
H.
Wan
g
,
“
Co
n
v
o
l
u
ti
o
n
a
l
LS
TM
Ne
two
r
k
:
A
M
a
c
h
in
e
Lea
rn
i
n
g
A
p
p
r
o
a
c
h
f
o
r
P
re
c
ip
it
a
ti
o
n
No
wc
a
stin
g
,
”
A
d
v
a
n
c
e
s in
Ne
u
ra
l
In
fo
rm
a
ti
o
n
Pr
o
c
e
s
sin
g
S
y
ste
ms
,
v
o
l.
2
8
,
p
p
.
8
0
2
-
8
1
0
,
2
0
1
5
.
[2
5
]
A.
M
u
l
y
a
d
i
a
n
d
E.
C.
Dja
m
a
l,
“
S
u
n
sh
i
n
e
Du
ra
ti
o
n
P
re
d
ictio
n
Us
in
g
1
D
Co
n
v
o
l
u
ti
o
n
a
l
Ne
u
ra
l
Ne
two
rk
s,”
2
0
1
9
6
t
h
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
I
n
stru
me
n
ta
t
io
n
,
Co
n
tro
l,
a
n
d
A
u
to
m
a
ti
o
n
(ICA)
,
2
0
1
9
.
[2
6
]
M
.
Qi
u
,
P
.
Z
h
a
o
,
K.
Zh
a
n
g
,
J.
Hu
a
n
g
,
X.
S
h
i,
a
n
d
X.
Wan
g
,
“
A
sh
o
rt
-
Term
Ra
in
fa
ll
P
re
d
ictio
n
M
o
d
e
l
Us
in
g
M
u
lt
i
-
tas
k
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
two
rk
s,”
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
D
a
ta
M
i
n
in
g
,
p
p
.
3
9
5
-
4
0
4
,
2
0
1
7
.
[2
7
]
F
.
J.
C
h
a
n
g
,
P
.
A.
Ch
e
n
,
Y.
R.
Lu
,
E.
Hu
a
n
g
,
a
n
d
K.
Y.
C
h
a
n
g
,
“
Re
a
l
-
ti
m
e
M
u
lt
i
-
S
tep
-
Ah
e
a
d
Wate
r
Le
v
e
l
F
o
re
c
a
stin
g
b
y
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
two
r
k
s fo
r
Urb
a
n
F
lo
o
d
C
o
n
tr
o
l,
”
J
o
u
rn
a
l
o
f
Hy
d
ro
l
o
g
y
,
v
o
l.
5
1
7
,
p
p
.
8
3
6
-
8
4
6
,
2
0
1
4
.
[2
8
]
Y.
Li
u
,
C.
G
o
n
g
,
L.
Ya
n
g
,
a
n
d
Y.
Ch
e
n
,
“
DST
P
-
RNN
:
a
d
u
a
l
-
sta
g
e
two
-
p
h
a
se
a
tt
e
n
ti
o
n
-
b
a
se
d
r
e
c
u
rre
n
t
n
e
u
ra
l
n
e
two
rk
f
o
r
l
o
n
g
-
term
a
n
d
m
u
lt
i
v
a
riate
ti
m
e
se
ries
p
re
d
ictio
n
,
”
Ex
p
e
rt S
y
ste
ms
W
it
h
Ap
p
li
c
a
ti
o
n
s
,
v
o
l.
1
4
3
,
2
0
2
0
.
[2
9
]
Ya
g
m
u
r
G
iz
e
m
Cin
a
r,
H.
M
iri
sa
e
e
a
,
P
.
G
o
sw
a
m
i,
E.
Ga
u
ss
iera
,
a
n
d
A.
-
A.
Ba
c
h
ir,
“
P
e
rio
d
-
a
wa
re
Co
n
ten
t
Atten
ti
o
n
RNN
s fo
r
Ti
m
e
S
e
ries
F
o
re
c
a
stin
g
with
M
issin
g
Va
lu
e
s,”
Ne
u
r
o
c
o
mp
u
ti
n
g
,
v
o
l.
3
1
2
,
p
p
.
1
7
7
-
1
8
6
,
2
0
1
8
.
[3
0
]
F
.
R.
Nin
g
sih
a
n
d
E.
C.
Dja
m
a
l,
“
Wi
n
d
S
p
e
e
d
F
o
re
c
a
stin
g
Us
in
g
R
e
c
u
rre
n
t
Ne
u
ra
l
Ne
two
rk
s
a
n
d
L
o
n
g
S
h
o
rt
Term
M
e
m
o
ry
,
”
6
t
h
I
n
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
In
str
u
me
n
ta
ti
o
n
,
C
o
n
tr
o
l,
a
n
d
Au
t
o
ma
ti
o
n
(ICA)
,
p
p
.
1
3
7
-
141
,
2
0
1
9
.
[3
1
]
S
.
Kim
,
S
.
Am
e
s,
J.
Lee
,
C.
Z
h
a
n
g
,
a
n
d
A
.
C.
