I
n
t
e
r
n
at
ion
al
Jou
r
n
al
of
E
lec
t
r
ical
an
d
Com
p
u
t
e
r
E
n
gin
e
e
r
in
g
(
I
JE
CE
)
Vol.
15
,
No.
1
,
F
e
br
ua
r
y
20
25
,
pp.
783
~
791
I
S
S
N:
2088
-
8708
,
DO
I
:
10
.
11591/i
jec
e
.
v
15
i
1
.
pp
7
83
-
791
783
Jou
r
n
al
h
omepage
:
ht
tp:
//
ij
e
c
e
.
iaes
c
or
e
.
c
om
S
e
ason
al
a
u
t
o
-
r
e
g
r
e
ssi
ve
i
n
t
e
gr
at
e
d
m
ovi
n
g av
e
r
ag
e
w
it
h
b
id
ir
e
c
t
io
n
al
l
on
g sh
o
r
t
-
t
e
r
m
m
e
m
o
r
y f
o
r
c
o
c
on
u
t
yi
e
ld
p
r
e
d
ic
t
io
n
Nir
an
j
an
S
h
a
d
ak
s
h
ar
ap
p
a
Jayan
n
a,
Ravip
r
ak
a
s
h
M
ad
e
n
u
r
L
in
gar
aj
u
D
e
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ie
nc
e
a
nd
E
ngi
ne
e
r
in
g
, K
a
lp
a
ta
r
u I
ns
ti
tu
te
of
T
e
c
hnol
ogy, T
ip
tu
r
,
a
f
f
il
ia
te
d t
o
V
is
ve
s
va
r
a
ya
T
e
c
hn
ol
ogi
c
a
l
U
ni
ve
r
s
it
y, B
e
la
ga
vi
, I
ndi
a
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
M
a
y
27,
2024
R
e
vis
e
d
S
e
p
5,
2024
Ac
c
e
pted
Oc
t
1,
2024
Cro
p
y
i
el
d
p
red
i
ct
i
o
n
h
e
l
p
s
farmers
mak
e
i
n
fo
rm
ed
d
e
ci
s
i
o
n
s
reg
ar
d
i
n
g
t
h
e
o
p
t
i
m
al
t
i
mi
n
g
f
o
r
cr
o
p
cu
l
t
i
v
a
t
i
o
n
,
t
a
k
i
n
g
i
n
t
o
acc
o
u
n
t
e
n
v
i
ro
n
m
en
t
a
l
fa
ct
o
r
s
t
o
e
n
h
a
n
ce
p
re
d
i
c
t
i
v
e
a
cc
u
rac
y
a
n
d
m
ax
i
m
i
z
e
y
i
e
l
d
s
.
T
h
e
e
x
i
s
t
i
n
g
me
t
h
o
d
s
req
u
i
re
a
m
as
s
i
v
e
a
mo
u
n
t
o
f
d
a
t
a,
w
h
i
c
h
i
s
c
o
m
p
l
ex
t
o
a
cq
u
i
re.
T
o
o
v
erc
o
m
e
t
h
i
s
i
s
s
u
e,
t
h
i
s
p
ap
er
p
r
o
p
o
s
e
d
a
s
ea
s
o
n
al
a
u
t
o
-
re
g
r
e
s
s
i
v
e
i
n
t
e
g
r
at
ed
mo
v
i
n
g
av
e
ra
g
e
-
b
i
d
i
r
ec
t
i
o
n
a
l
l
o
n
g
s
h
o
rt
-
t
erm
m
em
o
r
y
(S
A
RIMA
-
B
i
L
ST
M)
f
o
r
co
c
o
n
u
t
y
i
e
l
d
p
r
ed
i
c
t
i
o
n
.
T
h
e
c
o
l
l
ec
t
e
d
d
at
as
et
i
s
p
re
p
r
o
c
es
s
e
d
t
h
r
o
u
g
h
a
l
a
b
e
l
en
c
o
d
er
an
d
m
i
n
-
ma
x
n
o
rma
l
i
za
t
i
o
n
i
s
em
p
l
o
y
e
d
t
o
ch
a
n
g
e
n
o
n
-
n
u
m
er
i
c
fea
t
u
re
s
i
n
t
o
n
u
m
er
i
ca
l
fea
t
u
re
s
an
d
e
n
h
an
ce
mo
d
e
l
p
er
f
o
rma
n
c
e.
T
h
e
p
re
p
r
o
c
es
s
e
d
f
ea
t
u
re
s
ar
e
s
e
l
ec
t
e
d
t
h
r
o
u
g
h
a
n
a
d
a
p
t
i
v
e
s
t
ra
t
e
g
y
-
b
a
s
e
d
w
h
a
l
e
o
p
t
i
m
i
z
at
i
o
n
a
l
g
o
ri
t
h
m
(
A
S
-
W
O
A
)
t
o
a
v
o
i
d
l
o
ca
l
o
p
t
i
ma
i
s
s
u
e
s
.
T
h
e
n
,
t
h
e
s
e
l
ec
t
e
d
fe
at
u
r
es
a
re
g
i
v
e
n
t
o
t
h
e
SA
RIM
A
-
B
i
L
S
T
M
t
o
p
re
d
i
c
t
t
h
e
c
o
c
o
n
u
t
y
i
e
l
d
s
.
T
h
e
p
r
o
p
o
s
e
d
SA
RIM
A
-
Bi
L
ST
M
i
s
ad
a
p
t
a
b
l
e
t
o
h
an
d
l
i
n
g
a
w
i
d
es
p
read
o
f
v
ari
o
u
s
s
eas
o
n
al
p
a
t
t
er
n
s
a
n
d
cap
t
u
r
es
s
p
a
t
i
a
l
feat
u
re
s
.
T
h
e
SA
RIMA
-
B
i
L
ST
M
p
erfo
rma
n
ce
i
s
es
t
i
ma
t
ed
t
h
r
o
u
g
h
t
h
e
c
o
effi
c
i
en
t
o
f
d
et
erm
i
n
a
t
i
o
n
(R2
),
mean
ab
s
o
l
u
t
e
erro
r
(MA
E
),
mean
s
q
u
are
d
erro
r
(MSE
),
an
d
ro
o
t
mean
s
q
u
are
erro
r
(RMSE
).
SA
RIMA
-
Bi
L
ST
M
at
t
ai
n
s
0
.
8
4
o
f
R2
,
0
.
0
5
6
o
f
MA
E
,
0
.
0
8
1
o
f
MS
E
,
an
d
0
.
9
0
7
o
f
RMSE
w
h
i
ch
i
s
b
et
t
er
w
h
e
n
co
mp
are
d
t
o
e
x
i
s
t
i
n
g
t
ec
h
n
i
q
u
es
l
i
k
e
mu
l
t
i
l
a
y
er
s
t
ack
e
d
en
s
emb
l
e,
co
n
v
o
l
u
t
i
o
n
a
l
n
eu
ra
l
n
e
t
w
o
rk
an
d
d
ee
p
n
eu
ra
l
n
e
t
w
o
r
k
(CN
N
-
D
N
N
)
an
d
au
t
o
reg
re
s
s
i
v
e
mo
v
i
n
g
a
v
erag
e
(
A
RIM
A
).
K
e
y
w
o
r
d
s
:
B
idi
r
e
c
ti
ona
l
long
s
hor
t
-
ter
m
memor
y
C
oc
onut
yield
pr
e
diction
L
a
be
l
e
nc
ode
r
S
e
a
s
ona
l
a
uto
-
r
e
gr
e
s
s
ive
int
e
gr
a
ted
movi
ng
a
ve
r
a
ge
W
ha
le
opti
mi
z
a
ti
on
a
lgor
it
hm
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
Nir
a
njan
S
ha
da
ks
ha
r
a
ppa
J
a
ya
nna
De
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ienc
e
a
nd
E
nginee
r
ing
,
Ka
lpata
r
u
I
ns
ti
tut
e
of
T
e
c
hnology
,
a
f
f
il
iate
d
to
Vis
ve
s
va
r
a
ya
T
e
c
hnologi
c
a
l
Unive
r
s
it
y
NH
-
206,
Vidya
Na
ga
r
,
T
ipt
ur
,
Ka
r
na
taka
572201
,
I
ndia
E
mail:
ni
r
a
njans
j555@gm
a
il
.
c
om
1.
I
NT
RODU
C
T
I
ON
T
he
c
oc
onut
is
a
s
igni
f
ica
nt
c
r
op
whic
h
plays
a
c
r
uc
ial
r
ole
in
e
c
onomi
e
s
o
f
nume
r
ous
c
ountr
ies
,
including
I
ndia,
the
P
hil
ippi
ne
s
,
a
nd
I
ndone
s
ia
[
1
]
.
Ge
ne
r
a
ll
y,
i
t
is
known
a
s
the
tr
e
e
of
he
a
ve
n,
be
c
a
us
e
a
ll
pa
r
ts
of
the
plants
a
r
e
us
e
f
ul
a
nd
the
main
s
our
c
e
of
income
f
or
f
a
r
mer
s
[
2
]
.
W
or
ldwide
,
it
is
gr
o
wn
in
93
c
ountr
ies
in
12
mi
ll
ion
he
c
tar
e
s
a
r
e
a
s
with
a
ye
a
r
ly
pr
oduc
ti
on
of
59.
98
mi
ll
ion
nu
ts
.
I
ndia
ha
s
s
e
c
ur
e
d
the
thi
r
d
pos
it
ion
globally
,
p
r
oduc
ing
a
n
i
mpr
e
s
s
ive
10.
56
mi
ll
ion
c
oc
onuts
a
nnua
ll
y
[
3
]
,
[
4]
.
Ac
c
ur
a
tely
pr
e
dicting
c
oc
onut
yields
is
c
r
uc
ial
in
mi
ti
ga
ti
ng
potential
dis
a
s
ter
s
dur
ing
dif
f
e
r
e
nt
s
tage
s
of
c
r
op
gr
owth,
im
pa
c
ti
ng
yield
leve
ls
s
igni
f
ica
ntl
y.
M
onit
or
ing
c
ons
e
c
uti
ve
ti
me
s
e
r
ies
da
ta
thr
oughout
gr
owth
pe
r
iods
is
e
s
s
e
nti
a
l
f
or
e
f
f
e
c
ti
ve
c
oc
onut
yield
pr
e
diction
[
5]
,
[
6
]
.
T
he
yield
is
a
c
omp
lex
whic
h
va
r
ies
thr
oug
h
f
a
c
tor
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
15
,
No.
1
,
F
e
br
ua
r
y
20
25
:
783
-
791
784
li
ke
va
r
iety
a
nd
a
ge
.
T
he
matu
r
e
d
nuts
a
r
e
take
n
a
s
yield
s
ince
it
c
ontains
s
oli
d
e
ndos
pe
r
m
whic
h
is
uti
li
z
e
d
f
or
yield
pr
e
diction
[
7]
,
[
8
]
.
T
e
mper
a
tur
e
is
a
s
igni
f
ica
nt
f
a
c
tor
in
a
gr
owth
o
f
c
oc
onut
nuts
a
nd
huge
tempe
r
a
tur
e
da
mage
s
the
r
oo
t
gr
owth
[
9
]
–
[
11]
.
