I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
,
p
p
.
2300
~
2
3
1
3
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
3
.
p
p
2
3
0
0
-
2
3
1
3
2300
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Co
mpa
ring
glo
ba
l and v
a
riety
-
spec
ific
ense
mble mo
d
els for
a
v
o
ca
do
ma
turit
y
predic
tion wi
th
n
ea
r
-
infrared
Chris
t
ia
n O
v
a
lle,
J
ho
nn
G
a
rc
ia
J
im
enez
,
J
o
s
e
B
rio
nes
Z
uñ
ig
a
D
e
p
a
r
t
me
n
t
o
f
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
,
U
n
i
v
e
r
si
d
a
d
Te
c
n
o
l
ó
g
i
c
a
d
e
l
P
e
r
ú
,
Li
m
a
,
P
e
r
ú
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
1
9
,
2
0
2
5
R
ev
is
ed
Feb
1
6
,
2
0
2
6
Acc
ep
ted
Ap
r
2
0
,
2
0
2
6
E
n
s
u
r
i
n
g
a
c
c
u
ra
te,
n
o
n
-
d
e
s
tr
u
c
t
i
v
e
m
a
t
u
ri
t
y
c
la
ss
i
fi
c
a
t
i
o
n
o
f
a
v
o
c
a
d
o
s
i
s
c
ri
ti
c
a
l
t
o
s
u
p
p
l
y
c
h
a
i
n
o
p
ti
m
iza
ti
o
n
i
n
a
g
r
o
-
i
n
d
u
st
ria
l
s
y
st
e
m
s.
T
h
is
s
tu
d
y
p
re
se
n
t
s
a
p
re
d
ict
i
v
e
fra
m
e
w
o
r
k
t
h
a
t
i
n
t
e
g
ra
tes
n
e
a
r
-
i
n
f
ra
r
e
d
(N
IR)
sp
e
c
tr
o
sc
o
p
y
w
i
t
h
e
n
se
m
b
le
sta
c
k
i
n
g
m
a
c
h
i
n
e
l
e
a
r
n
in
g
(M
L)
t
o
e
n
h
a
n
c
e
th
e
p
re
c
is
i
o
n
o
f
a
v
o
c
a
d
o
ri
p
e
n
e
ss
a
ss
e
s
sm
e
n
t
.
T
h
e
p
r
o
p
o
se
d
m
e
t
h
o
d
o
l
o
g
y
c
o
m
p
a
re
s
g
l
o
b
a
l
v
e
rs
u
s
v
a
r
ie
ty
-
s
p
e
c
i
fi
c
m
o
d
e
l
s
f
o
r
'
Ha
ss
'
a
n
d
'
F
u
e
r
te'
a
v
o
c
a
d
o
t
y
p
e
s
,
le
v
e
ra
g
i
n
g
s
p
e
c
t
r
a
l
d
a
ta
(9
0
0
–
1
,
7
0
0
n
m
)
a
n
d
m
u
l
ti
p
le
b
a
s
e
c
las
s
if
ie
r
s,
i
n
c
l
u
d
i
n
g
ra
n
d
o
m
f
o
r
e
st
(
R
F
)
,
g
ra
d
ie
n
t
b
o
o
s
ti
n
g
(G
B)
,
s
u
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
s
(
S
V
M
s
)
,
d
e
c
is
i
o
n
t
re
e
s
(
DT
)
,
k
-
n
e
a
re
s
t
n
e
i
g
h
b
o
rs
(KN
N)
,
a
n
d
c
a
te
g
o
ric
a
l
b
o
o
st
i
n
g
(
Ca
tB
o
o
s
t
)
,
c
o
m
b
i
n
e
d
v
ia
l
i
n
e
a
r
r
e
g
re
s
si
o
n
a
s
a
m
e
ta
-
lea
r
n
e
r
.
E
x
p
e
rim
e
n
ta
l
re
s
u
l
ts
re
v
e
a
le
d
t
h
a
t
th
e
st
a
c
k
i
n
g
m
o
d
e
ls
o
u
t
p
e
rf
o
rm
e
d
i
n
d
i
v
i
d
u
a
l
le
a
r
n
e
r
s,
wi
t
h
v
a
rie
t
y
-
sp
e
c
i
fic
GB
m
o
d
e
l
a
c
h
ie
v
i
n
g
th
e
h
i
g
h
e
st
p
e
rf
o
rm
a
n
c
e
(M
a
t
t
h
e
ws
c
o
r
re
l
a
t
i
o
n
c
o
e
f
fi
c
ie
n
t
(
M
CC
)
=
0
.
6
7
9
,
a
re
a
u
n
d
e
r
t
h
e
c
u
r
v
e
(
AU
C
)
=
0
.
9
3
1
)
.
T
h
e
se
f
i
n
d
i
n
g
s
h
i
g
h
l
i
g
h
t
t
h
e
c
ri
ti
c
a
l
im
p
o
rt
a
n
c
e
o
f
v
a
r
iet
a
l
s
p
e
c
if
ic
it
y
i
n
m
o
d
e
l
c
a
l
i
b
ra
ti
o
n
a
n
d
d
e
m
o
n
stra
te
h
o
w
e
n
se
m
b
le
s
tr
a
te
g
ies
c
a
n
im
p
ro
v
e
r
o
b
u
s
tn
e
ss
,
sc
a
la
b
i
l
it
y
,
a
n
d
in
te
r
p
re
ta
b
i
li
t
y
in
in
te
ll
i
g
e
n
t
a
g
ric
u
lt
u
ra
l
s
y
s
tem
s.
T
h
e
p
r
o
p
o
se
d
m
o
d
e
l
p
r
o
v
i
d
e
s
a
c
o
m
p
u
tat
i
o
n
a
l
l
y
e
f
fic
ie
n
t
s
o
l
u
ti
o
n
f
o
r
re
a
l
-
ti
m
e
q
u
a
l
it
y
c
o
n
tr
o
l
a
n
d
s
u
p
p
o
r
ts
th
e
d
e
p
l
o
y
m
e
n
t
o
f
AI
-
p
o
we
re
d
s
y
s
tem
s
wi
t
h
in
a
g
r
ic
u
lt
u
ra
l
s
u
p
p
l
y
c
h
a
i
n
s
i
n
d
e
v
e
l
o
p
i
n
g
re
g
i
o
n
s.
K
ey
w
o
r
d
s
:
Av
o
ca
d
o
r
ip
en
ess
E
n
s
em
b
le
s
tack
in
g
Ma
ch
in
e
lear
n
in
g
Near
-
in
f
r
ar
e
d
s
p
ec
tr
o
s
co
p
y
No
n
-
d
estru
ctiv
e
q
u
ality
co
n
tr
o
l
Pre
cisi
o
n
ag
r
icu
ltu
r
e
Var
iety
-
s
p
ec
if
ic
m
o
d
els
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
:
C
h
r
is
tian
Ov
alle
Dep
ar
tm
en
t o
f
E
n
g
in
ee
r
in
g
,
Facu
lty
o
f
E
n
g
in
ee
r
i
n
g
,
Un
i
v
er
s
id
ad
T
ec
n
o
ló
g
ica
d
el
Per
ú
L
im
a,
Per
u
E
m
ail:
d
o
v
alle@
u
tp
.
ed
u
.
p
e
1.
I
NT
RO
D
UCT
I
O
N
I
n
r
ec
en
t
y
ea
r
s
,
th
e
av
o
ca
d
o
in
d
u
s
tr
y
h
as
co
n
s
o
lid
ated
its
p
o
s
itio
n
as
o
n
e
o
f
t
h
e
m
o
s
t
ec
o
n
o
m
ically
v
alu
ab
le
an
d
s
tr
ateg
ically
r
el
ev
an
t
s
ec
to
r
s
with
in
Per
u
v
ian
ag
r
icu
ltu
r
e
[
1
]
,
[
2
]
.
Per
u
cu
r
r
en
tly
r
a
n
k
s
am
o
n
g
th
e
wo
r
ld
’
s
lead
in
g
av
o
ca
d
o
ex
p
o
r
ter
s
,
co
n
s
is
ten
tly
o
cc
u
p
y
in
g
a
p
o
s
itio
n
with
in
th
e
to
p
t
h
r
ee
g
lo
b
al
p
r
o
d
u
ce
r
s
,
d
r
iv
en
b
y
s
u
s
tain
ed
in
ter
n
atio
n
al
d
em
a
n
d
an
d
th
e
in
cr
ea
s
in
g
p
r
ef
e
r
en
ce
f
o
r
h
ig
h
-
q
u
ality
f
r
u
it
in
g
lo
b
al
m
ar
k
ets
[
3
]
,
[
4
]
.
T
h
is
co
n
tex
t
h
ig
h
lig
h
ts
th
e
u
r
g
en
t
n
ee
d
f
o
r
ef
f
icien
t,
r
eliab
le,
an
d
s
ca
lab
le
m
eth
o
d
s
ca
p
ab
le
o
f
ass
ess
in
g
in
ter
n
al
q
u
ality
attr
ib
u
tes
th
at
d
eter
m
i
n
e
co
m
m
er
cial
m
atu
r
ity
,
s
h
elf
lif
e,
an
d
p
o
s
th
ar
v
est
p
er
f
o
r
m
an
ce
[
5
]
.
C
o
n
v
en
tio
n
al
m
atu
r
ity
ass
ess
m
en
t
tech
n
iq
u
es,
in
clu
d
in
g
f
ir
m
n
ess
test
in
g
an
d
d
r
y
m
atter
d
eter
m
in
atio
n
(
co
ef
f
icien
ts
o
f
d
eter
m
in
atio
n
o
r
R
2
)
,
ar
e
wid
e
ly
u
s
ed
b
u
t
p
r
esen
t
s
ig
n
if
ican
t
lim
itatio
n
s
.
T
h
ese
m
eth
o
d
s
ar
e
d
estru
ctiv
e,
la
b
o
r
-
in
te
n
s
iv
e,
tim
e
-
co
n
s
u
m
in
g
,
an
d
u
n
s
u
itab
le
f
o
r
lar
g
e
-
s
ca
le
o
r
r
ea
l
-
tim
e
in
d
u
s
tr
ial
ap
p
licatio
n
s
[
6
]
–
[
8
]
.
As
a
r
esp
o
n
s
e
to
th
ese
co
n
s
tr
ain
ts
,
n
ea
r
-
in
f
r
ar
ed
(
NI
R
)
s
p
ec
tr
o
s
co
p
y
h
as
em
er
g
ed
as
a
p
r
o
m
is
in
g
n
o
n
-
d
estru
ctiv
e
alter
n
ativ
e,
en
ab
li
n
g
r
ap
id
esti
m
atio
n
o
f
i
n
ter
n
a
l
q
u
ality
p
ar
am
eter
s
s
u
ch
as
d
r
y
m
atter
,
s
u
g
ar
co
n
t
en
t,
an
d
lip
id
co
n
ce
n
tr
atio
n
w
ith
o
u
t
d
am
ag
in
g
th
e
f
r
u
it
[
9
]
,
[
1
0
]
.