Wi
lso
n
,
“
Re
so
lu
t
io
n
Re
c
o
n
stru
c
ti
o
n
o
f
Cli
m
a
te
D
a
ta
with
P
ix
e
l
Re
c
u
rsiv
e
M
o
d
e
l,
”
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Da
t
a
M
in
i
n
g
W
o
rk
sh
o
p
s
,
p
p
.
3
1
3
-
3
2
1
,
2
0
1
7
.
[3
2
]
P
.
A.
Ch
e
n
,
L
.
C.
Ch
a
n
g
,
a
n
d
F
.
J.
Ch
a
n
g
,
“
Re
in
fo
rc
e
d
re
c
u
rr
e
n
t
n
e
u
ra
l
n
e
two
rk
s
fo
r
m
u
l
ti
-
st
e
p
-
a
h
e
a
d
flo
o
d
fo
re
c
a
sts,”
J
o
u
rn
a
l
o
f
Hy
d
ro
lo
g
y
,
v
o
l
.
4
9
7
,
p
p
.
7
1
-
7
9
,
2
0
1
3
.
[3
3
]
Y.
Lu
o
,
“
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
two
rk
s
f
o
r
Clas
sify
in
g
Re
lati
o
n
s
in
Cl
in
ica
l
No
tes
,
”
J
o
u
r
n
a
l
o
f
Bi
o
me
d
i
c
a
l
I
n
fo
rm
a
ti
c
s
,
v
o
l.
7
2
,
p
p
.
8
5
-
9
5
,
2
0
1
7.
[3
4
]
S
.
Al
h
a
g
ry
,
A.
A.
F
a
h
m
y
,
a
n
d
R.
A.
El
-
Kh
o
rib
i,
“
Emo
ti
o
n
Re
c
o
g
n
i
ti
o
n
b
a
se
d
o
n
EE
G
u
si
n
g
LS
T
M
R
e
c
u
rre
n
t
Ne
u
ra
l
Ne
two
rk
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Ad
v
a
n
c
e
d
Co
mp
u
ter
S
c
ien
c
e
a
n
d
Ap
p
li
c
a
ti
o
n
s
,
v
o
l.
8
,
n
o
.
1
0
,
p
p
.
3
5
6
-
3
5
8
,
2
0
1
7
.
[3
5
]
S
.
P
o
o
rn
ima
a
n
d
M
.
P
u
sh
p
a
lath
a
,
“
P
re
d
ictio
n
o
f
Ra
in
fa
ll
Us
i
n
g
I
n
ten
sified
LS
T
M
Ba
se
d
Re
c
u
rre
n
t
N
e
u
ra
l
Ne
two
rk
with
Weig
h
ted
Li
n
e
a
r
Un
it
s,”
At
mo
sp
h
e
re
,
v
o
l.
1
0
,
n
o
.
1
1
,
p
p
.
1
-
1
8
,
2
0
1
9
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Mi
shk
a
Ald
i
ty
a
Pri
a
tn
a
re
c
e
iv
e
d
h
is
b
a
c
h
e
lo
r'
s
d
e
g
re
e
i
n
th
e
De
p
a
rtme
n
t
o
f
M
e
teo
ro
l
o
g
y
fr
o
m
In
stit
u
t
Tek
n
o
l
o
g
i
Ba
n
d
u
n
g
i
n
2
0
1
9
.
E
-
m
a
il
:
m
ish
k
a
.
a
ld
i
ty
a
@g
m
a
il
.
c
o
m
Esm
e
r
a
ld
a
Co
n
te
ss
a
Dja
m
a
l
re
c
e
iv
e
d
a
B
a
c
h
e
lo
r
's
d
e
g
re
e
in
En
g
in
e
e
rin
g
P
h
y
sic
s
fro
m
In
stit
u
t
Tek
n
o
lo
g
i
Ba
n
d
u
n
g
in
1
9
9
4
,
a
M
a
ste
r
's
d
e
g
re
e
in
In
stru
m
e
n
t
a
n
d
Co
n
tro
l
fr
o
m
In
stit
u
t
Tek
n
o
lo
g
i
Ba
n
d
u
n
g
i
n
1
9
9
8
.
S
i
n
c
e
Ph
.
D.
d
isse
rtati
o
n
u
n
ti
l
n
o
w
,
re
s
e
a
rc
h
o
n
EE
G
c
las
sifica
ti
o
n
a
n
d
fin
ish
e
d
d
o
c
to
ra
l
p
ro
g
ra
m
fro
m
In
stit
u
t
Tek
n
o
l
o
g
i
Ba
n
d
u
n
g
i
n
2
0
0
5
.
S
h
e
is
a
lec
tu
re
r
o
f
In
fo
rm
a
ti
c
s De
p
a
rtme
n
t,
Un
i
v
e
rsi
tas
Je
n
d
e
ra
l
Ac
h
m
a
d
Ya
n
i.
Ema
il
:
e
sm
e
ra
ld
a
.
c
o
n
tes
sa
@le
c
tu
re
.
u
n
jan
i.
a
c
.
id
Evaluation Warning : The document was created with Spire.PDF for Python.