R
e
a
l
-
ti
me
a
nd
de
e
p
lea
r
ning
(
DL
)
tec
hniques
a
r
e
e
m
ployed
to
e
nha
nc
e
the
e
a
r
ly
de
tec
ti
on
a
nd
e
nha
nc
e
the
c
oc
onut
yield
whic
h
of
f
e
r
s
be
tt
e
r
yield
pr
e
diction
[
1
2]
,
[
13]
.
R
e
mot
e
s
e
ns
ing
plays
a
pr
omi
ne
nt
r
ole
in
c
a
lcula
ti
ng
c
r
op
da
mage
s
a
nd
pr
e
dicting
c
r
op
yield
.
B
y
a
pplyi
ng
thes
e
methods
,
the
pr
oduc
ti
on
of
hybr
id
c
oc
onut
s
a
nd
loca
l
tall
a
t
dua
l
lea
f
c
li
pping
leve
ls
is
e
f
f
e
c
ti
ve
ly
pr
e
dicte
d
[
14]
–
[
16
]
.
I
niyan
a
nd
J
e
ba
kumar
[
17]
de
ve
loped
a
mut
ua
l
inf
or
mation
f
e
a
tur
e
s
e
lec
ti
on
(
M
I
F
S
)
f
or
c
r
op
yield
pr
e
diction
us
ing
mul
ti
laye
r
s
tac
ke
d
e
ns
e
mbl
e
r
e
gr
e
s
s
ion
(
M
S
E
R
)
.
T
he
de
ve
loped
model
i
ntr
ica
ted
the
pr
e
diction
p
r
oc
e
dur
e
of
c
r
op
yield
on
s
oybe
a
n
a
nd
c
r
op
whic
h
obtaine
d
be
tt
e
r
pe
r
f
or
manc
e
a
c
c
o
r
ding
to
phe
notype
f
a
c
tor
s
.
S
e
ve
r
a
l
non
-
li
ne
a
r
a
nd
li
ne
a
r
tec
hniques
a
r
e
e
mpl
oye
d
in
the
model
f
or
pr
e
dicting
c
r
op
yield.
How
e
ve
r
,
c
r
op
da
ta
ne
c
e
s
s
it
a
tes
numer
ous
pr
e
pr
oc
e
s
s
ing
to
f
it
t
ing
ba
s
e
li
ne
s
uppli
e
s
that
a
r
e
not
ini
ti
a
ted
f
or
yield
p
r
e
diction.
Oikonomi
dis
e
t
al
.
[
18]
int
r
oduc
e
d
a
hybr
id
c
onv
olut
ional
ne
ur
a
l
ne
twor
k
a
nd
de
e
p
ne
ur
a
l
ne
twor
k
(
C
NN
-
DN
N)
f
or
c
r
op
y
ield
p
r
e
diction.
I
t
wa
s
e
f
f
e
c
ti
ve
f
or
p
r
e
diction
due
to
it
s
ou
tl
ier
s
with
huge
va
r
ianc
e
in
dif
f
e
r
e
nt
a
r
e
a
s
whic
h
we
r
e
e
mpl
oye
d
to
c
r
e
a
te
yield
e
s
ti
mation.
T
he
f
e
a
tur
e
s
e
lec
ti
on
ut
il
ize
d
f
e
a
tur
e
e
nginee
r
ing
tec
hniques
whic
h
we
r
e
e
mpl
oye
d
to
o
btain
e
f
f
icie
nt
pe
r
f
or
manc
e
.
T
he
int
r
oduc
e
d
model
r
e
quir
e
d
a
s
igni
f
ica
nt
c
a
pa
bil
it
y
o
f
da
ta
a
nd
it
c
a
nnot
c
o
ns
ider
the
ti
me
-
s
e
r
ies
da
ta.
P
e
ng
e
t
al
.
[
19
]
s
ug
ge
s
ted
a
n
incor
por
a
ted
tec
hnique
on
B
iL
S
T
M
-
s
ine
c
os
in
e
a
lgor
it
hm
(
S
C
A)
f
or
s
olar
r
a
diation
pr
e
diction
.
I
nit
ially,
c
ompl
e
te
e
ns
e
mbl
e
e
mpi
r
ica
l
mode
de
c
ompos
it
i
on
thr
ough
a
da
pti
ve
nois
e
(
C
E
E
M
DA
N)
wa
s
a
ppli
e
d
f
or
pr
e
diction.
T
he
a
uto
-
c
or
r
e
lation
f
unc
ti
on
(
AC
F
)
a
nd
the
pa
r
ti
a
l
a
uto
-
c
or
r
e
lation
f
unc
ti
on
(
P
AC
F
)
we
r
e
uti
li
z
e
d
to
identi
f
y
r
a
diation
pa
tt
e
r
n
of
de
c
ompos
e
d
s
ub
-
modes
.
How
e
ve
r
,
the
de
ve
loped
model
r
e
quir
e
d
a
va
s
t
a
mount
of
c
a
ptur
e
d
da
ta
whic
h
is
c
ompl
e
x
to
obtain
.
P
r
a
s
e
r
t
a
nd
R
ungr
e
unga
nun
[
20
]
p
r
e
s
e
nted
a
c
oc
onut
yield
pr
e
diction
thr
ough
a
uto
-
r
e
gr
e
s
s
ive
int
e
gr
a
ted
movi
ng
a
ve
r
a
ge
(
AR
I
M
A)
.
I
ts
a
ut
onomous
va
r
iable
s
of
f
e
r
e
d
e
nha
nc
e
d
a
c
c
ur
a
c
y
without
a
ny
indi
vidual
va
r
iable
s
by
de
s
ignating
a
ppr
opr
iate
f
a
c
tor
s
to
pr
oduc
e
the
pr
e
diction
model.
How
e
ve
r
,
huge
a
tt
e
nti
on
wa
s
r
e
quir
e
d
to
f
a
s
c
inate
a
nd
ins
pir
e
f
a
r
mer
s
in
c
oc
onut
pr
oduc
ti
on.
Nova
r
ianto
[
21]
im
pleme
nt
e
d
a
n
a
e
r
ial
photog
r
a
phy
tec
hnique
thr
ough
dr
one
s
f
or
de
ter
mi
ning
da
ta
e
f
f
e
c
ti
ve
ne
s
s
by
int
e
gr
a
ti
ng
s
tanda
r
d
s
a
mpl
e
population
in
loca
l
tall
c
oc
o
nut.
T
he
population
de
ns
it
y
o
f
e
ve
r
y
a
r
e
a
is
e
s
tablis
he
d
thr
ough
va
r
ious
f
a
c
tor
s
a
nd
pa
lm
s
a
r
e
e
xpi
r
e
d
a
nd
th
e
r
e
s
ult
s
of
pe
s
ts
a
r
e
not
a
ppr
opr
iate
f
or
c
oc
onut
p
r
odu
c
ti
on.
T
he
e
xis
ti
ng
tec
hniques
ha
ve
li
mi
tations
s
uc
h
a
s
r
e
quir
ing
e
xtens
ive
da
ta
p
r
oc
e
s
s
ing
a
nd
una
ble
to
ha
ndle
ti
me
-
s
e
r
ies
da
ta.
T
he
s
e
methods
ne
c
e
s
s
it
a
te
s
igni
f
ica
nt
pr
e
pr
oc
e
s
s
ing
to
f
it
ba
s
e
li
ne
models
that
a
r
e
not
ini
ti
a
ted
f
or
yield
pr
e
diction
a
nd
they
s
tr
u
ggled
to
f
a
s
c
inate
a
nd
mot
ivate
f
a
r
mer
s
in
c
oc
onut
pr
oduc
ti
on.
M
or
e
ove
r
,
they
a
r
e
una
ble
to
ha
ndle
a
wide
s
pr
e
a
d
of
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tu
r
e
s
be
c
a
us
e
of
loca
l
opti
ma
is
s
ue
s
.
T
o
ove
r
c
ome
th
is
is
s
ue
,
thi
s
r
e
s
e
a
r
c
h
pr
opos
e
d
s
e
a
s
ona
l
a
uto
-
r
e
gr
e
s
s
ive
int
e
gr
a
ted
movi
ng
a
ve
r
a
ge
-
bidi
r
e
c
ti
ona
l
long
s
h
or
t
-
ter
m
memor
y
s
e
a
s
ona
l
a
uto
-
r
e
gr
e
s
s
ive
int
e
gr
a
ted
movi
ng
a
ve
r
a
ge
-
bidi
r
e
c
ti
ona
l
long
s
hor
t
-
ter
m
memor
y
(
S
AR
I
M
A
-
B
i
L
S
T
M
)
f
or
c
oc
onut
yield
pr
e
diction
whic
h
is
a
doptable
to
ha
ndle
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
s
pa
ti
a
l
f
e
a
tur
e
s
.
Addit
ionally,
a
da
pti
ve
s
tr
a
tegy
-
ba
s
e
d
wha
le
opti
mi
z
a
ti
on
a
lgo
r
it
hm
(
AS
-
W
OA
)
is
us
e
d
f
o
r
f
e
a
tur
e
s
e
lec
ti
on
pr
oc
e
s
s
whic
h
a
voids
loca
l
opti
ma
is
s
ue
s
a
nd
e
nha
n
c
e
s
the
c
onve
r
ge
nc
e
r
a
te.
T
he
pa
pe
r
c
ontr
ibut
ion
is
a
s
:
a.
T
he
da
tas
e
t
is
pr
e
p
r
oc
e
s
s
e
d
us
ing
labe
l
e
nc
oding
a
nd
mi
n
-
max
no
r
maliza
ti
on
whic
h
e
ns
ur
e
s
non
-
numer
ic
da
ta
is
tr
a
ns
f
or
med
e
f
f
e
c
ti
ve
ly
ther
e
by
c
ontr
ibut
in
g
to
e
nha
nc
e
the
model
pe
r
f
o
r
manc
e
.
T
he
labe
l
e
n
c
ode
r
is
us
e
d
to
c
onve
r
t
c
a
tegor
ica
l
f
e
a
tur
e
s
int
o
numer
ica
l
f
e
a
tur
e
s
a
nd
mi
n
-
max
nor
maliza
ti
on
is
a
ppli
e
d
to
s
c
a
le
the
f
e
a
tur
e
s
.
b.
T
he
pr
e
pr
oc
e
s
s
e
d
f
e
a
tur
e
s
a
r
e
then
s
e
lec
ted
us
ing
AS
-
W
OA
,
whic
h
e
nha
nc
e
s
the
c
onve
r
ge
nc
e
r
a
te
a
nd
e
f
f
e
c
ti
ve
ly
mi
ti
ga
tes
the
r
is
k
of
ge
tt
ing
t
r
a
ppe
d
in
l
oc
a
l
opti
ma
ther
e
by
e
ns
ur
ing
opti
mal
f
e
a
tur
e
s
e
lec
ti
on.
c.