Owin
g
to
its
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
o
mp
a
r
in
g
g
lo
b
a
l a
n
d
va
r
iety
-
s
p
ec
ific e
n
s
emb
le
mo
d
els fo
r
a
vo
ca
d
o
ma
tu
r
ity
…
(
C
h
r
is
tia
n
Ova
lle)
2301
r
ap
id
ac
q
u
is
itio
n
,
p
o
r
tab
ilit
y
,
an
d
r
ea
g
e
n
t
-
f
r
ee
o
p
er
atio
n
,
NI
R
s
p
ec
tr
o
s
co
p
y
h
as b
ee
n
s
u
cc
ess
f
u
lly
d
ep
lo
y
ed
i
n
b
o
th
lab
o
r
ato
r
y
en
v
ir
o
n
m
en
ts
an
d
co
m
m
er
cial
p
ac
k
in
g
lin
es,
p
o
s
itio
n
in
g
it
as
a
p
r
ac
tic
al
to
o
l
f
o
r
q
u
ality
co
n
tr
o
l in
t
h
e
ag
r
i
-
f
o
o
d
in
d
u
s
tr
y
[
1
1
]
–
[
1
3
]
.
E
x
ten
s
iv
e
r
esear
ch
h
as
d
em
o
n
s
tr
ated
th
e
ap
p
licab
ilit
y
o
f
NI
R
s
p
ec
tr
o
s
co
p
y
ac
r
o
s
s
a
wi
d
e
r
an
g
e
o
f
f
r
u
its
,
in
clu
d
in
g
m
an
g
o
,
ap
p
le
,
k
iwi,
g
r
ap
e,
an
d
b
an
an
a
[
1
4
]
–
[
1
7
]
.
T
r
ad
itio
n
ally
,
r
eg
r
ess
io
n
-
b
ased
tec
h
n
iq
u
es
s
u
ch
as
p
ar
tial
least
s
q
u
ar
es
(
PLS)
an
d
its
v
ar
ian
ts
,
in
clu
d
in
g
d
i
-
PLS,
h
av
e
b
ee
n
e
m
p
lo
y
ed
to
m
o
d
el
f
r
u
it
m
atu
r
ity
.
Ho
wev
e
r
,
th
ei
r
p
r
ed
ictiv
e
p
er
f
o
r
m
a
n
ce
is
o
f
t
en
in
f
lu
e
n
ce
d
b
y
c
o
n
tex
tu
a
l
f
ac
to
r
s
s
u
ch
a
s
en
v
ir
o
n
m
en
tal
c
o
n
d
itio
n
s
,
s
ea
s
o
n
al
v
ar
ia
b
ilit
y
,
an
d
c
u
ltiv
ar
-
s
p
ec
if
ic
ch
ar
ac
ter
is
tics
[
1
8
]
,
[
1
9
]
.
I
n
th
e
s
p
ec
if
ic
ca
s
e
o
f
av
o
ca
d
o
,
p
r
e
v
io
u
s
s
tu
d
ies
u
s
in
g
NI
R
s
p
ec
tr
o
s
co
p
y
h
av
e
r
ep
o
r
ted
p
r
o
m
is
in
g
p
r
ed
i
ctiv
e
ca
p
ab
ilit
ies
f
o
r
m
atu
r
ity
-
r
elate
d
p
ar
am
eter
s
,
p
ar
ticu
lar
ly
d
r
y
m
atter
,
f
ir
m
n
es
s
,
an
d
lip
id
c
o
n
ten
t,
with
R
2
co
m
m
o
n
l
y
r
an
g
in
g
b
etwe
en
0
.
7
0
an
d
0
.
8
0
[
2
0
]
–
[
2
7
]
.
Alth
o
u
g
h
s
o
m
e
wo
r
k
s
h
av
e
ac
h
iev
ed
h
i
g
h
er
ac
cu
r
ac
y
lev
els
(
R
2
≈
0
.
9
8
)
th
r
o
u
g
h
o
p
tim
ized
ca
lib
r
atio
n
an
d
c
r
o
s
s
-
v
alid
atio
n
s
tr
ateg
ies
[
2
8
]
,
t
h
ese
f
in
d
in
g
s
a
ls
o
em
p
h
asize
t
h
e
s
en
s
itiv
ity
o
f
NI
R
-
b
ased
m
o
d
e
ls
to
m
ea
s
u
r
em
en
t c
o
n
d
itio
n
s
,
s
am
p
lin
g
p
r
o
to
co
ls
,
an
d
s
ea
s
o
n
al
ef
f
ec
ts
[
2
9
]
.
Mo
r
e
r
ec
e
n
t
ly
,
ad
v
an
c
e
s
in
m
ac
h
in
e
lea
r
n
in
g
(
ML
)
h
av
e
co
n
tr
ib
u
ted
t
o
f
u
r
t
h
er
im
p
r
o
v
em
en
t
s
i
n
p
r
ed
i
ct
iv
e
p
e
r
f
o
r
m
an
ce
.
Alg
o
r
ith
m
s
s
u
ch
a
s
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
e
two
r
k
s
(
C
NN
s
)
an
d
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
s
(
S
V
Ms
)
h
a
v
e
d
em
o
n
s
tr
a
ted
en
h
an
c
ed
f
ea
tu
r
e
e
x
tr
a
ct
io
n
an
d
cl
a
s
s
if
ic
at
io
n
c
ap
ab
il
it
ie
s
in
H
as
s
av
o
ca
d
o
d
a
ta
s
et
s
[
3
0
]
.
De
s
p
i
te
th
e
s
e
ad
v
an
c
e
s
,
m
o
s
t
ex
i
s
tin
g
s
tu
d
ie
s
f
o
cu
s
o
n
s
in
g
le
-
v
ar
i
ety
d
a
ta
s
e
t
s
o
r
d
ev
e
lo
p
g
en
e
r
a
li
ze
d
m
o
d
el
s
w
it
h
o
u
t
e
x
p
l
ic
it
ly
ad
d
r
e
s
s
in
g
th
e
tr
ad
e
-
o
f
f
b
e
tw
ee
n
g
lo
b
al
m
o
d
el
s
t
r
ai
n
ed
ac
r
o
s
s
m
u
l
tip
le
v
ar
ie
ti
es
an
d
v
ar
ie
ty
-
s
p
e
cif
ic
m
o
d
e
l
s
.
T
h
i
s
g
ap
is
p
ar
t
icu
lar
ly
r
el
ev
an
t
i
n
r
e
al
-
wo
r
l
d
p
r
o
d
u
ct
io
n
s
y
s
tem
s
,
wh
er
e
h
ete
r
o
g
en
eo
u
s
cu
lt
iv
ar
s
co
e
x
is
t
a
n
d
r
eq
u
ir
e
r
o
b
u
s
t
y
e
t
a
d
ap
t
ab
l
e
p
r
ed
ic
t
iv
e
s
o
lu
t
io
n
s
[
2
0
]
,
[
2
1
]
.
W
ith
in
th
is
f
r
am
ewo
r
k
,
a
n
d
with
th
e
o
b
jectiv
e
o
f
p
r
o
m
o
tin
g
th
e
a
d
o
p
tio
n
o
f
n
o
n
-
d
estru
ctiv
e
tech
n
o
lo
g
ies
in
th
e
Per
u
v
ian
ag
r
o
-
e
x
p
o
r
t
s
ec
to
r
,
t
h
is
s
tu
d
y
p
r
o
p
o
s
es
an
ap
p
r
o
ac
h
b
ased
o
n
NI
R
r
ef
lecta
n
ce
s
p
ec
tr
o
s
co
p
y
c
o
m
b
in
e
d
with
s
u
p
er
v
is
ed
ML
alg
o
r
ith
m
s
.
T
h
e
m
ain
o
b
jectiv
e
is
to
d
ev
el
o
p
a
n
o
n
-
in
v
asiv
e
s
y
s
tem
f
o
r
av
o
ca
d
o
m
at
u
r
ity
p
r
ed
ictio
n
a
n
d
t
o
s
y
s
tem
atica
lly
ev
alu
ate
p
er
f
o
r
m
a
n
ce
d
if
f
er
en
ce
s
b
etwe
en
g
lo
b
al
m
o
d
els
an
d
v
ar
iety
-
s
p
ec
if
ic
m
o
d
els.
B
y
d
o
i
n
g
s
o
,
t
h
e
s
tu
d
y
aim
s
to
o
p
tim
ize
m
a
tu
r
ity
class
if
icatio
n
an
d
p
o
s
th
ar
v
est
m
an
ag
em
e
n
t
p
r
o
ce
s
s
es,
u
ltima
tely
co
n
tr
i
b
u
tin
g
to
im
p
r
o
v
ed
o
p
er
atio
n
al
ef
f
icien
cy
a
n
d
co
m
p
etitiv
en
ess
in
in
ter
n
atio
n
al
m
ar
k
ets.
T
h
e
n
o
v
elty
o
f
th
is
r
esear
ch
li
es
in
th
e
ap
p
licatio
n
o
f
e
n
s
em
b
le
s
tack
in
g
as
a
m
atu
r
ity
class
if
icatio
n
s
tr
ateg
y
,
an
ap
p
r
o
ac
h
th
at
r
em
ain
s
u
n
d
er
ex
p
lo
r
ed
in
NI
R
-
b
ased
av
o
ca
d
o
ass
ess
m
en
t.
Un
lik
e
in
d
iv
id
u
al
p
r
ed
ictiv
e
m
o
d
els,
s
tack
in
g
in
teg
r
ates
th
e
co
m
p
lem
en
t
ar
y
s
tr
en
g
th
s
o
f
m
u
ltip
le
s
u
p
er
v
is
ed
lear
n
in
g
alg
o
r
ith
m
s
th
r
o
u
g
h
a
m
eta
-
lea
r
n
in
g
f
r
am
ewo
r
k
,
r
esu
ltin
g
in
im
p
r
o
v
e
d
p
r
ed
ictiv
e
ac
c
u
r
ac
y
an
d
g
r
ea
ter
m
o
d
el
s
tab
ilit
y
.
T
h
is
m
eth
o
d
o
lo
g
ical
co
n
tr
ib
u
tio
n
n
o
t
o
n
ly
en
h
an
c
es
th
e
r
o
b
u
s
tn
ess
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
tem
b
u
t
also
f
ac
ilit
ates
it
s
in
teg
r
atio
n
in
to
d
ig
ital
ag
r
ib
u
s
in
ess
p
latf
o
r
m
s
,
in
clu
d
in
g
I
o
T
-
b
ased
f
ield
m
o
n
ito
r
in
g
s
y
s
tem
s
,
d
ig
ital
tr
ac
ea
b
ilit
y
s
o
lu
tio
n
s
in
s
u
p
p
ly
ch
ain
s
,
an
d
Ag
r
ic
u
ltu
r
e
4
.
0
ap
p
licatio
n
s
.
Fu
r
th
e
r
m
o
r
e,
th
e
co
m
p
a
r
ativ
e
an
aly
s
is
b
etwe
en
g
l
o
b
al
a
n
d
v
ar
iety
-
s
p
ec
if
ic
m
o
d
els
p
r
o
v
id
es
p
r
ac
tical
g
u
id
eli
n
es
f
o
r
lar
g
e
-
s
ca
le
im
p
lem
en
tatio
n
i
n
co
m
m
er
ci
al
class
if
icatio
n
s
y
s
tem
s
.
Fin
ally
,
th
is
s
tu
d
y
co
n
tr
i
b
u
tes
t
o
th
e
tech
n
o
l
o
g
ical
ad
v
an
ce
m
e
n
t
o
f
d
ev
elo
p
in
g
c
o
u
n
tr
ies
b
y
p
r
o
p
o
s
in
g
co
m
p
u
t
atio
n
ally
ef
f
icien
t,
s
ca
lab
le,
a
n
d
in
d
u
s
tr
y
-
o
r
ien
ted
s
o
lu
tio
n
s
th
at
s
u
p
p
o
r
t
ex
p
o
r
t c
o
m
p
etitiv
en
ess
an
d
f
o
s
ter
in
n
o
v
atio
n
in
th
e
ag
r
icu
ltu
r
al
s
ec
to
r
.
2.
M
E
T
H
O
D
T
h
e
m
eth
o
d
o
lo
g
y
o
f
th
is
s
tu
d
y
was
s
tr
u
ctu
r
ed
f
o
llo
win
g
th
e
p
h
ases
o
f
t
h
e
cr
o
s
s
-
in
d
u
s
tr
y
s
tan
d
ar
d
p
r
o
ce
s
s
f
o
r
d
ata
m
in
in
g
(
C
R
I
SP
-
DM
)
f
r
am
ewo
r
k
[
3
1
]
.
T
h
e
ad
o
p
tio
n
o
f
th
is
f
r
a
m
ewo
r
k
,
a
wid
ely
r
ec
o
m
m
en
d
ed
p
r
ac
tice
f
o
r
en
s
u
r
in
g
r
o
b
u
s
tn
ess
in
ML
p
r
o
j
ec
ts
,
was
ad
ap
ted
h
er
e
to
ad
d
r
ess
a
ch
em
o
m
etr
ic
class
if
icatio
n
p
r
o
b
lem
.
Ad
h
er
i
n
g
to
a
s
y
s
tem
atic
wo
r
k
f
lo
w
a
lig
n
s
with
b
est
p
r
ac
tices
in
s
p
ec
tr
al
d
ata
an
al
y
s
is
,
wh
er
e
a
r
ig
o
r
o
u
s
p
r
o
ce
s
s
is
r
e
q
u
ir
ed
f
r
o
m
d
ata
p
r
ep
ar
atio
n
t
o
f
in
al
m
o
d
el
v
alid
atio
n
[
3
2
]
,
[
3
3
]
.
As
d
etailed
i
n
Fig
u
r
e
1
,
th
e
p
r
o
ce
s
s
b
eg
in
s
with
r
ef
lecta
n
ce
ca
p
tu
r
e
(
9
0
0
–
1
,
7
0
0
n
m
)
o
f
7
0
s
am
p
les
o
f
th
e
Hass
an
d
Fu
e
r
te
v
ar
ieties.
I
n
an
in
itial
p
h
ase,
t
h
e
d
ata
u
n
d
er
g
o
a
f
i
r
s
t
p
r
ep
ar
atio
n
p
ip
elin
e
(
Sav
itzk
y
-
Go
la
y
(
SG)
an
d
s
tan
d
ar
d
n
o
r
m
al
v
a
r
iate
(
SNV)
)
.
S
u
b
s
e
q
u
e
n
t
ly
,
a
p
r
i
n
c
i
p
a
l
co
m
p
o
n
en
t
a
n
a
ly
s
i
s
-
b
a
s
ed
o
u
t
l
ie
r
d
e
t
e
c
t
io
n
f
i
l
t
e
r
(
P
C
A
-
O
D
)
w
i
t
h
a
9
0
%
c
o
n
f
i
d
en
c
e
le
v
e
l
a
n
d
f
i
v
e
p
r
i
n
c
i
p
a
l
c
o
m
p
o
n
en
t
s
i
s
ap
p
l
i
ed
i
n
d
i
v
id
u
a
l
l
y
.
I
n
th
i
s
s
t
e
p
,
s
c
a
n
s
i
d
en
t
i
f
i
ed
a
s
o
u
t
l
i
er
s
a
r
e
d
i
s
c
ar
d
ed
,
an
d
t
h
e
r
e
m
a
in
i
n
g
s
p
e
c
tr
a
ar
e
t
h
e
n
a
v
er
a
g
e
d
f
o
r
e
a
ch
m
e
a
s
u
r
e
m
en
t
p
o
i
n
t.
S
u
b
s
e
q
u
e
n
t
ly
,
a
s
e
co
n
d
ch
e
m
o
m
e
t
r
i
c
p
i
p
e
l
i
n
e
i
s
a
p
p
li
e
d
g
l
o
b
a
l
l
y
t
o
a
l
l
s
p
e
c
tr
a
,
i
n
t
eg
r
a
t
in
g
S
N
V
a
n
d
m
in
i
m
u
m
n
o
i
s
e
f
r
a
c
t
io
n
(
M
N
F
)
t
r
a
n
s
f
o
r
m
a
t
i
o
n
s
,
a
s
e
c
o
n
d
SG
-
d
er
i
v
a
t
iv
e
(
SG
-
D2
)
w
i
t
h
a
s
p
e
c
t
r
a
l
c
l
i
p
p
in
g
to
1
,
0
7
0
–
1
,
6
2
7
n
m
,
a
n
d
G
r
a
m
-
S
ch
m
id
t
(
G
S
)
to
en
h
an
c
e
c
r
i
t
ic
a
l
c
h
e
m
i
c
a
l
p
a
t
te
r
n
s
.
T
h
e
s
y
s
t
e
m
's
s
u
c
c
e
s
s
l
i
e
s
i
n
co
r
r
e
l
a
t
i
n
g
th
e
s
e
d
a
t
a
w
i
t
h
th
e
r
ef
e
r
en
ce
v
a
r
i
ab
l
e
o
f
d
r
y
m
a
t
t
er
(
%D
M
)
,
o
b
t
a
i
n
ed
b
y
t
h
e
r
m
a
l
d
r
y
i
n
g
a
t
7
0
°
C
f
o
r
2
4
h
o
u
r
s
,
a
l
l
o
w
in
g
m
a
t
u
r
i
t
y
to
b
e
c
a
t
eg
o
r
i
z
ed
in
t
o
th
r
ee
th
r
e
s
h
o
l
d
s
:
<2
4
.
5
%
,
2
4
.
5
–
2
6
.
5
%
,
an
d
≥
2
6
.
5
%
.
Fi
n
a
l
l
y
,
t
h
e
d
a
t
a
f
e
e
d
s
th
e
M
L
m
o
d
e
l
i
n
g
b
lo
ck
,
w
h
e
r
e
c
l
a
s
s
i
f
i
e
r
s
s
u
ch
a
s
a
d
a
p
t
iv
e
b
o
o
s
t
in
g
(
A
d
a
B
o
o
s
t
)
,
r
an
d
o
m
f
o
r
e
s
t
(
RF
)
,
g
r
a
d
i
en
t
b
o
o
s
t
i
n
g
(
G
B
)
,
S
V
M
,
an
d
k
-
n
e
a
r
e
s
t
n
e
i
g
h
b
o
r
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
3
0
0
-
2
3
1
3
2302
(
K
N
N
)
a
r
e
d
e
p
lo
y
ed
,
o
p
t
i
m
iz
e
d
t
h
r
o
u
g
h
a
2
0
-
f
o
l
d
c
r
o
s
s
-
v
a
l
i
d
a
t
io
n
.
R
o
b
u
s
t
n
e
s
s
i
s
v
er
i
f
i
e
d
u
s
in
g
r
ec
e
i
v
e
r
o
p
er
a
t
i
n
g
ch
a
r
ac
t
e
r
i
s
t
i
c
(
R
O
C
)
c
u
r
v
e
s
an
d
c
o
n
f
u
s
io
n
m
a
tr
i
c
e
s
,
an
d
p
er
f
o
r
m
an
c
e
i
s
e
v
al
u
a
t
e
d
w
i
t
h
m
e
t
r
ic
s
s
u
c
h
a
s
a
r
e
a
u
n
d
e
r
t
h
e
c
u
r
v
e
(
AUC
)
,
c
l
a
s
s
i
f
i
c
a
t
io
n
a
c
cu
r
a
c
y
,
F1
-
s
c
o
r
e
,
ac
c
u
r
a
c
y
,
r
e
ca
l
l
,
an
d
M
a
tt
h
e
w
s
c
o
r
r
e
l
a
t
i
o
n
co
e
f
f
i
c
i
en
t
(
M
C
C
)
.
Fig
u
r
e
1
.
Me
th
o
d
o
lo
g
ical
wo
r
k
f
lo
w
f
o
r
th
e
n
o
n
-
d
estru
ctiv
e
class
if
icatio
n
o
f
av
o
ca
d
o
m
atu
r
ity
u
s
in
g
NI
R
s
p
ec
tr
o
s
co
p
y
an
d
ML
2
.
1
.
Sa
m
ple c
o
llect
io
n a
nd
s
elec
t
io
n
T
h
is
s
tu
d
y
was
co
n
d
u
cted
u
s
in
g
Hass
an
d
Fu
er
te
av
o
ca
d
o
s
am
p
les
co
llected
f
r
o
m
a
ce
r
tifie
d
co
m
m
er
cial
f
ar
m
lo
ca
ted
in
th
e
Say
a
n
–
Hu
ar
a
ar
ea
,
r
e
p
r
esen
tativ
e
o
f
Per
u
’
s
e
x
p
o
r
t
-
o
r
ien
ted
av
o
ca
d
o
p
r
o
d
u
ctio
n
.
Fru
its
wer
e
in
ten
tio
n
ally
s
elec
ted
at
d
if
f
er
en
t
m
atu
r
ity
s
tag
es,
p
r
elim
in
ar
ily
d
eter
m
in
ed
th
r
o
u
g
h
v
is
u
al
in
s
p
ec
tio
n
(
p
ee
l
co
l
o
r
)
,
tactile
f
ir
m
n
ess
,
an
d
p
o
s
th
ar
v
est
co
n
d
itio
n
.
T
h
is
in
ten
tio
n
al
s
tr
atif
icatio
n
en
s
u
r
ed
ad
e
q
u
ate
v
a
r
iab
ilit
y
a
n
d
r
ep
r
esen
tativ
en
ess
o
f
th
e
d
ataset
f
o
r
b
o
th
s
p
ec
tr
al
a
n
aly
s
is
an
d
class
if
icatio
n
m
o
d
elin
g
.
T
h
e
d
is
tr
ib
u
tio
n
o
f
s
am
p
les b
y
v
ar
iety
a
n
d
o
r
ig
in
is
s
u
m
m
ar
ized
in
T
ab
le
1
.
T
ab
le
1
.
Descr
ip
tio
n
o
f
s
am
p
l
es u
s
ed
V
a
r
i
e
t
y
R
e
g
i
o
n
Zo
n
e
Q
u
a
n
t
i
t
y
H
a
ss
F
i
l
e
S
a
y
a
n
70
S
t
r
o
n
g
F
i
l
e
S
a
y
a
n
70
2
.
2
.
Sp
ec
t
ra
l
a
cquis
it
io
n sy
s
t
em
Sp
ec
tr
al
d
ata
wer
e
ac
q
u
ir
ed
u
s
in
g
an
Av
aSp
ec
-
NI
R
-
1
.
7
s
p
ec
tr
o
m
eter
o
p
er
ated
th
r
o
u
g
h
Av
aSo
f
t
8
.
1
6
s
o
f
twar
e.
T
h
e
s
y
s
tem
co
v
e
r
ed
a
wav
elen
g
th
r
an
g
e
o
f
9
0
0
–
1
,
7
0
0
n
m
,
with
a
n
in
teg
r
atio
n
t
im
e
o
f
5
m
s
an
d
a
n
av
er
ag
e
o
f
8
0
s
ca
n
s
p
er
m
ea
s
u
r
em
en
t
an
d
an
a
v
o
ca
d
o
d
is
tan
ce
o
f
3
–
5
m
m
,
p
r
o
v
id
i
n
g
s
u
f
f
icien
t
s
p
ec
tr
al
r
eso
lu
tio
n
to
ca
p
tu
r
e
th
e
m
ai
n
ab
s
o
r
p
tio
n
b
an
d
s
ass
o
ciate
d
with
wate
r
,
lip
id
s
,
ca
r
b
o
h
y
d
r
ate
s
an
d
d
r
y
m
atter
in
av
o
ca
d
o
p
u
lp
[
2
9
]
,
[
3
4
]
.