T
he
s
e
lec
ted
f
e
a
tur
e
s
a
r
e
then
f
e
d
int
o
S
AR
I
M
A
-
B
i
L
S
T
M
model
whic
h
is
a
da
ptable
a
nd
c
a
pa
ble
of
ha
ndli
ng
a
wide
r
a
nge
of
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
e
f
f
e
c
ti
ve
ly
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
whic
h
lea
ds
a
c
c
ur
a
te
a
nd
r
e
li
a
ble
pr
e
dictions
.
T
his
r
e
s
e
a
r
c
h
pa
pe
r
is
pr
e
pa
r
e
d
a
s
s
e
c
ti
on
2
de
f
i
ne
s
pr
opos
e
d
methodology
.
S
e
c
ti
on
3
de
f
ines
the
r
e
s
ult
s
a
nd
dis
c
us
s
ion
.
T
he
c
onc
lus
ion
of
thi
s
pa
pe
r
is
given
in
s
e
c
ti
on
4.
2.
P
ROP
OS
E
D
M
E
T
HO
D
T
he
S
AR
I
M
A
-
B
iL
S
T
M
is
pr
opos
e
d
in
thi
s
manus
c
r
ipt
f
or
p
r
e
dicting
c
oc
onut
yield
in
Ke
r
a
la.
T
he
c
oll
e
c
ted
da
tas
e
t
is
pr
e
pr
oc
e
s
s
e
d
thr
ough
labe
l
e
nc
ode
r
a
nd
mi
n
-
max
no
r
maliza
ti
on
whic
h
is
e
m
ployed
to
c
ha
nge
non
-
numer
ic
f
e
a
tur
e
s
int
o
numer
ica
l
f
e
a
tur
e
s
a
nd
e
nha
nc
e
model
pe
r
f
or
manc
e
.
T
he
pr
e
p
r
oc
e
s
s
e
d
f
e
a
tur
e
s
a
r
e
s
e
lec
ted
thr
ough
AS
-
W
OA
whic
h
e
nh
a
nc
e
s
the
c
onve
r
ge
nc
e
r
a
te
a
nd
a
voids
loca
l
opti
m
a
is
s
ue
s
.
T
he
n,
the
s
e
lec
ted
f
e
a
tur
e
s
a
r
e
given
to
the
S
AR
I
M
A
-
B
i
L
S
T
M
f
or
pr
e
dicting
the
c
oc
onut
yields
.
F
igur
e
1
pr
e
s
e
nts
the
block
diagr
a
m
o
f
pr
opos
e
d
methodolo
gy.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
E
lec
&
C
omp
E
ng
I
S
S
N:
2088
-
8708
Se
as
onal
auto
-
r
e
gr
e
s
s
ive
int
e
gr
ated
mov
ing
av
e
r
a
ge
w
it
h
…
(
N
ir
anjan
Shadak
s
har
appa
J
ay
anna
)
785
F
igur
e
1.
B
lock
diagr
a
m
of
the
pr
opos
e
d
methodol
ogy
2.
1.
Dat
as
e
t
T
he
c
oc
onut
holds
a
pr
omi
ne
nt
pos
it
ion
a
s
a
c
ult
iv
a
ti
on
c
r
op
in
Ke
r
a
la
,
whic
h
c
ove
r
s
a
ppr
oxim
a
tely
39%
o
f
the
r
e
gion.
T
he
da
tas
e
t
s
our
c
e
d
f
r
om
the
de
pa
r
tm
e
nt
of
e
c
onomi
c
s
a
nd
s
tatis
ti
c
s
,
Ke
r
a
la,
s
pa
ns
f
r
om
2011
to
2021
a
nd
e
nc
ompas
s
e
s
ye
a
r
ly
c
oc
onut
yields
f
r
om
10
dis
tr
icts
:
Ala
ppuz
ha
,
Koz
hikode,
I
dukki,
Kottaya
m,
Kollam,
E
r
na
kulam
,
T
h
r
is
s
ur
,
W
a
ya
na
d,
Ka
s
a
r
a
god
a
nd
P
a
lakka
d.
T
h
is
da
tas
e
t
c
ompr
is
e
s
a
r
ound
120
r
e
c
or
ds
a
nnua
ll
y,
e
qua
ti
ng
to
a
n
a
ppr
oxim
a
te
pr
oduc
ti
on
of
8
mi
ll
ion
nuts
.
2.
2.
P
r
e
p
r
oc
e
s
s
in
g
T
he
pr
e
pr
oc
e
s
s
ing
tec
hniques
e
mpl
oye
d
in
thi
s
r
e
s
e
a
r
c
h
a
r
e
labe
l
e
nc
ode
r
a
nd
mi
n
-
max
nor
maliza
ti
on.
T
he
s
e
tec
hniques
im
pr
ove
d
the
da
ta
qua
li
ty
whic
h
is
e
xplaine
d
in
the
f
oll
owing
s
ub
-
s
e
c
ti
ons
.
L
a
be
l
e
nc
oding
tr
a
ns
f
or
ms
non
-
numer
ic
f
e
a
tur
e
s
i
nto
numer
ica
l
one
s
,
f
a
c
il
it
a
ti
ng
the
model's
a
bil
it
y
to
lea
r
n
f
r
om
the
da
ta
.
W
he
n
de
a
li
ng
with
da
tas
e
ts
,
c
e
r
tain
f
e
a
tur
e
s
may
no
t
be
in
the
a
ppr
opr
iate
f
o
r
mat,
ne
c
e
s
s
it
a
ti
ng
their
c
onve
r
s
ion
int
o
numer
ica
l
va
lues
thr
ough
labe
l
e
nc
oding.
T
his
pr
oc
e
s
s
r
e
duc
e
s
da
ta
volum
e
by
e
nc
oding
numer
ic
va
lues
,
r
e
s
ult
ing
i
n
e
f
f
icie
nt
memo
r
y
us
a
ge
.
T
he
mi
n
-
max
nor
mali
z
a
ti
on
is
uti
li
z
e
d
f
or
c
ha
nging
r
a
w
da
ta
int
o
a
s
tanda
r
dize
d
f
or
mat
whic
h
e
nha
nc
e
s
the
model
pe
r
f
o
r
man
c
e
.
T
he
maximum
s
c
or
e
of
the
f
e
a
tur
e
s
is
c
ha
nge
d
to
1,
th
e
mi
nim
um
s
c
or
e
o
f
the
f
e
a
tur
e
s
is
c
ha
nge
d
to
0
a
nd
other
s
c
or
e
s
of
f
e
a
tur
e
s
a
r
e
c
ha
nge
d
be
twe
e
n
0
a
nd
1
[
2
2]
.
I
t
is
e
xpr
e
s
s
e
d
in
(
1)
,
′
is
a
nor
malize
d
s
c
or
e
,
i
s
a
n
a
c
tual
va
lue,
(
)
a
nd
(
)
is
a
maximum
a
nd
mi
nim
um
s
c
or
e
of
y.
y
′
=
y
−
m
i
n
(
y
)
m
a
x
(
xy
)
−
m
i
n
(
y
)
(
1)
2.
3.
Adap
t
ive
s
t
r
at
e
gy
-
b
as
e
d
w
h
ale
op
t
im
izat
io
n
alg
or
it
h
m
f
or
f
e
at
u
r
e
s
e
lec
t
ion
T
he
wha
le
op
ti
mi
z
a
ti
on
a
lgo
r
it
hm
(
W
OA
)
is
e
mpl
oye
d
f
or
f
e
a
tur
e
s
e
lec
ti
on
o
f
c
oc
onut
yield
pr
e
diction
whic
h
a
voids
loca
l
opti
ma
is
s
ue
s
.
I
t
is
a
population
-
ba
s
e
d
a
lgor
it
hm
ins
pir
e
d
by
wha
le’
s
bubble
-
ne
t
f
e
e
ding.
W
OA
c
ompr
is
e
s
thr
e
e
s
tage
s
e
nc
ir
c
li
ng,
e
xploi
tation
a
nd
e
xplor
a
ti
on
whic
h
de
notes
the
wha
les
'
hunti
ng
pr
oc
e
dur
e
.
T
he
e
nc
ir
c
li
ng
p
r
e
y
identif
i
e
s
pr
e
y
pos
it
ions
a
nd
e
nc
ir
c
les
the
tar
ge
ts
,
e
xp
loi
tation
de
notes
s
pir
a
l
a
tt
a
c
k,
a
nd
e
xplor
a
ti
on
de
notes
r
a
nd
om
s
e
a
r
c
h
pr
e
y.
2.
3.
1.
E
n
c
irc
li
n
g
p
r
e
y
T
his
pha
s
e
is
us
e
d
to
identif
y
the
pr
e
y
pos
it
io
ns
a
nd
it
de
s
ignate
s
s
e
a
r
c
h
a
ge
nts
a
c
c
or
ding
to
dif
f
e
r
e
nt
wha
le
dis
tanc
e
s
f
r
om
pr
e
y.
Af
ter
identif
ying
the
s
e
a
r
c
h
a
ge
nt,
e
ve
r
y
wha
le
upda
tes
the
p
os
it
ions
a
c
c
or
ding
to
the
s
e
a
r
c
h
a
ge
nt.
T
he
op
ti
mum
p
r
e
y
pos
it
ions
a
r
e
unidentif
ied
a
t
the
in
it
ial
opti
mum
s
olut
ion
c
los
e
r
to
the
pr
oba
ble
s
olut
ion
is
c
ons
ider
e
d
th
r
ou
gh
W
OA
.
A
f
ter
e
s
ti
mating
the
be
s
t
s
olut
ions
,
s
e
a
r
c
h
a
ge
nts
tr
y
to
upda
te
their
pos
it
ions
to
the
be
s
t
s
e
a
r
c
h
a
ge
nt
f
or
a
tt
a
ini
ng
the
c
ur
r
e
nt
be
s
t
loca
ti
ons
that
a
r
e
e
xpr
e
s
s
e
d
in
(
2
)
to
(
5)
,
D
⃗
⃗
=
β
∙
X
∗
⃗
⃗
⃗
⃗
(
t
)
−
X
⃗
⃗
(
t
)
(
2)
X
⃗
⃗
(
t
+
1
)
=
X
∗
⃗
⃗
⃗
⃗
(
t
)
−
α
∙
D
⃗
⃗
(
3)
=
2
∙
∙
−
(
4)
=
2
∙
(
5)
w
he
r
e
,
⃗
⃗
de
notes
pr
e
s
e
nt
pos
it
ion
ve
c
tor
o
f
opti
mal
metr
ics
,
t
is
a
pr
e
s
e
nt
it
e
r
a
ti
on
,
∗
⃗
⃗
⃗
⃗
(
)
is
a
pr
e
s
e
nt
be
s
t
a
r
r
a
nge
ment
loca
ti
on
ve
c
tor
a
t
t
th
it
e
r
a
ti
on.
(
)
de
notes
pos
it
ion
of
s
e
a
r
c
h
a
ge
nt,
is
r
a
nge
s
a
mon
g
[
−
,
]
,
he
r
e
m
r
e
duc
e
s
li
ne
a
r
ly
f
r
om
2
to
0
ove
r
the
whole
e
xploi
tation
a
nd
e
xplor
a
ti
on
pha
s
e
s
.
T
he
m
is
e
s
ti
mate
d
by
=
2
−
2
∗
⁄
whic
h
s
tays
s
im
il
a
r
thr
ough
whol
e
pr
oc
e
s
s
.