All m
ea
s
u
r
em
en
ts
wer
e
r
ec
o
r
d
ed
in
ab
s
o
r
b
an
ce
m
o
d
e.
T
o
en
s
u
r
e
m
ea
s
u
r
em
en
t
c
o
n
s
is
ten
cy
an
d
m
i
n
im
ize
v
ar
iab
ili
ty
,
a
f
ib
er
-
o
p
tic
p
r
o
b
e
eq
u
i
p
p
ed
with
an
in
teg
r
ated
h
alo
g
en
lig
h
t
s
o
u
r
c
e
was
p
o
s
itio
n
ed
p
er
p
e
n
d
icu
la
r
ly
to
th
e
f
r
u
it
s
u
r
f
ac
e
u
s
in
g
a
m
ec
h
an
ical
h
o
l
d
er
th
at
m
ain
tain
ed
a
f
i
x
ed
a
cq
u
i
s
itio
n
d
is
tan
ce
.
Prio
r
to
s
p
ec
tr
al
ac
q
u
is
itio
n
,
th
e
s
y
s
tem
was
ca
lib
r
ated
u
s
in
g
a
wh
ite
r
ef
er
en
ce
s
tan
d
ar
d
(
T
ef
lo
n
)
a
n
d
a
d
ar
k
r
e
f
er
en
ce
to
co
r
r
ec
t
f
o
r
e
n
v
ir
o
n
m
en
t
al
an
d
in
s
tr
u
m
en
tal
in
ter
f
er
en
ce
s
.
T
h
e
ex
p
e
r
im
en
t
al
s
etu
p
an
d
in
s
tr
u
m
e
n
tatio
n
a
r
e
illu
s
tr
ated
in
Fig
u
r
e
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
o
mp
a
r
in
g
g
lo
b
a
l a
n
d
va
r
iety
-
s
p
ec
ific e
n
s
emb
le
mo
d
els fo
r
a
vo
ca
d
o
ma
tu
r
ity
…
(
C
h
r
is
tia
n
Ova
lle)
2303
Fig
u
r
e
2
.
Av
aSp
ec
-
NI
R
-
1
.
7
N
I
R
d
ev
ice
2
.
3
.
Sa
m
pli
ng
a
nd
s
pect
ra
l a
cquis
it
io
n
A
s
tan
d
ar
d
ized
s
am
p
lin
g
p
r
o
t
o
co
l
was
d
esig
n
ed
t
o
ca
p
tu
r
e
th
e
r
ep
r
esen
tativ
e
s
p
ec
tr
al
v
ar
iab
ilit
y
o
f
ea
ch
f
r
u
it.
Fo
u
r
eq
u
id
is
tan
t
m
ea
s
u
r
em
en
t
p
o
in
ts
wer
e
m
ar
k
ed
alo
n
g
th
e
eq
u
ato
r
ial
r
eg
io
n
o
f
ea
ch
av
o
ca
d
o
.
At
ea
ch
p
o
in
t,
s
p
ec
tr
a
wer
e
ac
q
u
ir
ed
u
s
in
g
th
e
Av
aSp
ec
-
NI
R
-
1
.
7
s
p
ec
tr
o
m
eter
o
p
er
atin
g
in
d
if
f
u
s
e
r
ef
lecta
n
ce
m
o
d
e
an
d
co
u
p
led
to
a
f
ib
e
r
-
o
p
tic
p
r
o
b
e.
T
o
m
a
x
im
ize
th
e
s
ig
n
al
-
to
-
n
o
is
e
r
ati
o
(
SNR
)
an
d
en
s
u
r
e
r
o
b
u
s
t
s
p
ec
tr
al
m
ea
s
u
r
em
en
ts
,
ea
ch
f
in
al
s
p
ec
tr
u
m
was
o
b
tain
ed
b
y
av
er
ag
i
n
g
8
0
co
n
s
ec
u
tiv
e
s
ca
n
s
at
th
e
s
am
e
p
o
in
t.
T
h
is
p
r
o
ce
d
u
r
e
r
esu
lted
i
n
f
o
u
r
h
i
g
h
-
q
u
ality
r
e
p
r
esen
tativ
e
s
p
ec
tr
a
p
er
f
r
u
it,
wh
ich
wer
e
s
u
b
s
eq
u
en
tly
u
s
ed
f
o
r
ch
em
o
m
etr
ic
an
aly
s
is
an
d
m
o
d
el
d
e
v
elo
p
m
en
t.
All
s
p
ec
tr
al
d
ata
wer
e
co
llected
with
in
th
e
9
0
0
–
1
,
7
0
0
n
m
r
an
g
e
.
T
h
e
s
am
p
le
p
r
e
p
ar
atio
n
a
n
d
s
p
ec
tr
al
ac
q
u
is
itio
n
p
r
o
ce
s
s
is
s
h
o
wn
in
Fig
u
r
e
3
,
wh
er
e
Fig
u
r
e
3
(
a)
s
h
o
ws
th
e
l
ab
el
o
f
s
am
p
les
f
o
r
id
en
tific
at
io
n
,
Fig
u
r
e
3
(
b
)
s
h
o
ws
th
e
m
a
r
k
in
g
o
f
th
e
a
r
ea
o
f
an
aly
s
is
o
f
th
e
s
am
p
les
,
an
d
Fi
g
u
r
e
3
(
c)
s
h
o
ws th
e
ac
q
u
is
itio
n
o
f
s
p
ec
tr
a
.
(
a)
(
b
)
(
c)
Fig
u
r
e
3
.
Pre
p
a
r
atio
n
a
n
d
ac
q
u
is
itio
n
o
f
s
p
ec
tr
a
in
a
v
o
ca
d
o
s
am
p
les
of
(
a)
lab
el
o
f
s
am
p
le
s
f
o
r
id
en
tific
atio
n
,
(
b
)
m
ar
k
in
g
o
f
th
e
ar
ea
o
f
an
al
y
s
is
o
f
th
e
s
am
p
les,
an
d
(
c
)
ac
q
u
is
itio
n
o
f
s
p
ec
tr
a
2
.
4
.
B
a
s
eline
a
na
ly
s
is
a
nd
m
a
t
urit
y
cla
s
s
if
ica
t
io
n
Fo
r
m
o
d
el
t
r
ain
in
g
a
n
d
v
alid
a
tio
n
,
%DM
was
s
elec
ted
as
th
e
r
ef
er
e
n
ce
v
a
r
iab
le
to
d
ef
in
e
av
o
ca
d
o
m
atu
r
ity
s
tag
es.
T
h
is
p
ar
am
et
er
is
wid
ely
r
ec
o
g
n
ized
in
th
e
s
cien
tific
liter
atu
r
e
a
n
d
b
y
th
e
av
o
ca
d
o
i
n
d
u
s
tr
y
as
a
r
eliab
le
in
d
icato
r
o
f
b
o
th
p
h
y
s
io
lo
g
ical
an
d
e
d
ib
le
m
atu
r
ity
[
3
5
]
,
[
3
6
]
.
Dr
y
m
atter
co
n
ten
t
was
d
eter
m
in
ed
u
s
in
g
th
e
s
tan
d
ar
d
g
r
a
v
im
etr
ic
m
et
h
o
d
.
Ap
p
r
o
x
i
m
ately
1
0
g
o
f
h
o
m
o
g
en
ized
p
u
lp
f
r
o
m
ea
ch
f
r
u
it
wer
e
weig
h
ed
u
s
in
g
an
a
n
aly
tical
b
alan
ce
with
a
p
r
ec
is
io
n
o
f
±
0
.
0
0
1
g
an
d
s
u
b
s
eq
u
en
tl
y
d
r
ied
in
a
f
o
r
ce
d
-
air
co
n
v
ec
tio
n
o
v
en
at
7
0
°C
f
o
r
2
4
h
o
u
r
s
u
n
til
co
n
s
tan
t
weig
h
t
was
ac
h
iev
ed
.
T
h
e
co
m
p
lete
d
r
y
in
g
a
n
d
an
al
y
s
is
p
r
o
ce
s
s
is
illu
s
tr
ated
in
Fig
u
r
e
4
,
wh
er
e
Fig
u
r
e
4
(
a)
s
h
o
w
s
th
e
p
r
e
p
ar
atio
n
o
f
f
r
esh
s
a
m
p
les,
Fig
u
r
e
4
(
b
)
s
h
o
ws
th
e
h
ea
t
tr
ea
tm
en
t
i
n
o
v
en
,
an
d
Fig
u
r
e
4
(
c
)
s
h
o
ws
th
e
d
eh
y
d
r
ated
s
am
p
le
s
.
Fin
ally
,
f
o
r
th
e
d
ev
elo
p
m
e
n
t
o
f
th
e
class
if
icatio
n
m
o
d
els,
ea
ch
f
r
u
it
was
ass
ig
n
ed
to
o
n
e
o
f
th
r
ee
ca
te
g
o
r
ies
‘
im
m
atu
r
e,
’
‘
o
p
tim
al
m
atu
r
ity
,
’
o
r
‘
o
v
e
r
r
ip
e’
b
ased
o
n
its
%DM
v
alu
e.
T
h
e
th
r
esh
o
l
d
s
u
s
ed
f
o
r
th
is
ca
teg
o
r
izatio
n
,
d
er
iv
ed
f
r
o
m
c
o
m
m
er
cial
s
tan
d
ar
d
s
o
f
t
h
e
ex
p
o
r
t in
d
u
s
tr
y
[
3
7
]
,
[
3
8
]
,
a
r
e
d
etailed
in
T
ab
le
2
.
T
ab
le
2
.
Av
o
ca
d
o
class
if
icatio
n
th
r
esh
o
ld
s
ac
co
r
d
in
g
to
th
e
%DM
M
a
t
u
r
i
t
y
c
a
t
e
g
o
r
y
%M
S
u
m
b
r
a
l
U
n
r
i
p
e
<
2
4
.
5
%
O
p
t
i
mal
ma
t
u
r
i
t
y
2
4
.
5
%
≤
%D
M
≤
2
6
.
5
%
O
v
e
r
r
i
p
e
≥
2
6
.
5
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
3
0
0
-
2
3
1
3
2304
(
a)
(
b
)
(
c)
Fig
u
r
e
4
.
Dr
y
in
g
p
r
o
ce
s
s
o
f
a
v
o
ca
d
o
s
am
p
les f
o
r
p
h
y
s
ical
-
ch
em
ical
an
aly
s
is
of
(
a)
p
r
ep
ar
at
io
n
o
f
f
r
esh
s
am
p
les
,
(
b
)
h
ea
t tr
ea
tm
e
n
t in
o
v
en
,
a
n
d
(
c
)
d
e
h
y
d
r
ate
d
s
am
p
les
2
.
5
.
P
re
pro
ce
s
s
ing
o
f
s
pect
r
a
l da
t
a
R
aw
NI
R
s
p
ec
tr
a
ar
e
o
f
ten
af
f
ec
ted
b
y
in
s
tr
u
m
en
tal
n
o
is
e,
lig
h
t
s
ca
tter
in
g
ef
f
ec
ts
,
a
n
d
b
aselin
e
v
ar
iatio
n
s
.
T
o
m
itig
ate
th
ese
in
ter
f
er
en
ce
s
a
n
d
e
n
h
an
ce
ch
em
ically
r
elev
an
t
in
f
o
r
m
at
io
n
,
an
o
p
tim
ized
p
r
ep
r
o
ce
s
s
in
g
p
ip
elin
e
was
ap
p
lied
p
r
io
r
to
ch
em
o
m
etr
ic
m
o
d
elin
g
[
3
9
]
–
[
4
1
]
.
T
o
o
p
tim
ize
th
e
SNR
an
d
co
r
r
ec
t
f
o
r
p
h
y
s
ical
d
is
tu
r
b
a
n
ce
s
in
th
e
s
am
p
les,
a
p
r
ep
r
o
c
ess
in
g
wo
r
k
f
lo
w
d
iv
id
ed
in
to
two
cr
itical
p
h
ases
was im
p
lem
en
ted
,
f
o
llo
win
g
a
d
u
al
-
p
ip
elin
e
ar
ch
itectu
r
e
.
i)
I
n
d
iv
id
u
al
clea
n
in
g
p
h
ase:
f
i
r
s
t,
th
e
r
aw
s
p
ec
tr
a
o
f
e
ac
h
m
ea
s
u
r
e
m
en
t
p
o
in
t
u
n
d
er
wen
t
s
ig
n
al
n
o
r
m
aliza
tio
n
u
s
in
g
SG
s
m
o
o
th
in
g
(
5
1
-
p
o
i
n
t
win
d
o
w,
s
ec
o
n
d
-
o
r
d
er
p
o
ly
n
o
m
ial)
,
SNV
s
m
o
o
th
in
g
,
an
d
th
en
SG
an
d
th
e
f
ir
s
t
d
er
iv
ativ
e
(
1
9
-
p
o
in
t
win
d
o
w,
s
ec
o
n
d
-
o
r
d
er
p
o
ly
n
o
m
ial)
.
T
o
en
s
u
r
e
s
am
p
le
p
u
r
ity
,
o
u
tlier
d
etec
tio
n
was
p
er
f
o
r
m
ed
u
s
in
g
PC
A
-
OD
with
a
9
0
%
co
n
f
id
en
ce
lev
el
a
n
d
5
co
m
p
o
n
e
n
ts
.
Scan
s
id
en
tifie
d
as
o
u
tlier
s
wer
e
s
y
s
tem
atica
lly
r
em
o
v
e
d
,
a
n
d
o
n
ly
th
e
s
p
ec
tr
a
v
alid
f
o
r
r
e
p
r
esen
tin
g
ea
ch
f
r
u
it
zo
n
e
wer
e
av
e
r
ag
ed
.
ii)
Glo
b
al
ch
em
o
m
etr
ic
p
ip
elin
e
p
h
ase:
o
n
ce
th
e
d
ataset
was
co
n
s
o
lid
ated
,
a
n
ad
v
an
ce
d
p
ip
elin
e
wa
s
ap
p
lied
.
T
h
e
s
eq
u
en
ce
o
f
th
e
s
ec
o
n
d
p
ip
elin
e
is
illu
s
tr
ated
in
Fig
u
r
e
5
,
w
h
er
e
Fig
u
r
e
5
(
a)
s
h
o
ws
th
e
r
aw
ab
s
o
r
b
an
ce
s
p
ec
tr
a
o
f
all
p
o
in
ts
an
d
v
ar
iety
,
Fig
u
r
e
5
(
b
)
s
h
o
ws
af
ter
SNV
co
r
r
ec
tio
n
,
Fig
u
r
e
5
(
c)
s
h
o
ws
af
ter
MN
F
n
o
is
e
r
ed
u
ctio
n
,
Fig
u
r
e
5
(
d
)
s
h
o
ws
af
ter
t
h
e
s
ec
o
n
d
d
e
r
iv
ativ
e
o
f
SG,
an
d
Fig
u
r
e
5
(
e)
s
h
o
ws
th
e
f
in
al
s
p
ec
tr
a
af
ter
Gau
s
s
ian
s
m
o
o
th
in
g
an
d
s
p
ec
tr
al
clip
p
in
g
; a
n
d
c
o
n
s
is
ted
o
f
th
e
f
o
llo
win
g
s
tep
s
:
‒
SNV:
a
s
a
f
ir
s
t
s
tep
,
SNV
wa
s
ap
p
lied
to
c
o
r
r
ec
t
lig
h
t
s
ca
tter
in
g
v
a
r
iatio
n
s
an
d
o
p
tical
p
a
th
len
g
t
h
d
if
f
er
en
ce
s
am
o
n
g
s
am
p
les,
s
t
an
d
ar
d
izin
g
th
e
s
ca
le
o
f
e
ac
h
s
p
ec
tr
u
m
.
‒
MN
F
d
en
o
is
in
g
:
f
o
r
d
ee
p
er
n
o
is
e
r
ed
u
ctio
n
th
a
n
co
n
v
en
t
io
n
al
s
m
o
o
th
in
g
m
et
h
o
d
s
,
th
e
MN
F
tr
an
s
f
o
r
m
with
f
iv
e
co
m
p
o
n
e
n
ts
was
em
p
lo
y
ed
.
T
h
is
ad
v
a
n
ce
d
tech
n
iq
u
e
was
cr
u
cial
to
s
eg
r
eg
ate
an
d
elim
in
ate
m
u
ltiv
ar
iate
n
o
is
e,
r
etain
in
g
o
n
ly
th
e
co
m
p
o
n
e
n
ts
with
th
e
h
ig
h
est
SNR
f
o
r
s
u
b
s
eq
u
en
t stag
es.
‒
SG
s
m
o
o
th
in
g
an
d
s
ec
o
n
d
d
er
iv
ativ
e:
a
SG
f
ilter
(
1
7
-
p
o
i
n
t w
in
d
o
w,
s
ec
o
n
d
-
o
r
d
e
r
p
o
ly
n
o
m
ial)
was
th
en
ap
p
lied
f
o
r
s
m
o
o
th
i
n
g
an
d
s
im
u
ltan
eo
u
s
ca
lcu
latio
n
o
f
th
e
s
ec
o
n
d
d
er
iv
ativ
e.
T
h
is
co
m
b
in
atio
n
ef
f
ec
tiv
ely
r
em
o
v
es
b
aselin
e
s
h
if
ts
an
d
r
eso
lv
es
o
v
er
lap
p
in
g
p
ea
k
s
,
am
p
lify
i
n
g
s
u
b
tle
d
if
f
er
en
ce
s
ass
o
ciate
d
with
ch
em
ical
co
m
p
o
s
itio
n
.
‒
Gau
s
s
ian
s
m
o
o
th
in
g
:
em
p
ir
ic
al
test
s
s
h
o
wed
th
at
a
f
in
al
s
m
o
o
th
in
g
s
tep
u
s
in
g
a
lo
w
-
im
p
ac
t
Gau
s
s
ian
f
ilter
(
σ
=1
.
0
)
c
o
n
tr
i
b
u
ted
t
o
a
d
d
itio
n
al
s
p
ec
tr
u
m
s
tab
ilizatio
n
,
elim
in
atin
g
r
esid
u
al
h
ig
h
-
f
r
eq
u
e
n
cy
n
o
is
e
in
tr
o
d
u
ce
d
b
y
d
er
iv
atio
n
with
o
u
t d
is
to
r
tin
g
p
ea
k
s
o
f
in
ter
est.
‒
Sp
ec
tr
al
tr
im
m
in
g
:
f
in
ally
,
th
e
s
p
ec
tr
al
r
an
g
e
was
cr
o
p
p
e
d
t
o
1
,
0
7
0
–
1
,
6
2
7
n
m
t
o
f
o
cu
s
th
e
an
aly
s
is
o
n
th
e
m
o
s
t in
f
o
r
m
ativ
e
r
eg
io
n
with
th
e
h
ig
h
est
SNR
f
o
r
d
r
y
m
atter
p
r
e
d
ictio
n
in
a
v
o
ca
d
o
.
I
t
is
ac
k
n
o
wled
g
e
d
th
at
t
h
is
p
ip
elin
e
is
m
o
r
e
elab
o
r
ate
th
a
n
co
n
v
e
n
tio
n
al
p
r
ep
r
o
ce
s
s
in
g
s
eq
u
en
ce
s
.
Ho
wev
er
,
d
u
r
in
g
th
e
m
o
d
el
o
p
tim
izatio
n
p
h
ase,
it
was
co
n
f
ir
m
ed
th
at
o
m
itti
n
g
a
n
y
o
f
th
e
s
e
s
tep
s
p
ar
ticu
lar
ly
MN
F
d
en
o
is
in
g
an
d
th
e
f
in
al
Gau
s
s
ian
s
m
o
o
th
in
g
led
to
a
q
u
an
tifia
b
le
d
ec
r
ea
s
e
in
class
if
i
ca
tio
n
p
er
f
o
r
m
an
c
e
(
lo
wer
AUC
an
d
MC
C
in
c
r
o
s
s
-
v
alid
atio
n
)
.
T
h
e
r
ef
o
r
e
,
th
e
co
m
p
lete
s
eq
u
en
ce
was
r
etain
ed
to
en
s
u
r
e
m
ax
im
u
m
r
eliab
ilit
y
a
n
d
r
o
b
u
s
tn
ess
o
f
th
e
f
in
al
m
o
d
els p
r
esen
ted
in
th
is
wo
r
k
.
2
.
6
.
Dev
el
o
pm
ent
a
nd
ev
a
lu
a
t
io
n o
f
cla
s
s
if
ica
t
io
n
m
o
dels
T
h
e
d
ev
elo
p
m
en
t
o
f
th
e
clas
s
if
ier
s
was
d
esig
n
ed
u
n
d
er
a
p
r
ed
ictiv
e
co
m
p
u
tatio
n
al
ar
ch
itectu
r
e
s
tr
u
ctu
r
ed
ar
o
u
n
d
two
ap
p
r
o
ac
h
es:
i)
a
g
lo
b
al
m
o
d
el
tr
ain
ed
with
co
m
b
in
ed
d
ata
f
r
o
m
b
o
th
‘
Hass
’
an
d
‘
Fu
er
te’
v
a
r
ieties
an
d
ii)
v
ar
iety
-
s
p
ec
if
ic
m
o
d
els
in
d
iv
id
u
ally
ca
lib
r
ated
f
o
r
ea
ch
cu
ltiv
ar
.
T
h
is
m
eth
o
d
o
l
o
g
ical
co
n
t
r
ast
en
ab
led
th
e
ev
alu
atio
n
o
f
wh
eth
e
r
s
p
ec
tr
al
d
if
f
er
e
n
ce
s
b
etwe
e
n
cu
ltiv
ar
s
ju
s
tify
d
is
tin
ct
ca
lib
r
atio
n
s
tr
ateg
ies
[
2
1
]
,
[
4
2
]
,
[
4
3
]
.
T
h
e
p
r
e
d
ictiv
e
p
r
o
ce
s
s
in
g
wo
r
k
f
lo
w
was
b
ased
o
n
a
n
e
n
s
em
b
le
s
tack
in
g
p
ip
elin
e,
in
teg
r
atin
g
m
u
ltip
le
alg
o
r
ith
m
s
as
b
ase
lear
n
er
s
an
d
a
m
eta
-
m
o
d
e
l
f
o
r
co
m
b
i
n
atio
n
.