T
he
n
is
r
a
ndom
number
withi
n
[
0
,
1
]
,
is
a
maximum
it
e
r
a
ti
on.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
15
,
No.
1
,
F
e
br
ua
r
y
20
25
:
783
-
791
786
2.
3.
2.
E
xp
loi
t
a
t
ion
s
t
age
T
his
pha
s
e
c
ontains
two
va
r
ious
mec
ha
ni
s
ms
s
uc
h
a
s
s
hr
inki
ng
e
nr
iching
pr
oc
e
s
s
a
nd
s
pir
a
l
upda
ti
ng
pos
it
ion
[
23]
whic
h
is
e
xplaine
d
in
be
ll
o
w:
I
n
the
s
hr
ink
ing
e
n
r
iching
p
r
oc
e
s
s
,
s
e
a
r
c
h
a
ge
nt
pos
it
ion
upda
te
is
c
las
s
if
ied
with
s
e
a
r
c
h
s
pa
c
e
by
be
s
t
s
ol
uti
on
of
a
ge
nt.
I
n
s
pir
a
l
upda
ti
ng
pos
it
ion,
c
a
ndi
da
te
ne
w
pos
it
ion
is
e
va
luate
d.
T
he
goa
l
of
the
s
hr
inki
ng
e
nr
iching
mec
ha
nis
m
is
to
mi
nim
ize
the
e
va
luate
d
va
l
ue
of
m
whic
h
or
igi
na
tes
the
humpbac
k
wha
le
be
ha
vior
.
T
he
s
e
a
r
c
h
a
ge
nt
pos
it
ion
upda
te
is
c
a
tegor
ize
d
withi
n
the
s
e
a
r
c
h
s
pa
c
e
li
mi
ted
th
r
ough
the
be
s
t
s
olut
ion
of
t
he
a
ge
nt
whic
h
is
de
s
ignate
d
thr
ough
α
r
a
ndom
va
l
ue
s
.
T
he
c
a
ndidate
's
ne
w
pos
it
ions
a
r
e
e
va
luate
d
by
uti
li
z
ing
s
pir
a
l
upda
ti
ng
pos
it
ion
mec
ha
nis
m.
T
he
s
hr
inki
ng
e
nr
iching
mec
ha
nis
m
a
nd
s
pir
a
l
upda
ti
ng
pos
it
ion
a
r
e
e
xpr
e
s
s
e
d
in
(
6
)
to
(
9)
,
D
⃗
⃗
=
X
∗
⃗
⃗
⃗
⃗
(
t
)
−
X
⃗
⃗
(
t
)
(
6)
X
⃗
⃗
(
t
+
1
)
=
D
⃗
⃗
∙
e
wl
∙
co
s
(
2π
l
)
+
X
∗
⃗
⃗
⃗
⃗
(
t
)
(
7)
(
+
1
)
=
∗
⃗
⃗
⃗
⃗
(
)
−
∙
⃗
⃗
,
<
0
.
5
(
8)
(
+
1
)
=
⃗
⃗
∙
∙
co
s
(
2
)
+
∗
⃗
⃗
⃗
⃗
(
)
≥
0
.
5
(
9)
h
e
r
e
,
⃗
⃗
is
highes
t
dis
tanc
e
a
mong
wha
le
a
nd
pr
e
y,
de
notes
c
ons
tant
whic
h
e
xplains
ge
ometr
y
of
logar
it
hmi
c
s
pir
a
l
,
a
nd
is
p
r
oduc
e
d
withi
n
the
r
a
n
ge
of
[
−
1
,
1
]
c
or
r
e
s
pondingl
y.
T
he
de
c
is
ion
a
bout
s
pe
c
if
i
c
pr
oc
e
s
s
is
de
s
ignate
d
thr
ough
a
r
a
ndom
number
of
p
∈
[
0
,
1
]
f
oc
us
e
s
on
unif
or
m
dis
tr
ibut
ion.
I
f
p
<
0
.
5
,
the
a
ge
nts
c
onti
nue
to
the
lea
de
r
a
c
c
or
ding
to
the
s
hr
i
nking
e
nc
ir
c
li
ng
p
r
oc
e
s
s
.
I
f
p
≥
0
.
5
,
s
e
a
r
c
h
a
ge
nt
pos
it
io
n
is
moder
nize
d
ba
s
e
d
on
s
pir
a
l
upda
ti
ng
pos
it
ion.
2.
3.
3.
E
xp
lorat
ion
s
t
age
I
n
thi
s
s
tage
,
the
wha
les
r
a
ndoml
y
s
e
a
r
c
h
a
c
c
or
din
g
to
their
pos
it
ions
a
nd
α
is
de
s
igned
a
s
a
r
a
ndom
va
lue
e
it
he
r
les
s
e
r
than
1
or
gr
e
a
ter
than
-
1.
T
he
α
≥
1
c
ondit
ion
ha
s
ha
ppe
ne
d,
a
nd
e
xplor
a
ti
on
is
f
or
c
e
d
on
humpbac
k
wha
les
to
global
opti
mum
a
nd
r
e
mov
e
loca
l
mi
nim
a
.
T
he
e
xplor
a
ti
on
is
e
xp
r
e
s
s
e
d
in
(
10)
a
nd
(
11)
,
D
⃗
⃗
de
notes
dis
tanc
e
be
twe
e
n
i
th
wha
le
a
nd
p
r
e
y,
X
r
a
nd
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
de
notes
r
a
ndom
loca
ti
on
ve
c
tor
.
T
he
a
r
bit
r
a
r
il
y
s
or
ted
wha
le
f
r
om
pr
e
s
e
ntl
y
c
ons
ider
e
d
c
omm
uni
ty.
T
he
a
da
pti
ve
s
tr
a
tegy
-
ba
s
e
d
W
OA
is
e
xplaine
d
in
the
f
oll
owing
s
e
c
ti
on.
⃗
⃗
=
∙
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
−
(
10)
(
+
1
)
=
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
(
)
−
∙
⃗
⃗
(
11)
2.
3.
4.
Adap
t
ive
s
t
r
at
e
gy
-
b
as
e
d
W
OA
T
he
population
of
indi
vidual
diver
s
it
y
will
be
r
e
duc
e
d
whe
n
the
a
lgor
it
hm
is
it
e
r
a
ted
to
the
f
ur
ther
s
tage
s
.
T
he
de
duc
ti
on
c
a
us
e
s
indi
viduals
to
s
e
a
r
c
h
s
pa
c
e
f
or
de
c
a
y
in
the
pos
it
ion
of
loca
l
opti
mum
.
Dur
ing
thi
s
ti
me,
a
s
mall
number
of
wha
le
populatio
ns
a
r
e
s
e
a
r
c
he
d
c
onti
nuous
ly
a
nd
it
output
s
pr
e
matur
e
c
onve
r
ge
nc
e
a
nd
e
a
s
il
y
f
a
ll
s
int
o
loca
l
opti
mu
m.
S
o
,
a
n
e
f
f
icie
nt
ba
lanc
e
of
gr
owth
a
nd
e
xplor
a
ti
on
c
a
pa
bil
it
ies
of
the
a
lgor
i
thm
is
to
e
nha
nc
e
s
e
a
r
c
hing
a
bil
it
ies
.
T
he
a
da
pti
ve
ine
r
ti
a
we
ight
s
tr
a
tegy
is
int
r
oduc
e
d
int
o
tr
a
dit
ional
W
OA
to
ba
lanc
e
the
lo
c
a
l
a
nd
global
s
e
a
r
c
h
c
a
pa
bil
it
ies
of
W
OA
whic
h
e
nr
iche
s
population
diver
s
it
y.
I
t
e
na
bles
the
a
lgor
it
hm
to
maintain
a
s
pe
c
if
ic
int
e
ns
it
y
of
the
s
e
a
r
c
h
s
tate
i
n
f
ur
ther
it
e
r
a
ti
on
s
tage
s
whic
h
a
voids
loca
l
opti
ma
is
s
ue
s
a
nd
pr
e
matur
e
c
onve
r
ge
nc
e
.
T
he
iner
ti
a
upda
te
s
t
r
a
tegy
is
e
xpr
e
s
s
e
d
in
(
12)
a
nd
(
13)
,
T
he
ω
is
mathe
matica
ll
y
e
xpr
e
s
s
e
d
in
(
14
)
to
(
16)
,
(
+
1
)
=
{
∙
∗
⃗
⃗
⃗
⃗
(
)
−
∙
⃗
⃗
,
<
0
.
5
∙
⃗
⃗
∙
∙
co
s
(
2
)
+
∗
⃗
⃗
⃗
⃗
(
)
≥
0
.
5
(
12)
(
+
1
)
=
∙
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
(
)
−
∙
⃗
⃗
(
13)
=
×
√
×
(
1
−
)
(
14)
=
√
+
2
(
15)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
E
lec
&
C
omp
E
ng
I
S
S
N:
2088
-
8708
Se
as
onal
auto
-
r
e
gr
e
s
s
ive
int
e
gr
ated
mov
ing
av
e
r
a
ge
w
it
h
…
(
N
ir
anjan
Shadak
s
har
appa
J
ay
anna
)
787
=
∑
(
,
−
,
)
2
=
1
(
16)
w
he
r
e
,
ω
is
a
n
a
da
pti
ve
iner
ti
a
we
ight
that
is
a
djus
ted
thr
ough
the
number
o
f
i
ter
a
ti
ons
.
C
=
0
.
1
,
a
nd
a
r
e
pr
e
s
e
nt
a
nd
maximum
it
e
r
a
ti
ons
,
is
many
di
mens
ions
.
T
he
√
a
nd
plays
a
major
r
ole
in
the
ini
t
ial
s
tage
of
the
i
ter
a
ti
on
pr
oc
e
s
s
.
T
his
pr
oc
e
s
s
e
na
bl
e
s
the
indi
vidual
populations
to
e
s
c
a
pe
f
r
om
thei
r
p
r
e
s
e
nt
pos
it
ions
a
nd
e
xplor
a
ti
on
with
lar
ge
s
e
a
r
c
he
s
to
e
nha
nc
e
the
a
lgor
it
hm
c
a
pa
bil
it
ies
to
a
void
loca
l
opti
ma
is
s
ue
s
.
2.
4
.
P
r
e
d
ict
ion
T
he
c
oc
onut
yield
is
p
r
e
dicte
d
th
r
ough
s
e
a
s
ona
l
a
uto
-
r
e
gr
e
s
s
ive
int
e
gr
a
ted
movi
ng
a
ve
r
a
ge
(
S
AR
I
M
A
)
with
the
B
iL
S
T
M
model
whic
h
is
a
da
ptable
to
ha
ndli
ng
a
wide
s
pr
e
a
d
of
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
.
T
he
B
iL
S
T
M
pr
oc
e
s
s
e
s
the
da
ta
int
o
dua
l
ne
twor
ks
s
uc
h
a
s
f
or
wa
r
d
a
nd
ba
c
kwa
r
d
long
s
h
or
t
-
ter
m
memor
y
(
L
T
S
M
)
a
nd
the
r
e
s
ult
o
f
thes
e
ne
twor
ks
a
r
e
int
e
gr
a
ted
a
t
e
ve
r
y
ti
me.
T
he
B
iL
S
T
M
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
a
nd
bidi
r
e
c
ti
ona
l
ti
me
de
pe
nde
nc
ies
f
r
om
his
tor
ica
l
da
ta
whic
h
im
p
r
ove
s
the
model
pe
r
f
or
ma
nc
e
.
2.
4.
1
.
S
e
as
on
al
AR
I
M
A
T
he
AR
I
M
A
is
a
ti
me
s
e
r
ies
f
or
e
c
a
s
ti
ng
tec
hnique
us
e
d
f
or
c
ha
nging
s
tationar
y
f
r
om
non
-
s
tationar
y
ti
me
s
e
r
ies
da
ta
thr
ough
a
pplyi
ng
va
r
ianc
e
s
.
I
t
de
ter
mi
ne
s
c
ur
r
e
nt
ti
me
-
s
e
r
ies
va
lue
s
thr
ough
pr
e
vious
ly
pr
e
dicte
d
e
r
r
o
r
s
a
nd
va
lues
.
S
im
il
a
r
ly,
the
S
AR
I
M
A
e
mpl
oye
d
pa
s
t
va
lues
but
c
ons
ider
e
d
s
e
a
s
ona
li
ty
a
s
a
pa
r
a
mete
r
.
T
he
im
pr
ovis
a
ti
on
of
the
AR
I
M
A
is
a
s
e
a
s
ona
l
AR
I
M
A
(
S
AR
I
M
A)
.
T
he
S
AR
I
M
A
is
e
xpr
e
s
s
e
d
in
(
17)
.
S
A
RI
M
A
=
A
RI
M
A
(
p
,
d
,
q
)
(
P
,
D
,
Q
)
s
(
17)
He
r
e
,
a
nd
de
notes
a
utor
e
gr
e
s
s
ive
a
nd
s
e
a
s
ona
l
a
utor
e
gr
e
s
s
ive
or
de
r
,
a
nd
de
notes
dif
f
e
r
e
nc
e
a
nd
s
e
a
s
ona
l
dif
f
e
r
e
nc
e
or
de
r
,
a
nd
de
notes
movi
n
g
a
ve
r
a
ge
a
nd
s
e
a
s
ona
l
movi
ng
a
ve
r
a
ge
or
de
r
,
is
a
s
e
a
s
ona
l
length
in
a
da
ta.
T
he
S
AR
I
M
A
c
a
ptur
e
s
both
s
e
a
s
on
a
l
a
nd
non
-
s
e
a
s
ona
l
pa
tt
e
r
n
s
in
da
ta
whic
h
c
r
e
a
tes
f
or
e
c
a
s
ted
model.
T
he
S
AR
I
M
A
is
a
d
a
pta
ble
to
ha
ndli
ng
a
wide
s
pr
e
a
d
of
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
.
T
he
c
e
r
tain
modeling
pr
oc
e
dur
e
of
S
AR
I
M
A
is
de
f
ined:
a.
S
tationar
it
y
tes
t:
T
he
a
ugmente
d
dicke
y
-
f
ull
e
r
(
AD
F
)
a
nd
a
uto
-
c
or
r
e
lation
f
unc
ti
on
(
AC
F
)
a
r
e
uti
li
z
e
d
to
f
ind
whe
ther
the
ti
me
-
s
e
r
ies
da
ta
wa
s
s
table
or
not.
I
f
it
is
not
s
table
,
then
the
d
a
nd
D
-
or
de
r
dif
f
e
r
e
nc
e
s
a
r
e
a
ppli
e
d.
b.
L
jung
-
B
ox
te
s
t:
I
t
is
a
c
c
ompl
is
he
d
on
a
s
e
que
n
c
e
,
if
the
p
s
c
or
e
is
les
s
e
r
than
the
s
igni
f
ica
nc
e
leve
l,
the
s
e
que
nc
e
is
c
ons
tant.
I
f
the
s
e
que
nc
e
is
not
c
ons
tan
t,
the
modeling
is
c
onti
nue
d
.
c.
M
ode
l
identif
ica
ti
on
a
nd
o
r
de
r
de
ter
mi
na
t
ion:
T
o
f
it
the
S
AR
I
M
A,
the
P
ython
gr
id
s
e
a
r
c
h
is
e
mpl
oye
d.
d.
M
ode
l
s
e
lec
ti
on:
T
he
opti
mum
model
is
s
e
lec
ted
u
s
ing
the
lea
s
t
Aka
ike
inf
or
mation
c
r
it
e
r
ion
(
A
I
C
)
.
e.
M
ode
l
tes
t:
B
y
uti
li
z
ing
the
r
e
s
idual
white
nois
e
te
s
t,
the
model
s
uc
c
e
s
s
f
it
ti
ng
is
de
ter
mi
ne
d.
f.
P
r
e
diction:
T
he
c
ons
tr
uc
ted
model
is
uti
li
z
e
d
f
o
r
p
r
e
diction.
2.
4.
2
.
B
id
ire
c
t
ion
al
lon
g
s
h
or
t
-
t
e
r
m
m
e
m
or
y
T
he
L
S
T
M
s
tor
e
s
the
input
da
ta
in
a
hidden
laye
r
including
the
ti
me
s
e
r
ies
c
onc
e
pt.
T
hr
ough
thi
s
,
the
input
da
ta
is
s
tac
ke
d
in
a
ti
me
s
e
que
n
c
e
a
t
the
hidden
laye
r
,
a
nd
the
ne
w
input
da
ta
is
r
e
pli
c
a
ted
in
the
r
e
s
ult
.
T
he
B
iL
S
T
M
is
a
n
im
p
r
ove
d
ve
r
s
ion
of
c
onve
nti
ona
l
L
S
T
M
whic
h
take
s
pa
s
t
a
nd
upc
omi
ng
s
tate
s
to
e
nha
nc
e
the
pr
e
diction
pe
r
f
or
manc
e
.
T
he
B
iL
S
T
M
pr
oc
e
s
s
e
s
the
da
ta
int
o
dua
l
ne
twor
ks
s
uc
h
a
s
f
o
r
wa
r
d
a
nd
ba
c
kwa
r
d
L
S
T
M
[
24]
a
nd
the
r
e
s
ult
o
f
thes
e
ne
t
wor
ks
a
r
e
int
e
gr
a
ted
a
t
e
ve
r
y
ti
me.
F
igu
r
e
2
p
r
e
s
e
nts
the
a
r
c
hit
e
c
tur
e
of
B
iL
S
T
M
.
I
n
a
f
or
wa
r
d
laye
r
,
it
s
e
va
luation
is
a
c
c
ompl
is
he
d
f
r
o
m
t
im
e
[
1
,
t
]
a
nd
it
s
output
is
a
tt
a
ined
a
nd
s
a
ve
d
a
t
e
ve
r
y
ti
me.
I
n
the
ba
c
kgr
ound
laye
r
,
th
e
e
va
luation
is
r
e
ve
r
s
e
d
ove
r
ti
me
[
,
1
]
a
nd
it
s
outpu
t
is
a
tt
a
ined
a
nd
s
a
ve
d
a
t
e
ve
r
y
ti
me.
L
a
s
tl
y,
a
t
e
ve
r
y
mom
e
nt,
the
f
inal
r
e
s
ult
is
a
tt
a
ined
th
r
ough
int
e
gr
a
ti
ng
output
f
or
the
r
e
s
pe
c
ti
ve
ti
me
of
the
f
or
wa
r
d
a
nd
b
a
c
kwa
r
d
laye
r
s
.
T
he
B
iL
S
T
M
is
mathe
matica
ll
y
e
xpr
e
s
s
e
d
in
(
18
)
to
(
20)
.
h
⃗
t
=
ta
nh
(
W
x
h
⃗
⃗
x
t
+
W
h
⃗
⃗
h
⃗
⃗
+
b
h
⃗
⃗
)
(
18
)
ℎ
⃖
⃗
=
ℎ
(
ℎ
⃖
⃗
⃗
+
ℎ
⃖
⃗
⃗
ℎ
⃖
⃗
⃗
+
ℎ
⃖
⃗
⃗
)
(
19)
y
t
=
W
h
⃗
⃗
h
⃗
t
+
W
h
⃖
⃗
⃗
y
h
⃖
⃗
+
b
y
(
20)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
15
,
No.
1
,
F
e
br
ua
r
y
20
25
:
783
-
791
788
whe
r
e
,
ℎ
⃗
a
nd
ℎ
⃖
⃗
a
r
e
f
or
wa
r
d
a
nd
ba
c
kwa
r
d
hidden
laye
r
output
s
whic
h
a
r
e
p
r
e
s
e
nted
thr
ough
s
upe
r
s
c
r
ipt
s
(
.
)
a
nd
(
.
⃖
)
c
or
r
e
s
pondingl
y.
T
he
ℎ
⃗
⃗
,
ℎ
⃗
⃗
ℎ
⃗
⃗
a
nd
ℎ
⃖
⃗
⃗
a
r
e
we
ight
mat
r
i
c
e
s
of
input
,
hidden
a
nd
ou
tput
laye
r
s
.
T
he
,
ℎ
⃗
⃗
a
nd
ℎ
⃖
⃗
⃗
a
r
e
output
bias
ve
c
tor
s
of
input
,
hidde
n
a
nd
outpu
t
laye
r
s
c
or
r
e
s
pondingl
y.
T
he
B
iL
S
T
M
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
a
nd
bidi
r
e
c
ti
ona
l
ti
me
de
pe
nde
nc
ies
f
r
om
his
tor
ica
l
da
ta
whic
h
im
p
r
ove
s
t
he
model
pe
r
f
or
manc
e
.
F
igur
e
2.
Ar
c
hit
e
c
tur
e
of
B
iL
S
T
M
3.
RE
S
UL
T
S
AN
D
DI
S
CU
S
S
I
ON
T
he
S
AR
I
M
A
-
B
iL
S
T
M
is
e
va
luate
d
thr
ough
P
yt
hon
3.
10
wi
th
R
AM
8
GB
,
int
e
l
c
or
e
i5
a
nd
OS
windows
10.
T
he
pr
opos
e
d
S
AR
I
M
A
-
B
iL
S
T
M
i
s
a
da
ptable
to
ha
ndli
ng
a
wide
s
pr
e
a
d
of
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tu
r
e
s
.
T
he
p
r
opos
e
d
S
AR
I
M
A
-
B
i
L
S
T
M
is
a
da
ptable
to
ha
ndli
ng
a
wide
s
pr
e
a
d
of
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
c
a
ptu
r
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
.
T
he
met
r
ics
li
ke
R
2,
M
AE
,
M
S
E
,
a
nd
R
M
S
E
a
r
e
e
mpl
oye
d
f
o
r
e
va
luating
model
pe
r
f
or
m
a
nc
e
[
25]
.
3.
1.