Sp
ec
if
ically
,
th
e
b
ase
class
if
ier
s
in
clu
d
ed
R
F,
GB
,
XGBo
o
s
t,
ca
teg
o
r
ical
b
o
o
s
tin
g
(
C
at
B
o
o
s
t
)
,
KNN,
an
d
d
ec
is
io
n
tr
ee
s
(
DT
)
,
wh
ile
th
e
f
u
s
io
n
lay
er
c
o
n
s
is
ted
o
f
a
lin
ea
r
r
eg
r
ess
io
n
m
eta
-
m
o
d
el,
r
esp
o
n
s
ib
le
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
o
mp
a
r
in
g
g
lo
b
a
l a
n
d
va
r
iety
-
s
p
ec
ific e
n
s
emb
le
mo
d
els fo
r
a
vo
ca
d
o
ma
tu
r
ity
…
(
C
h
r
is
tia
n
Ova
lle)
2305
o
p
tim
izin
g
th
e
weig
h
ted
co
m
b
in
atio
n
o
f
in
d
iv
i
d
u
al
p
r
ed
ict
io
n
s
.
T
o
en
s
u
r
e
tr
ain
in
g
r
o
b
u
s
tn
ess
an
d
m
itig
ate
o
v
er
f
itti
n
g
,
a
2
0
-
f
o
l
d
cr
o
s
s
-
v
alid
atio
n
s
ch
em
e
was
ap
p
lie
d
i
n
all
ex
p
e
r
im
en
ts
.
T
h
is
s
tatis
ti
ca
l
d
esig
n
en
s
u
r
e
d
th
at
ea
ch
in
s
tan
ce
o
f
th
e
d
ata
s
et
was
u
s
ed
f
o
r
b
o
th
tr
ain
in
g
an
d
v
alid
atio
n
,
th
er
e
b
y
r
ed
u
cin
g
b
ias
in
er
r
o
r
esti
m
atio
n
an
d
s
tr
en
g
th
e
n
in
g
m
o
d
el
g
e
n
er
aliza
tio
n
.
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
class
if
ier
s
was
ev
alu
ated
u
s
in
g
a
s
et
o
f
co
m
p
lem
e
n
tar
y
m
etr
ics:
co
n
f
u
s
io
n
m
atr
ix
,
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
AUC,
an
d
MCC
,
th
e
latter
b
ein
g
p
ar
ticu
lar
ly
r
o
b
u
s
t
in
m
u
lticlas
s
cla
s
s
if
icatio
n
co
n
tex
ts
[
4
4
]
.
Hy
p
e
r
p
ar
am
ete
r
s
f
o
r
ea
ch
alg
o
r
ith
m
wer
e
au
to
m
atica
lly
o
p
tim
ized
with
in
th
e
cr
o
s
s
-
v
alid
atio
n
f
r
a
m
ewo
r
k
t
o
m
a
x
im
ize
p
r
e
d
ictiv
e
p
er
f
o
r
m
a
n
ce
.
I
n
t
h
is
way
,
t
h
e
p
r
esen
t
s
tu
d
y
n
o
t
o
n
ly
co
m
p
ar
es
th
e
ef
f
ec
tiv
en
e
s
s
o
f
g
lo
b
al
v
er
s
u
s
v
ar
iety
-
s
p
ec
if
ic
m
o
d
els
b
u
t
also
v
alid
ates
th
e
r
elev
an
ce
o
f
en
s
em
b
le
s
tack
in
g
as
a
n
ad
v
an
ce
d
m
eth
o
d
o
lo
g
ical
s
tr
at
eg
y
to
im
p
r
o
v
e
t
h
e
s
tab
ilit
y
,
s
ca
lab
ilit
y
,
an
d
in
ter
p
r
etab
ilit
y
o
f
av
o
ca
d
o
m
a
tu
r
ity
class
if
icatio
n
s
y
s
tem
s
.
(
a)
(
b
)
(
c
)
(
d
)
(
e)
Fig
u
r
e
5
.
Vis
u
aliza
tio
n
o
f
th
e
s
p
ec
tr
al
p
r
ep
r
o
ce
s
s
in
g
wo
r
k
f
l
o
w
f
o
r
(
a)
r
aw
a
b
s
o
r
b
an
ce
s
p
ec
tr
a
o
f
all
p
o
in
ts
an
d
v
ar
iety
,
(
b
)
af
ter
SNV
co
r
r
ec
tio
n
,
(
c)
af
ter
MN
F n
o
is
e
r
e
d
u
ctio
n
,
(
d
)
a
f
ter
th
e
s
ec
o
n
d
d
er
iv
ativ
e
o
f
SG
,
an
d
(
e)
f
in
al
s
p
ec
tr
a
af
ter
Gau
s
s
ian
s
m
o
o
th
in
g
an
d
s
p
ec
tr
al
c
lip
p
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
3
0
0
-
2
3
1
3
2306
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
k
e
y
f
in
d
in
g
s
o
f
th
e
s
tu
d
y
,
in
te
g
r
atin
g
th
e
ex
p
lo
r
ato
r
y
an
al
y
s
is
o
f
th
e
s
p
ec
tr
al
d
ata
with
th
e
q
u
a
n
titativ
e
ev
alu
atio
n
o
f
th
e
cla
s
s
if
icatio
n
m
o
d
els.
Fiv
e
ML
alg
o
r
ith
m
s
wer
e
co
m
p
ar
ed
u
n
d
er
two
s
tr
ateg
ic
ap
p
r
o
ac
h
es: a
g
lo
b
al
m
o
d
el
an
d
v
ar
iety
-
s
p
ec
if
ic
m
o
d
els.
I
n
a
d
d
itio
n
,
th
e
im
p
ac
t
o
f
th
e
e
n
s
em
b
le
s
tack
in
g
tech
n
iq
u
e
was
ass
ess
ed
ag
ain
s
t
in
d
iv
id
u
al
m
o
d
els,
with
th
e
aim
o
f
id
e
n
tify
in
g
th
e
m
o
s
t r
o
b
u
s
t a
n
d
ac
cu
r
ate
s
tr
at
eg
y
f
o
r
th
e
n
o
n
-
d
estru
ctiv
e
cla
s
s
if
icatio
n
o
f
av
o
ca
d
o
m
atu
r
it
y
s
tag
es.
3
.
1
.
E
x
plo
ra
t
o
ry
da
t
a
a
na
l
y
s
is
Fig
u
r
e
6
p
r
esen
ts
th
e
av
er
ag
e
p
r
ep
r
o
ce
s
s
ed
NI
R
s
p
ec
tr
a
f
o
r
ea
ch
m
atu
r
ity
class
(
u
n
r
ip
e,
o
p
tim
al
m
atu
r
ity
,
an
d
o
v
er
r
i
p
e
)
ac
r
o
s
s
th
e
wav
elen
g
th
r
a
n
g
e
o
f
1
,
0
7
0
–
1
,
6
2
7
n
m
(
x
-
ax
is
:
wav
elen
g
th
in
n
m
;
y
-
ax
is
:
ab
s
o
r
b
an
ce
)
.
Dis
tin
ct
s
p
ec
tr
al
p
atter
n
s
ar
e
o
b
s
er
v
ed
am
o
n
g
th
e
th
r
ee
m
atu
r
ity
s
tag
es,
in
d
icatin
g
th
at
NI
R
s
p
ec
tr
o
s
co
p
y
ef
f
ec
tiv
ely
ca
p
tu
r
es
th
e
p
h
y
s
ico
ch
em
ical
ch
an
g
es
ass
o
ciate
d
with
th
e
av
o
ca
d
o
r
ip
en
in
g
p
r
o
ce
s
s
,
p
ar
ticu
lar
ly
th
o
s
e
r
elate
d
to
d
r
y
m
atter
ac
cu
m
u
latio
n
a
n
d
co
m
p
o
s
itio
n
al
s
h
if
ts
.
T
o
f
u
r
th
e
r
ex
p
lo
r
e
th
e
in
tr
in
s
ic
s
tr
u
ctu
r
e
o
f
th
e
s
p
ec
tr
al
d
ata
,
PC
A
wa
s
ap
p
lied
.
Fig
u
r
e
7
illu
s
tr
ates
th
e
P
C
A
s
co
r
e
p
lo
t,
wh
er
e
a
clea
r
s
ep
ar
atio
n
o
f
s
am
p
les
alo
n
g
th
e
f
ir
s
t
p
r
in
cip
al
co
m
p
o
n
en
t
(
PC
1
)
is
o
b
s
er
v
ed
.
PC
1
i
s
s
tr
o
n
g
ly
co
r
r
elate
d
with
th
e
m
atu
r
ity
s
tag
e,
with
Un
r
i
p
e
s
am
p
les
(
d
r
y
m
atter
<
2
4
.
5
%
)
clu
s
ter
in
g
p
r
e
d
o
m
in
a
n
tly
o
n
th
e
p
o
s
itiv
e
s
id
e
o
f
PC
1
,
wh
ile
o
p
tim
al
m
atu
r
ity
a
n
d
o
v
e
r
r
ip
e
s
am
p
les o
cc
u
p
y
th
e
n
eg
ativ
e
PC
1
r
eg
io
n
.
Fig
u
r
e
6
.
Av
e
r
ag
e
NI
R
s
p
ec
tr
a
f
o
r
ea
c
h
m
atu
r
ity
class
Fig
u
r
e
7
.
P
r
in
cip
al
co
m
p
o
n
en
t
an
aly
s
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
o
mp
a
r
in
g
g
lo
b
a
l a
n
d
va
r
iety
-
s
p
ec
ific e
n
s
emb
le
mo
d
els fo
r
a
vo
ca
d
o
ma
tu
r
ity
…
(
C
h
r
is
tia
n
Ova
lle)
2307
B
ey
o
n
d
m
at
u
r
ity
-
r
elate
d
s
ep
ar
atio
n
,
th
e
PC
A
also
r
ev
ea
ls
a
p
ar
tial
b
u
t
co
n
s
is
ten
t
s
ep
ar
atio
n
b
etwe
en
th
e
two
av
o
ca
d
o
v
ar
ieties.
Hass
s
am
p
les
ar
e
m
ain
ly
co
n
ce
n
tr
ated
o
n
o
n
e
s
id
e
o
f
th
e
s
p
ec
tr
al
s
p
ac
e,
wh
er
ea
s
Fu
er
te
s
am
p
les
s
p
an
a
b
r
o
a
d
er
r
a
n
g
e
ac
r
o
s
s
m
atu
r
it
y
lev
els.
T
h
is
v
ar
ietal
h
eter
o
g
en
ei
ty
p
r
o
v
id
es
s
tr
o
n
g
em
p
ir
ical
ev
id
en
ce
s
u
p
p
o
r
tin
g
th
e
h
y
p
o
th
esis
th
at
a
s
in
g
l
e
g
lo
b
al
m
o
d
el
m
a
y
b
e
lim
it
ed
b
y
i
n
ter
-
v
a
r
ietal
s
p
ec
tr
al
d
if
f
er
en
ce
s
,
wh
er
ea
s
v
ar
iety
-
s
p
ec
if
ic
m
o
d
els
ca
n
b
etter
ex
p
lo
it
cu
ltiv
ar
-
d
ep
en
d
en
t
s
p
ec
tr
al
s
ig
n
atu
r
es.
T
h
ese
f
in
d
i
n
g
s
d
ir
e
ctly
ju
s
tify
th
e
co
m
p
ar
ativ
e
m
o
d
elin
g
s
tr
ateg
y
ad
o
p
ted
in
th
i
s
s
tu
d
y
.
3
.
2
.