Qu
an
t
it
a
t
ive
an
d
q
u
al
it
at
ive
an
alys
is
T
he
qua
nti
tative
a
nd
qua
li
tative
a
na
lys
is
of
the
p
r
opos
e
d
S
AR
I
M
A
-
B
iL
S
T
M
is
a
s
s
e
s
s
e
d
with
R
2,
M
AE
,
M
S
E
,
a
nd
R
M
S
E
met
r
ics
.
T
he
B
iL
S
T
M
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
a
nd
bidi
r
e
c
ti
ona
l
ti
me
de
pe
nde
nc
ies
f
r
om
his
tor
ica
l
da
ta
whic
h
im
pr
ove
s
the
model
pe
r
f
or
manc
e
.
T
he
pr
opos
e
d
S
AR
I
M
A
-
B
iL
S
T
M
is
a
da
ptable
to
ha
ndli
ng
a
wide
s
pr
e
a
d
of
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
.
T
a
ble
1
pr
e
s
e
nts
the
AS
-
W
OA
pe
r
f
or
manc
e
a
nd
T
a
ble
2
pr
e
s
e
nts
the
S
AR
I
M
A
-
B
iL
S
T
M
pe
r
f
or
manc
e
.
T
a
ble
1.
P
e
r
f
o
r
manc
e
of
AS
-
W
OA
M
e
th
od
R2
M
A
E
M
S
E
R
M
S
E
PSO
0.93
0.089
0.098
0.983
G
W
O
0.90
0.085
0.093
0.957
C
S
O
0.88
0.078
0.089
0.931
W
O
A
0.86
0.073
0.084
0.926
AS
-
W
O
A
0.84
0.056
0.081
0.907
T
a
ble
2.
P
e
r
f
o
r
manc
e
of
S
AR
I
M
A
-
B
iL
S
T
M
mode
l
M
e
th
od
R2
M
A
E
M
S
E
R
M
S
E
C
N
N
0.92
0.092
0.092
0.957
R
N
N
0.89
0.086
0.089
0.949
L
S
T
M
0.87
0.081
0.086
0.926
B
iL
S
T
M
0.85
0.075
0.083
0.915
S
A
R
I
M
A
-
B
iL
S
T
M
0.84
0.056
0.081
0.907
T
a
ble
1
p
r
e
s
e
nts
the
pe
r
f
o
r
manc
e
of
AS
-
W
OA
thr
ough
R
2,
M
AE
,
M
S
E
a
nd
R
M
S
E
.
T
he
pe
r
f
or
manc
e
of
pa
r
ti
c
le
s
wa
r
m
opt
im
iza
ti
on
(
P
S
O
)
,
g
r
e
y
wolf
opti
mi
z
a
ti
on
(
GW
O)
,
c
a
t
s
wa
r
m
opti
mi
z
a
ti
on
(
C
S
O)
a
nd
W
OA
a
r
e
matc
he
d
with
A
S
-
W
OA
.
T
h
e
AS
-
W
OA
a
tt
a
ins
be
tt
e
r
r
e
s
ult
s
thr
ough
R
2,
M
A
E
,
M
S
E
,
a
nd
R
M
S
E
va
lues
of
a
bout
0.
84,
0.
056
,
0
.
08
1,
a
nd
0.
907
a
ppr
op
r
iate
ly
whe
n
c
ompar
e
d
to
e
xis
ti
ng
a
lgor
it
hms
.
T
a
ble
2
pr
e
s
e
nts
the
pe
r
f
or
manc
e
of
S
AR
I
M
A
-
B
i
L
S
T
M
thr
ough
R
2,
M
AE
,
M
S
E
,
a
nd
R
M
S
E
.
T
he
c
onvolut
ional
ne
ur
a
l
ne
two
r
k
(
C
NN
)
,
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
(
R
NN
)
,
L
S
T
M
,
a
nd
B
iL
S
T
M
pe
r
f
or
manc
e
a
r
e
matc
he
d
with
S
AR
I
M
A
-
B
iL
S
T
M
pe
r
f
o
r
manc
e
.
T
he
S
AR
I
M
A
-
B
i
L
S
T
M
a
tt
a
ins
be
tt
e
r
r
e
s
ult
s
thr
ough
R
2,
M
AE
,
M
S
E
,
a
nd
R
M
S
E
va
lues
of
a
bout
0
.
84,
0.
056
,
0.
081
,
a
nd
0
.
907
a
ppr
op
r
iate
ly
whe
n
c
ompar
e
d
to
e
xis
ti
ng
a
lgor
it
hms
.
F
igu
r
e
s
3
,
4,
a
nd
5
pr
e
s
e
nt
the
e
poc
h
v/s
a
c
c
ur
a
c
y,
e
poc
h
v/s
l
os
s
,
a
nd
c
onf
us
ion
matr
ix
f
or
S
AR
I
M
A
-
B
iL
S
T
M
r
e
s
pe
c
ti
v
e
ly.
C
onf
us
ion
matr
ix
is
a
table
e
mpl
oye
d
to
e
s
tablis
h
the
pr
e
diction
pe
r
f
or
manc
e
that
s
umm
a
r
ize
s
a
nd
v
is
ua
li
z
e
s
the
model
pe
r
f
or
manc
e
.
T
he
e
poc
hs
v/s
a
c
c
ur
a
c
y
a
nd
los
s
gr
a
phs
a
r
e
vis
ua
li
z
ing
the
pr
oc
e
s
s
whe
n
tr
a
ini
ng
the
model
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
E
lec
&
C
omp
E
ng
I
S
S
N:
2088
-
8708
Se
as
onal
auto
-
r
e
gr
e
s
s
ive
int
e
gr
ated
mov
ing
av
e
r
a
ge
w
it
h
…
(
N
ir
anjan
Shadak
s
har
appa
J
ay
anna
)
789
F
igur
e
3.
E
poc
h
v/s
a
c
c
ur
a
c
y
F
igur
e
4.
E
poc
h
v/s
los
s
F
igur
e
5.
C
onf
us
ion
m
a
tr
ix
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
15
,
No.
1
,
F
e
br
ua
r
y
20
25
:
783
-
791
790
3.
2.
Com
p
ar
a
t
ive
a
n
alys
is
T
he
S
AR
I
M
A
-
B
iL
S
T
M
is
c
ompar
e
d
with
e
xis
ti
n
g
methods
thr
ough
metr
ics
li
ke
R
2,
M
AE
,
M
S
E
a
nd
R
M
S
E
s
hown
in
T
a
ble
3.
T
he
e
xis
ti
ng
metho
ds
li
ke
M
ult
il
a
ye
r
s
tac
ke
d
e
ns
e
mbl
e
[
17
]
,
C
NN
-
D
NN
[
18]
,
C
E
N
-
S
C
A
-
B
iL
S
T
M
[
19]
,
a
nd
AR
I
M
A
[
20]
a
r
e
c
o
mpar
e
d
with
the
S
AR
I
M
A
-
B
iL
S
T
M
model.
T
he
B
iL
S
T
M
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
a
nd
bidi
r
e
c
ti
ona
l
ti
me
de
pe
nde
nc
ies
f
r
om
his
tor
ica
l
da
ta
whic
h
im
p
r
ove
s
t
he
model
pe
r
f
or
manc
e
.
T
he
p
r
opos
e
d
S
AR
I
M
A
-
B
iL
S
T
M
is
a
da
ptable
to
ha
ndli
ng
a
wide
s
pr
e
a
d
of
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
.
F
r
om
T
a
ble
3,
the
S
AR
I
M
A
-
B
iL
S
T
M
a
tt
a
ins
be
tt
e
r
r
e
s
ult
s
thr
ough
R
2,
M
AE
,
M
S
E
,
a
nd
R
M
S
E
va
lues
of
a
bout
0.
84
,
0
.
05
6,
0.
081
,
a
nd
0.
907
r
e
s
pe
c
ti
ve
ly.
T
a
ble
3.
C
ompar
a
ti
ve
a
na
lys
is
M
e
th
od
R2
M
A
E
M
S
E
R
M
S
E
M
ul
ti
la
ye
r
s
ta
c
ke
d e
n
s
e
mbl
e
[
17]
N
/A
6.63
N
/A
10.545
C
N
N
-
D
N
N
[
18]
0.87
0.199
0.071
0.266
C
E
N
-
S
C
A
-
B
iL
S
T
M
[
19]
0.974
35.77
46.98
68.42
A
R
I
M
A
[
20]
0.961
N
/A
N
/A
0.853
S
A
R
I
M
A
-
B
iL
S
T
M
0.84
0.056
0.081
0.907
3.
3.
Dis
c
u
s
s
ion
T
he
a
dva
ntage
s
of
S
AR
I
M
A
-
B
iL
S
T
M
a
nd
the
d
r
a
wba
c
ks
of
e
xis
ti
ng
tec
hniques
a
r
e
dis
c
us
s
e
d
in
thi
s
s
e
c
ti
on.
T
he
mul
ti
laye
r
s
tac
ke
d
e
ns
e
mbl
e
[
1
7]
ne
c
e
s
s
it
a
te
s
numer
ous
pr
e
pr
oc
e
s
s
ing
to
f
it
the
ba
s
e
li
ne
s
uppli
e
s
that
a
r
e
not
ini
ti
a
ted
f
o
r
yield
pr
e
diction
.
T
he
C
NN
-
DN
N
[
18]
r
e
quir
e
s
a
s
igni
f
ica
nt
c
a
pa
bil
it
y
o
f
da
ta
a
nd
i
t
c
a
nnot
c
ons
ider
the
t
im
e
-
s
e
r
ies
da
ta.
T
he
C
E
N
-
S
C
A
-
B
iL
S
T
M
[
19]
r
e
quir
e
d
a
va
s
t
a
m
ount
of
c
a
ptur
e
d
da
ta
whic
h
is
c
ompl
e
x
to
obtain.
I
n
A
R
I
M
A
[
20]
,
high
a
tt
e
nti
on
wa
s
r
e
quir
e
d
to
f
a
s
c
inate
a
nd
s
ti
mul
a
te
f
a
r
mer
s
in
c
oc
onut
pr
oduc
ti
on
.
T
he
S
AR
I
M
A
-
B
iL
S
T
M
outper
f
or
ms
thes
e
e
xis
ti
ng
model
li
mi
tations
.
T
he
AS
-
W
OA
is
uti
li
z
e
d
f
or
s
e
lec
ti
ng
the
be
s
t
f
e
a
tur
e
s
whic
h
e
nha
nc
e
s
the
c
onve
r
ge
nc
e
r
a
te
a
nd
a
voids
loca
l
opti
ma
is
s
ue
s
.
T
he
S
AR
I
M
A
-
B
iL
S
T
M
is
a
da
ptable
to
ha
ndli
ng
a
wide
s
pr
e
a
d
of
va
r
iou
s
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
.
4.
CONC
L
USI
ON
T
his
pa
pe
r
pr
opos
e
s
a
S
AR
I
M
A
-
B
i
L
S
T
M
f
or
c
oc
onut
yield
pr
e
diction.
T
he
c
oll
e
c
ted
da
tas
e
t
is
pr
e
pr
oc
e
s
s
e
d
thr
ough
a
labe
l
e
nc
ode
r
a
nd
mi
n
-
max
nor
maliza
ti
on
is
e
mpl
oye
d
to
c
ha
nge
non
-
numer
ic
f
e
a
tur
e
s
int
o
numer
ica
l
f
e
a
tur
e
s
a
nd
e
nha
nc
e
mo
de
l
pe
r
f
or
manc
e
.
T
he
pr
e
pr
oc
e
s
s
e
d
f
e
a
tur
e
s
a
r
e
s
e
lec
ted
thr
ough
AS
-
W
OA
whic
h
a
voids
loca
l
opti
ma
is
s
ue
s
.
T
he
n,
the
s
e
lec
ted
f
e
a
tur
e
s
a
r
e
g
iven
to
the
S
AR
I
M
A
-
B
iL
S
T
M
f
or
pr
e
dicting
the
c
oc
onut
yields
.
T
he
S
AR
I
M
A
-
B
iL
S
T
M
is
a
da
ptable
to
ha
ndli
ng
a
wide
s
pr
e
a
d
of
va
r
ious
s
e
a
s
ona
l
pa
tt
e
r
ns
a
nd
c
a
ptur
e
s
s
pa
ti
a
l
f
e
a
tur
e
s
.
T
he
S
AR
I
M
A
-
B
iL
S
T
M
pe
r
f
or
manc
e
is
e
s
ti
mate
d
thr
ough
R
2,
M
AE
,
M
S
E
a
nd
R
M
S
E
.
T
he
S
AR
I
M
A
-
B
iL
S
T
M
a
tt
a
ins
0.
84
of
R
2,
0.
056
of
M
AE
,
0.
081
of
M
S
E
,
a
nd
0
.
907
of
R
M
S
E
whic
h
is
be
tt
e
r
whe
n
c
ompar
e
d
to
e
xis
ti
ng
tec
hniques
.
I
n
the
f
utur
e
,
a
hybr
id
opti
mi
z
a
ti
on
a
lgor
it
hm
c
a
n
be
e
mpl
oye
d
to
e
nha
nc
e
the
S
AR
I
M
A
-
B
iL
S
T
M
pe
r
f
or
manc
e
.
RE
F
E
RE
NC
E
S
[
1]
P
.
S
in
gl
a
,
M
.
D
uha
n,
a
nd
S
.
S
a
r
oha
,
“
A
n
e
ns
e
mbl
e
me
th
od
to
f
or
e
c
a
s
t
24
-
h
a
he
a
d
s
ol
a
r
ir
r
a
di
a
nc
e
us
in
g
w
a
ve
le
t
de
c
ompos
it
i
on
a
nd
B
iL
S
T
M
de
e
p
le
a
r
ni
ng
ne
twor
k,”
E
ar
th
Sc
ie
n
c
e
I
nf
or
m
at
i
c
s
,
vol
.
15,
no.
1,
pp.
291
–
306,
M
a
r
.
2022,
doi
:
10.1007/s
121
45
-
021
-
00723
-
1.
[
2]
B
.
D
.
N
a
ik
a
nd
M
.
U
d
a
ya
kuma
r
,
“
O
pt
im
iz
a
ti
on
a
nd
c
ha
r
a
c
t
e
r
iz
a
ti
on
s
tu
di
e
s
on
th
e
pr
oduc
ti
on
of
bi
o
-
di
e
s
e
l
f
r
om
W
S
O
us
in
g
c
a
r
bon
c
a
ta
ly
s
t
de
r
iv
e
d
f
r
om
c
oc
onut
me
a
l
r
e
s
id
ue
,”
E
ne
r
gy
S
our
c
e
s
,
P
ar
t
A
:
R
e
c
ov
e
r
y
,
U
ti
li
z
at
io
n,
and
E
nv
ir
on
m
e
nt
al
E
ff
e
c
ts
,
vol
. 45, no. 4, pp. 9864
–
9879, Oc
t.
2023, doi:
10.1080/15567036.
2019.1683096.
[
3]
D
.
S
a
r
a
vi
a
e
t
al
.
,
“
Y
ie
ld
p
r
e
di
c
ti
on
of
f
our
be
a
n
(
pha
s
e
ol
us
vul
ga
r
is
)
c
ul
ti
va
r
s
us
in
g
ve
ge
ta
ti
on
in
di
c
e
s
ba
s
e
d
on
mul
ti
s
pe
c
tr
a
l
im
a
ge
s
f
r
om UAV
i
n a
n a
r
id
z
one
of
P
e
r
u,”
D
r
one
s
, vol
. 7, no. 5, M
a
y 2023, doi:
10.3390/dr
one
s
7050325.
[
4]
S
.
C
hi
nna
ppa
n,
“
D
R
I
S
nor
ms
f
or
id
e
nt
if
yi
ng
yi
e
ld
li
mi
ti
ng
mi
c
r
onut
r
ie
nt
s
in
c
oc
oa
unde
r
c
oc
onut
in
te
r
c
r
oppi
ng
s
ys
te
ms
,”
J
our
nal
of
P
la
nt
N
ut
r
it
io
n
, vol
. 45, no. 8, pp. 1214
–
1222, M
a
y
2022, doi:
10.1080/01904167.
2021.1994592.
[
5]
R
.
K
.
M
e
ga
li
nga
m,
S
.
K
.
M
a
noha
r
a
n,
D
.
H
. T
.
A
.
B
a
bu,
G
.
S
r
ir
a
m,
K
.
L
oke
s
h,
a
nd
S
.
K
.
S
udhe
e
s
h,
“
C
o
c
onut
tr
e
e
s
c
la
s
s
if
ic
a
t
io
n
ba
s
e
d
on
he
ig
ht
,
in
c
li
na
ti
on,
a
nd
or
ie
nt
a
ti
on
us
in
g
M
I
N
-
S
V
M
a
lg
or
it
hm,”
N
e
ur
al
C
om
put
in
g
and
A
ppl
ic
at
io
ns
,
vol
.
35,
no.
16,
pp. 12055
–
12071, J
un. 2023, doi:
10.1007/s
00521
-
023
-
08339
-
w.
[
6]
M
.
N
.
H
a
di
,
“
I
mpl
e
me
nt
a
ti
on
of
tr
a
di
ti
ona
l
r
is
k
ma
na
g
e
me
nt
a
s
lo
s
s
pr
e
ve
nt
io
n
in
c
oc
onut
pr
oduc
ti
on
r
e
s
ul
ts
,”
A
K
A
D
E
M
I
K
:
J
ur
nal
M
ahas
is
w
a E
k
onomi & B
is
ni
s
, vol
. 2, no. 2, pp. 92
–
102,
M
a
y 2022, doi:
10.37481/j
me
b.v2i2.554.
[
7]
R
.
S
ir
is
a
ngs
a
w
a
ng
a
nd
N
.
P
he
ty
im
,
“
O
pt
im
iz
a
ti
on
of
ta
nni
n
e
xt
r
a
c
ti
on
f
r
om
c
oc
onut
c
oi
r
th
r
ough
r
e
s
pons
e
s
ur
f
a
c
e
me
th
odol
ogy,”
H
e
li
y
on
, vol
. 9, no. 2, F
e
b. 2023, doi:
10.1016/j
.he
li
yon.2023.e
13377.
[
8]
K
.
B
.
H
e
bba
r
e
t
al
.
,
“
P
r
e
di
c
ti
ng
th
e
pot
e
nt
ia
l
s
ui
ta
bl
e
c
li
ma
te
f
or
c
oc
onut
(
C
oc
os
nuc
if
e
r
a
L
.
)
c
ul
ti
va
ti
on
in
I
ndi
a
und
e
r
c
li
ma
te
c
ha
nge
s
c
e
na
r
io
s
us
in
g t
he
M
a
xE
nt
mode
l,
”
P
la
nt
s
, vol
. 11, no. 6,
p
p.
731, M
a
r
. 2022, doi:
10.3390/pl
a
nt
s
11060731.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
E
lec
&
C
omp
E
ng
I
S
S
N:
2088
-
8708
Se
as
onal
auto
-
r
e
gr
e
s
s
ive
int
e
gr
ated
mov
ing
av
e
r
a
ge
w
it
h
…
(
N
ir
anjan
Shadak
s
har
appa
J
ay
anna
)
791
[
9]
B
.
D
a
s
,
D
.
M
ur
ga
onka
r
,
S
.
N
a
vy
a
s
hr
e
e
,
a
nd
P
.
K
uma
r
,
“
N
ove
l
c
ombi
na
ti
on
a
r
ti
f
ic
ia
l
ne
ur
a
l
ne
twor
k
mode
ls
c
oul
d
not
out
pe
r
f
or
m
in
di
vi
dua
l
mode
ls
f
or
w
e
a
th
e
r
-
ba
s
e
d
c
a
s
he
w
yi
e
l
d
pr
e
di
c
ti
on,”
I
nt
e
r
nat
io
nal
J
our
nal
of
B
io
m
e
te
o
r
ol
ogy
,
vol
.
66,
no. 8, pp. 1627
–
1638, Aug. 2022, do
i:
10.1007/s
00484
-
022
-
02306
-
1.
[
10]
V
.
M
a
ma
th
a
a
nd
J
. C
.
K
a
vi
th
a
,
“
M
a
c
hi
ne
le
a
r
ni
ng
ba
s
e
d c
r
op
g
r
ow
th
ma
na
ge
me
nt
in
gr
e
e
nhous
e
e
nvi
r
onme
nt
us
in
g
hydr
oponic
s
f
a
r
mi
ng t
e
c
hni
que
s
,”
M
e
as
ur
e
m
e
nt
:
Se
ns
o
r
s
, vol
. 25, F
e
b. 2023
, doi
:
10.1016/j
.me
a
s
e
n.2023.100665.
[
11]
K
.
P
.
W
a
id
ya
r
a
th
ne
, T
.
H
.
C
ha
ndr
a
th
il
a
ke
,
a
nd
W
.
S
.
W
ic
kr
a
m
a
r
a
c
hc
hi
,
“
A
ppl
ic
a
ti
on
of
a
r
ti
f
ic
ia
l
ne
ur
a
l
ne
twor
k
to
pr
e
di
c
t
c
o
p
r
a
c
onve
r
s
io
n
f
a
c
to
r
,”
N
e
u
r
al
C
om
put
in
g
and
A
ppl
ic
at
io
ns
,
vol
.
34,
no.
10,
pp.
7909
–
7918,
M
a
y
2022,
doi
:
10.1007/s
00521
-
0
22
-
06893
-
3.
[
12]
Z
.
A
hme
d,
P
.
P
.
N
a
ge
s
w
a
r
a
R
a
o,
a
nd
P
.
S
r
ik
a
nt
h,
“
A
r
e
a
e
s
t
im
a
ti
on
of
ma
ngo
a
nd
c
oc
onut
c
r
ops
us
in
g
ma
c
hi
ne
le
a
r
ni
ng
in
H
e
s
a
r
a
gha
tt
a
H
obl
i
of
B
e
nga
lu
r
u
U
r
ba
n
D
is
tr
ic
t,
K
a
r
na
ta
ka
,
”
J
our
nal
of
G
e
om
at
ic
s
,
vol
.
17,
no.
1,
pp.
85
–
92,
A
pr
.
2023,
d
oi
:
10.58825/j
og.2023.17.1.75.
[
13]
P
.