Co
m
pa
riso
n o
f
a
pp
ro
a
c
hes
:
g
lo
ba
l m
o
del v
s
.
s
pecif
ic
m
o
dels
T
o
d
ete
r
m
i
n
e
t
h
e
m
o
s
t
ef
f
ec
ti
v
e
m
o
d
e
li
n
g
a
p
p
r
o
a
c
h
,
a
s
y
s
t
em
atic
c
o
m
p
ar
is
o
n
was
c
o
n
d
u
c
te
d
b
e
twe
en
a
g
l
o
b
a
l
m
o
d
el
an
d
v
a
r
ie
ty
-
s
p
ec
i
f
ic
m
o
d
els
f
o
r
F
u
e
r
t
e
a
n
d
Hass
.
F
o
r
ea
c
h
a
p
p
r
o
a
ch
,
f
iv
e
ML
al
g
o
r
it
h
m
s
we
r
e
im
p
le
m
e
n
t
ed
,
w
it
h
h
y
p
e
r
p
a
r
a
m
et
er
s
o
p
ti
m
i
ze
d
v
ia
c
r
o
s
s
-
v
al
id
a
ti
o
n
t
o
e
n
s
u
r
e
a
f
a
ir
a
n
d
r
o
b
u
s
t
co
m
p
a
r
is
o
n
.
T
h
e
q
u
a
n
tit
ati
v
e
r
esu
lts
a
r
e
s
u
m
m
a
r
iz
e
d
i
n
T
ab
les
3
t
o
5
.
F
o
r
t
h
e
g
l
o
b
al
m
o
d
el
(
T
ab
le
3
)
,
t
h
e
b
e
s
t
p
e
r
f
o
r
m
a
n
c
e
was
ac
h
ie
v
e
d
u
s
in
g
GB
,
wit
h
a
n
MCC
o
f
0
.
5
2
6
a
n
d
a
n
AUC
o
f
0
.
8
6
3
.
W
h
i
le
th
ese
v
al
u
es
in
d
ic
at
e
r
e
aso
n
a
b
l
e
p
r
e
d
ic
ti
v
e
ca
p
ab
ilit
y
,
th
ey
w
er
e
c
o
n
s
is
t
e
n
tl
y
s
u
r
p
ass
e
d
b
y
t
h
e
v
ar
iet
y
-
s
p
ec
if
ic
m
o
d
els
.
T
ab
le
3
.
G
lo
b
al
m
o
d
el
m
et
r
ics r
esu
lts
M
o
d
e
l
s
AUC
A
c
c
u
r
a
c
y
F1
-
sc
o
r
e
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
M
C
C
S
V
M
0
.
8
0
5
0
.
6
7
6
0
.
6
7
6
0
.
6
7
7
0
.
6
7
6
0
.
5
0
5
A
d
a
Bo
o
st
0
.
7
0
8
0
.
6
1
1
0
.
6
1
1
0
.
6
1
1
0
.
6
1
3
0
.
4
1
1
GB
0
.
8
6
3
0
.
6
9
0
0
.
6
8
9
0
.
6
8
9
0
.
6
9
0
0
.
5
2
6
RF
0
.
8
4
7
0
.
6
8
1
0
.
6
8
0
0
.
6
8
1
0
.
6
8
1
0
.
5
1
2
K
N
N
0
.
8
3
4
0
.
6
2
4
0
.
6
2
4
0
.
6
2
4
0
.
6
2
4
0
.
4
2
6
T
ab
le
4
.
Stro
n
g
v
a
r
iety
s
p
ec
if
i
c
m
o
d
el
m
etr
ic
r
esu
lts
M
o
d
e
l
s
AUC
A
c
c
u
r
a
c
y
F1
-
sc
o
r
e
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
M
C
C
S
V
M
0
.
9
1
5
0
.
7
8
9
0
.
7
8
6
0
.
7
8
8
0
.
7
8
9
0
.
6
3
3
A
d
a
Bo
o
st
0
.
8
2
2
0
.
8
0
3
0
.
8
0
2
0
.
8
0
2
0
.
8
0
3
0
.
6
6
2
GB
0
.
9
3
1
0
.
8
1
4
0
.
8
1
2
0
.
8
1
4
0
.
8
1
4
0
.
6
7
9
RF
0
.
9
3
3
0
.
8
1
7
0
.
8
1
5
0
.
8
1
5
0
.
8
1
6
0
.
6
8
4
K
N
N
0
.
9
1
8
0
.
7
8
1
0
.
7
8
2
0
.
7
8
2
0
.
7
8
1
0
.
6
2
7
T
ab
le
5
.
Hass
m
an
if
o
ld
s
p
ec
if
ic
m
o
d
el
m
etr
ic
r
esu
lts
M
o
d
e
l
s
AUC
A
c
c
u
r
a
c
y
F1
-
sc
o
r
e
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
M
C
C
S
V
M
0
.
8
5
8
0
.
7
2
0
0
.
7
2
0
0
.
7
3
2
0
.
7
2
0
0
.
5
8
0
A
d
a
Bo
o
st
0
.
7
3
3
0
.
6
4
1
0
.
6
4
1
0
.
6
4
1
0
.
6
4
1
0
.
4
5
9
GB
0
.
8
7
7
0
.
7
1
4
0
.
7
1
4
0
.
7
1
5
0
.
7
1
4
0
.
5
6
8
RF
0
.
8
6
6
0
.
6
9
1
0
.
6
8
9
0
.
6
9
3
0
.
6
9
1
0
.
5
3
3
K
N
N
0
.
8
4
5
0
.
6
8
1
0
.
6
8
0
0
.
6
8
1
0
.
6
8
1
0
.
5
1
8
Fo
r
th
e
Fu
er
te
v
a
r
iety
(
T
a
b
le
4
)
,
R
F
ac
h
iev
ed
th
e
h
ig
h
est
p
er
f
o
r
m
a
n
ce
,
r
ea
c
h
in
g
a
n
MCC
o
f
0
.
6
8
4
an
d
a
n
AUC
o
f
0
.
9
3
3
.
Similar
ly
,
f
o
r
t
h
e
Hass
v
ar
iety
(
T
ab
l
e
5
)
,
th
e
SVM
class
if
ier
y
ield
ed
th
e
b
est
b
alan
c
e
b
etwe
en
d
is
cr
im
in
atio
n
an
d
s
tab
ilit
y
,
with
an
MCC
o
f
0
.
6
3
3
an
d
a
n
AUC
o
f
0
.
8
5
8
.
T
h
ese
r
esu
lts
clea
r
ly
d
em
o
n
s
tr
ate
th
at
v
ar
iety
-
s
p
ec
i
f
ic
ca
lib
r
atio
n
s
tr
ateg
ies
o
u
tp
er
f
o
r
m
th
e
g
lo
b
al
ap
p
r
o
ac
h
a
cr
o
s
s
all
ev
alu
atio
n
m
etr
ics.
T
ab
le
6
p
r
esen
ts
th
e
o
p
tim
ized
h
y
p
er
p
a
r
am
eter
s
f
o
r
all
ev
alu
ated
m
o
d
els.
N
o
tab
ly
,
th
e
o
p
tim
al
co
n
f
ig
u
r
atio
n
s
d
if
f
e
r
s
u
b
s
tan
t
ially
b
etwe
en
g
lo
b
al
a
n
d
v
a
r
i
ety
-
s
p
ec
if
ic
m
o
d
els,
f
u
r
th
er
r
ein
f
o
r
cin
g
th
e
id
ea
th
at
cu
ltiv
ar
-
d
ep
en
d
en
t
s
p
ec
t
r
al
ch
ar
ac
ter
is
tics
r
eq
u
ir
e
tai
lo
r
ed
m
o
d
elin
g
s
tr
ateg
ies
r
at
h
er
th
a
n
a
u
n
if
ie
d
g
lo
b
al
ca
lib
r
atio
n
.
Fro
m
a
m
eth
o
d
o
lo
g
ical
p
er
s
p
ec
tiv
e,
e
n
s
em
b
le
-
b
ased
alg
o
r
ith
m
s
p
a
r
ticu
lar
ly
GB
an
d
R
F
co
n
s
is
ten
tly
o
u
tp
er
f
o
r
m
ed
s
i
m
p
ler
class
if
ier
s
s
u
ch
as
Ad
aBo
o
s
t
an
d
KNN.
T
h
is
tr
en
d
s
u
g
g
ests
th
at
th
e
r
elatio
n
s
h
ip
b
etwe
en
NI
R
s
p
ec
tr
al
f
ea
tu
r
es
an
d
av
o
ca
d
o
m
atu
r
ity
is
g
o
v
er
n
ed
b
y
co
m
p
lex
,
n
o
n
-
lin
ea
r
in
ter
ac
tio
n
s
th
at
ar
e
b
etter
ca
p
tu
r
ed
b
y
en
s
em
b
le
lear
n
in
g
tec
h
n
iq
u
es.
3
.
3
.
Ana
ly
s
is
o
f
t
he
bes
t
perf
o
rm
a
nce
m
o
dels
Fo
llo
win
g
th
e
co
n
f
ir
m
atio
n
o
f
th
e
s
u
p
er
io
r
ity
o
f
v
ar
iety
-
s
p
ec
if
ic
m
o
d
els,
a
d
etailed
a
n
aly
s
is
wa
s
co
n
d
u
cte
d
o
n
th
e
b
est
-
p
e
r
f
o
r
m
in
g
class
if
ier
f
o
r
ea
c
h
v
ar
ie
ty
.
Fo
r
th
e
Fu
e
r
te
v
ar
iety
,
R
F
d
em
o
n
s
tr
ated
th
e
h
ig
h
est
r
o
b
u
s
tn
ess
an
d
d
is
cr
im
in
atio
n
ca
p
ab
ilit
y
.
Fig
u
r
e
8
s
h
o
ws
th
e
co
n
f
u
s
io
n
m
atr
i
x
,
wh
er
e
a
s
tr
o
n
g
co
n
ce
n
tr
atio
n
o
f
p
r
ed
ictio
n
s
alo
n
g
th
e
m
ain
d
ia
g
o
n
al
in
d
ic
ates
h
ig
h
class
if
icatio
n
ac
cu
r
ac
y
ac
r
o
s
s
m
atu
r
ity
s
tag
es.
Mo
s
t
m
is
class
if
icatio
n
s
o
cc
u
r
r
ed
b
etwe
en
th
e
o
p
tim
al
m
atu
r
ity
an
d
o
v
er
r
ip
e
class
es,
wh
ich
is
co
n
s
is
ten
t
with
th
e
o
v
er
lap
o
b
s
er
v
ed
in
th
e
PC
A
(
Fig
u
r
e
7
)
an
d
r
ef
lects
th
e
co
n
tin
u
o
u
s
n
a
tu
r
e
o
f
th
e
r
ip
en
i
n
g
p
r
o
ce
s
s
.
T
h
e
co
r
r
esp
o
n
d
in
g
R
OC
cu
r
v
es
(
Fig
u
r
e
9
)
f
u
r
t
h
er
v
alid
ate
th
is
r
o
b
u
s
tn
ess
,
with
AUC
v
alu
e
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
3
0
0
-
2
3
1
3
2308
ex
ce
ed
in
g
0
.
9
0
f
o
r
all
m
at
u
r
it
y
class
es,
co
n
f
ir
m
in
g
th
e
m
o
d
el’
s
s
tr
o
n
g
d
is
cr
im
in
ato
r
y
p
o
wer
.
I
n
t
h
e
ca
s
e
o
f
th
e
Hass
v
ar
iety
,
th
e
SVM
c
lass
if
ier
ac
h
iev
ed
th
e
b
est
co
m
p
r
o
m
is
e
b
etwe
en
ac
cu
r
ac
y
an
d
s
tab
ilit
y
.