S
.
S
.
G
opi
a
nd
M
.
K
a
r
th
ik
e
ya
n,
“
R
e
d
f
ox
opt
im
iz
a
ti
on
w
it
h
e
ns
e
mbl
e
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
f
or
c
r
op
r
e
c
omm
e
nda
ti
on
a
nd
yi
e
ld
pr
e
di
c
ti
on
mode
l,
”
M
ul
ti
m
e
di
a
T
ool
s
and
A
ppl
ic
at
io
ns
,
vol
.
83,
no.
5,
pp.
13159
–
13179,
J
ul
.
2023,
doi
:
10.1007/s
110
42
-
023
-
16113
-
2.
[
14]
R
.
P
a
ndi
s
e
lv
a
m
e
t
al
.
,
“
C
e
nt
r
a
l
c
ompos
it
e
de
s
ig
n,
pa
r
e
to
a
na
ly
s
is
,
a
nd
a
r
ti
f
ic
ia
l
ne
ur
a
l
ne
twor
k
f
o
r
mode
li
ng
o
f
mi
c
r
ow
a
ve
pr
oc
e
s
s
in
g pa
r
a
me
te
r
s
f
or
t
e
nde
r
c
oc
onut
w
a
te
r
,”
M
e
as
ur
e
m
e
nt
:
F
ood
, vol
. 5, F
e
b. 2022, doi:
10.1016/j
.me
a
f
oo.2021.100015.
[
15]
R
.
E
.
S
.
C
a
r
o
e
t
al
.
,
“
M
in
in
g
a
nd
v
a
li
da
ti
on
of
nove
l
s
im
pl
e
s
e
qu
e
nc
e
r
e
pe
a
t
(
S
S
R
)
ma
r
ke
r
s
de
r
iv
e
d
f
r
om
c
o
c
onut
(
C
o
c
os
nuc
if
e
r
a
L
.
)
ge
nome
a
s
s
e
mbl
y,”
J
ou
r
nal
of
G
e
ne
ti
c
E
ngi
ne
e
r
in
g
and
B
io
te
c
hnol
ogy
,
vol
.
20,
no.
1,
p
p
.
71,
D
e
c
.
2022,
doi
:
10.1186/s
43141
-
022
-
00354
-
z.
[
16]
P
.
S
.
S
.
G
opi
a
nd
M
.
K
a
r
th
ik
e
ya
n,
“
M
ul
ti
moda
l
ma
c
hi
n
e
le
a
r
ni
ng
ba
s
e
d
c
r
op
r
e
c
omm
e
nda
ti
on
a
nd
yi
e
ld
pr
e
di
c
ti
on
mod
e
l,
”
I
nt
e
ll
ig
e
nt
A
ut
om
at
io
n & Sof
t
C
om
put
in
g
, vol
. 36, no. 1, pp. 31
3
–
326, 2023, doi:
10.32604/i
a
s
c
.2023.029756.
[
17]
S
. I
ni
ya
n a
nd R
. J
e
ba
kuma
r
, “
M
ut
ua
l
in
f
or
ma
ti
on
f
e
a
tu
r
e
s
e
le
c
ti
on (
M
I
F
S
)
ba
s
e
d c
r
op yie
ld
pr
e
di
c
ti
on on
c
or
n a
nd s
oybe
a
n c
r
op
s
us
in
g mul
ti
la
ye
r
s
ta
c
ke
d e
ns
e
mbl
e
r
e
gr
e
s
s
io
n (
M
S
E
R
)
,”
W
ir
e
le
s
s
P
e
r
s
onal
C
o
m
m
uni
c
at
io
ns
, vol
. 126, no. 3, pp. 1935
–
1964, Oc
t.
2022, doi:
10.1007/s
11277
-
021
-
08712
-
9.
[
18]
A
.
O
ik
onomi
di
s
,
C
.
C
a
ta
l,
a
nd
A
.
K
a
s
s
a
hun,
“
H
ybr
id
de
e
p
le
a
r
ni
ng
-
ba
s
e
d
mode
ls
f
or
c
r
op
yi
e
ld
pr
e
di
c
ti
on,”
A
ppl
ie
d
A
r
ti
fi
c
ia
l
I
nt
e
ll
ig
e
nc
e
, vol
. 36, no. 1, De
c
. 2022, doi:
10.1080/08839514.
2022.2031823.
[
19]
T
.
P
e
ng,
C
.
Z
ha
ng,
J
.
Z
hou,
a
nd
M
.
S
.
N
a
z
ir
,
“
A
n
in
te
gr
a
te
d
f
r
a
me
w
or
k
of
B
i
-
di
r
e
c
ti
ona
l
lo
ng
-
s
hor
t
te
r
m
me
mor
y
(
B
i
L
S
T
M
)
ba
s
e
d
on
s
in
e
c
os
in
e
a
lg
or
it
hm
f
or
hou
r
ly
s
ol
a
r
r
a
di
a
ti
on
f
or
e
c
a
s
ti
ng,”
E
ne
r
gy
,
vol
.
221,
A
pr
.
2021,
d
oi
:
10.1016/j
.e
ne
r
gy.2021.119887.
[
20]
T
.
P
r
a
s
e
r
t
a
nd
V
.
R
ungr
e
unga
nun,
“
T
ha
i
c
oc
onut
pr
ic
e
f
or
e
c
a
s
ti
ng
us
in
g
A
r
im
a
mode
l,
”
I
nt
e
r
nat
io
nal
J
our
nal
of
I
ndus
t
r
ia
l
E
ngi
ne
e
r
in
g R
e
s
e
ar
c
h and
D
e
v
e
lo
pm
e
nt
, vol
. 12, no. 1, 2021, doi
:
10.34218/I
J
A
R
E
T
.12.1.2020.087.
[
21]
H
.
N
ova
r
ia
nt
o,
“
E
s
ti
ma
ti
ng
c
oc
onut
pr
oduc
ti
on
a
nd
pr
oduc
t
iv
it
y
of
lo
c
a
l
ta
ll
in
T
a
li
a
bu
I
s
la
nd
u
s
in
g
dr
one
a
nd
s
a
mpl
in
g
popula
ti
on,”
C
O
R
D
, vol
. 38, pp. 33
–
42, J
un. 2022, doi:
10.37833/cor
d.v38i.453.
[
22]
S
.
S
ubha
a
nd
S
.
B
.
P
r
iy
a
,
“
P
r
e
di
c
ti
on
mode
l
f
o
r
a
good
le
a
r
ni
ng
e
nvi
r
onme
nt
us
in
g
a
n
e
ns
e
mbl
e
a
ppr
oa
c
h,”
C
om
put
e
r
Sy
s
t
e
m
s
Sc
ie
nc
e
and E
ngi
ne
e
r
in
g
, vol
. 44, no. 3, pp. 2081
–
2093, 2023,
doi
:
10.32604/cs
s
e
.2023.028451.
[
23]
S
.
D
e
y,
P
.
K
.
R
oy,
a
nd
A
.
S
a
r
ka
r
,
“
A
d
a
pt
iv
e
I
I
R
mode
l
i
de
nt
if
ic
a
ti
on
us
in
g
c
h
a
ot
ic
oppos
it
io
n
-
ba
s
e
d
w
h
a
le
opt
im
iz
a
t
io
n
a
lg
or
it
hm,”
J
our
nal
of
E
le
c
tr
ic
al
Sy
s
te
m
s
and I
nf
or
m
at
io
n T
e
c
h
nol
ogy
, vol
. 10, no. 1, 2023, doi
:
10.1186/s
43067
-
023
-
00102
-
4.
[
24]
C
.
W
a
ng
a
nd
J
.
Q
ia
o,
“
C
on
s
tr
uc
ti
on
pr
oj
e
c
t
c
o
s
t
pr
e
di
c
ti
on
me
th
od
ba
s
e
d
on
im
pr
ove
d
B
iL
S
T
M
,”
A
ppl
ie
d
Sc
ie
n
c
e
s
,
vol
.
14,
no.
3, J
a
n. 2024, doi:
10.3390/app140309
78.
[
25]
D
.
G
.
da
S
il
va
,
M
.
T
.
B
.
G
e
ll
e
r
,
M
.
S
.
dos
S
.
M
our
a
,
a
nd
A
.
A
.
de
M
.
M
e
ne
s
e
s
,
“
P
e
r
f
or
ma
nc
e
e
va
lu
a
ti
on
of
L
S
T
M
ne
ur
a
l
ne
twor
ks
f
or
c
ons
umpt
io
n
pr
e
di
c
ti
on,”
e
-
P
r
ime
-
A
dv
anc
e
s
in
E
le
c
tr
ic
al
E
ngi
ne
e
r
in
g,
E
le
c
tr
oni
c
s
and
E
ne
r
gy
,
vol
.
2,
N
ov.
2
022,
doi
:
10.1016/j
.pr
im
e
.2022.100030
B
I
OG
RA
P
HI
E
S
OF
AU
T
HO
RS
N
i
r
a
nja
n
Sha
d
a
ks
h
a
ra
pp
a
J
a
y
a
nn
a
i
s
cu
rren
t
l
y
w
o
rk
i
n
g
as
an
a
s
s
i
s
t
an
t
p
ro
fe
s
s
o
r
i
n
t
h
e
D
e
p
art
me
n
t
o
f
Co
m
p
u
t
er
Sci
e
n
ce
an
d
E
n
g
i
n
eeri
n
g
,
K
a
l
p
a
t
aru
In
s
t
i
t
u
t
e
o
f
T
ech
n
o
l
o
g
y
,
T
i
p
t
u
r,
w
h
i
c
h
i
s
a
ffi
l
i
a
t
ed
t
o
V
i
s
v
es
v
aray
a
T
ech
n
o
l
o
g
i
ca
l
U
n
i
v
er
s
i
t
y
,
Bel
ag
a
v
i
,
K
arn
a
t
ak
a.
H
e
can
b
e
co
n
t
ac
t
ed
at
ema
i
l
:
n
i
ra
n
j
a
n
s
j
5
5
5
@
g
mai
l
.
co
m.
R
a
v
i
pra
k
a
s
h
M
a
denur
Li
n
g
a
r
a
ju
i
s
c
u
rren
t
l
y
w
o
rk
i
n
g
as
a
p
r
o
fes
s
o
r
i
n
t
h
e
D
ep
ar
t
men
t
o
f
Co
m
p
u
t
er
Sci
e
n
ce
an
d
E
n
g
i
n
eer
i
n
g
,
K
al
p
at
ar
u
In
s
t
i
t
u
t
e
o
f
T
ech
n
o
l
o
g
y
,
T
i
p
t
u
r,
w
h
i
ch
i
s
a
ffi
l
i
at
e
d
t
o
V
i
s
v
e
s
v
ara
y
a
T
ec
h
n
o
l
o
g
i
cal
U
n
i
v
ers
i
t
y
,
Bel
a
g
av
i
,
K
arn
a
t
ak
a.
H
e
can
b
e
co
n
t
act
e
d
at
emai
l
:
rav
i
p
r
ak
a
s
h
m
l
@
g
mai
l
.
co
m.
Evaluation Warning : The document was created with Spire.PDF for Python.