T
h
e
co
n
f
u
s
io
n
m
atr
ix
p
r
esen
te
d
i
n
Fig
u
r
e
1
0
in
d
icate
s
h
ig
h
s
en
s
itiv
ity
f
o
r
th
e
o
p
tim
al
m
atu
r
ity
class
(
8
0
.
2
%),
wh
ile
p
er
f
o
r
m
an
ce
f
o
r
th
e
e
x
tr
em
e
class
es wa
s
m
o
r
e
m
o
d
e
r
a
te:
6
9
.
8
% f
o
r
u
n
r
ip
e
an
d
6
4
.
1
% f
o
r
o
v
e
r
r
ip
e.
T
ab
le
6
.
Mo
d
el
-
s
p
ec
if
ic
an
d
g
lo
b
al
h
y
p
er
p
ar
a
m
eter
s
M
o
d
e
l
s
S
t
r
o
n
g
H
a
t
r
e
d
G
l
o
b
a
l
S
V
M
C
o
s
t
:
1
0
Ep
si
l
o
n
r
e
g
r
e
ssi
o
n
:
0
.
1
0
K
e
r
n
e
l
R
B
F
g
:
1
.
0
5
C
o
s
t
:
8
.
5
Ep
si
l
o
n
r
e
g
r
e
ssi
o
n
:
0
.
2
0
K
e
r
n
e
l
R
B
F
g
:
1
.
0
5
C
o
s
t
:
1
.
0
5
Ep
si
l
o
n
r
e
g
r
e
ssi
o
n
:
0
.
2
0
K
e
r
n
e
l
R
B
F
g
:
0
.
9
7
A
d
a
Bo
o
st
N
u
mb
e
r
o
f
e
s
t
i
m
a
t
o
r
s:
50
La
r
n
i
n
g
r
a
t
e
:
0
.
0
1
0
Lo
s
s
r
e
g
r
e
ssi
o
n
:
s
q
u
a
r
e
N
u
mb
e
r
o
f
e
s
t
i
m
a
t
o
r
s:
50
La
r
n
i
n
g
r
a
t
e
:
0
.
0
1
0
Lo
s
s r
e
g
r
e
ssi
o
n
:
s
q
u
a
r
e
N
u
mb
e
r
o
f
e
s
t
i
m
a
t
o
r
s:
50
La
r
n
i
n
g
r
a
t
e
:
0
.
0
1
0
Lo
s
s r
e
g
r
e
ssi
o
n
:
s
q
u
a
r
e
GB
M
e
t
h
o
d
:
X
G
B
o
o
st
n
u
m
b
e
r
o
f
t
r
e
e
s:
2
0
0
Le
a
r
n
i
n
g
r
a
t
e
:
0
.
3
0
M
e
t
h
o
d
:
X
G
B
o
o
st
n
u
m
b
e
r
o
f
t
r
e
e
s:
2
0
0
Le
a
r
n
i
n
g
r
a
t
e
:
0
.
3
0
M
e
t
h
o
d
:
X
G
B
o
o
st
n
u
m
b
e
r
o
f
t
r
e
e
s:
1
0
0
Le
a
r
n
i
n
g
r
a
t
e
:
0
.
3
0
RF
N
u
mb
e
r
o
f
t
r
e
e
s:
350
N
u
mb
e
r
o
f
t
r
e
e
:
250
N
u
mb
e
r
o
f
t
r
e
e
:
100
K
N
N
N
u
mb
e
r
o
f
n
e
i
g
h
b
o
r
s:
3
M
e
t
r
i
c
:
E
u
c
l
i
d
e
a
n
W
e
i
g
h
t
:
b
y
d
i
s
t
a
n
c
e
s
N
u
mb
e
r
o
f
n
e
i
g
h
b
o
r
s:
3
M
e
t
r
i
c
:
E
u
c
l
i
d
e
a
n
W
e
i
g
h
t
:
b
y
d
i
s
t
a
n
c
e
s
N
u
mb
e
r
o
f
n
e
i
g
h
b
o
r
s:
8
M
e
t
r
i
c
:
E
u
c
l
i
d
e
a
n
W
e
i
g
h
t
:
b
y
d
i
s
t
a
n
c
e
s
Fig
u
r
e
8
.
S
tr
o
n
g
av
o
ca
d
o
co
n
f
u
s
io
n
m
atr
ix
T
h
e
an
aly
s
is
o
f
th
e
m
atr
i
x
r
e
v
ea
ls
h
ig
h
s
en
s
itiv
ity
f
o
r
th
e
‘
r
i
p
e’
class
(
2
4
.
5
–
2
6
.
5
)
,
with
an
ac
cu
r
ac
y
o
f
8
0
.
2
%.
Ho
wev
e
r
,
p
er
f
o
r
m
a
n
ce
was
m
o
r
e
m
o
d
e
r
ate
f
o
r
th
e
ex
tr
em
e
class
es,
with
s
en
s
itiv
ities
o
f
6
9
.
8
%
f
o
r
‘
u
n
r
ip
e
’
(
<
2
4
.
5
)
a
n
d
6
4
.
1
%
f
o
r
‘
o
v
er
r
ip
e’
(
≥
2
6
.
5
)
.
No
ta
b
ly
,
th
e
m
ain
s
o
u
r
ce
o
f
er
r
o
r
f
o
r
b
o
th
‘
u
n
r
ip
e’
an
d
‘
o
v
er
r
i
p
e’
was
th
eir
m
is
class
if
icatio
n
as
‘
r
ip
e’
(
2
5
.
0
%
an
d
2
8
.
3
%
o
f
th
e
tim
e,
r
esp
ec
tiv
ely
)
.
T
h
is
s
u
g
g
ests
th
at,
f
o
r
th
e
‘
Hass
’
v
a
r
iety
,
th
e
s
p
ec
tr
al
ch
ar
ac
ter
is
tics
o
f
th
e
in
ter
m
ed
iate
class
ten
d
to
o
v
er
lap
s
ig
n
if
ican
tly
with
th
o
s
e
o
f
th
e
ea
r
l
y
an
d
la
te
s
tag
es,
p
r
esen
tin
g
a
m
o
r
e
c
o
m
p
lex
class
if
icatio
n
ch
allen
g
e
co
m
p
ar
ed
to
th
e
‘
Fu
er
te’
v
ar
iety
.
T
h
e
m
o
d
el
d
em
o
n
s
tr
ates
s
tr
o
n
g
d
is
cr
im
in
atio
n
a
b
ilit
y
ac
r
o
s
s
all
class
es,
with
AUC
v
alu
es
o
f
0
.
8
7
1
f
o
r
‘
u
n
r
ip
e
,
’
0
.
8
9
8
f
o
r
‘
r
ip
e
,
’
an
d
0
.
8
6
3
f
o
r
‘
o
v
er
r
ip
e.
’
T
h
es
e
co
n
s
is
ten
tly
h
ig
h
v
alu
es
co
n
f
ir
m
th
at
th
e
SVM
m
o
d
el
p
o
s
s
ess
e
s
s
tr
o
n
g
p
r
e
d
ictiv
e
p
o
ten
tial
f
o
r
th
e
‘
Hass
’
v
ar
iety
.
T
h
e
ap
p
ar
en
t
d
is
cr
ep
an
cy
b
etwe
en
th
e
h
ig
h
AUC
an
d
t
h
e
m
o
r
e
m
o
d
er
ate
s
en
s
itiv
ities
in
th
e
ex
tr
e
m
e
class
es
in
d
icate
s
th
at,
alth
o
u
g
h
th
e
m
o
d
el
is
ef
f
ec
tiv
e
at
r
an
k
in
g
p
r
o
b
ab
il
ities
,
d
ef
in
in
g
a
s
in
g
le
cu
to
f
f
th
r
esh
o
l
d
f
o
r
f
in
al
class
if
ic
atio
n
is
in
h
e
r
en
tly
ch
allen
g
in
g
d
u
e
to
class
o
v
er
lap
.
T
o
co
n
tex
tu
alize
th
e
f
i
n
d
in
g
s
o
f
th
is
s
tu
d
y
,
it
is
ess
en
tial
to
co
m
p
ar
e
th
e
m
with
th
e
cu
r
r
en
t
s
tate
o
f
th
e
ar
t.
Pre
v
io
u
s
r
esear
c
h
o
n
NI
R
s
p
ec
tr
o
s
co
p
y
f
o
r
av
o
ca
d
o
q
u
ality
ass
ess
m
en
t
h
as
p
r
ed
o
m
in
an
tly
em
p
lo
y
ed
r
eg
r
ess
io
n
-
b
ased
m
eth
o
d
s
,
p
ar
tic
u
lar
ly
PLS,
to
p
r
ed
ict
q
u
a
n
titativ
e
p
ar
am
eter
s
s
u
ch
as
d
r
y
m
atter
co
n
ten
t.
R
ep
o
r
ted
R
2
co
ef
f
icien
ts
ty
p
ically
r
an
g
e
f
r
o
m
0
.
6
2
to
0
.
9
8
u
n
d
er
co
n
tr
o
lled
la
b
o
r
ato
r
y
co
n
d
itio
n
s
[
1
0
]
,
[
2
6
]
,
[
2
8
]
.
Alth
o
u
g
h
th
e
s
e
m
o
d
els
h
av
e
p
r
o
v
e
n
v
alu
a
b
le
f
o
r
q
u
an
titativ
e
p
r
e
d
ictio
n
,
th
eir
ap
p
licatio
n
in
in
d
u
s
tr
ial
en
v
ir
o
n
m
e
n
ts
r
em
ain
s
lim
ited
d
u
e
to
th
e
n
ee
d
f
o
r
f
r
eq
u
e
n
t
r
ec
alib
r
atio
n
an
d
th
e
ch
allen
g
es
ass
o
ciate
d
with
tr
an
s
f
er
r
in
g
m
o
d
els
to
d
y
n
am
ic,
r
ea
l
-
tim
e
class
if
icat
io
n
s
y
s
tem
s
.
I
n
c
o
n
tr
ast,
th
e
p
r
esen
t
s
tu
d
y
ap
p
r
o
ac
h
es
th
e
p
r
o
b
lem
f
r
o
m
a
class
if
icatio
n
p
er
s
p
ec
tiv
e,
wh
ich
is
m
o
r
e
co
m
p
at
ib
le
with
au
to
m
ated
s
o
r
tin
g
s
y
s
tem
s
an
d
p
o
s
th
ar
v
e
s
t q
u
ality
m
an
ag
em
e
n
t o
p
e
r
atio
n
s
.
R
eg
ar
d
in
g
t
h
e
e
n
s
em
b
le
s
tack
in
g
s
tr
ateg
y
,
r
esu
lts
d
em
o
n
s
tr
ated
th
at
co
m
b
i
n
in
g
m
u
ltip
le
alg
o
r
ith
m
s
co
n
s
is
ten
tly
o
u
tp
er
f
o
r
m
ed
in
d
iv
id
u
al
class
if
ier
s
,
im
p
r
o
v
in
g
b
o
th
ac
cu
r
ac
y
an
d
p
r
ed
ictiv
e
s
tab
ilit
y
.
T
h
is
tr
en
d
alig
n
s
with
f
in
d
in
g
s
f
r
o
m
r
ec
en
t
s
tu
d
ies
o
n
tr
o
p
ical
f
r
u
it
cl
ass
if
icatio
n
u
s
in
g
ML
[
3
0
]
,
wh
er
e
s
tack
in
g
-
b
ased
f
r
am
ewo
r
k
s
ac
h
iev
e
d
ac
cu
r
ac
y
g
ain
s
o
f
5
–
1
0
%
co
m
p
ar
ed
t
o
b
aselin
e
m
o
d
els.
I
n
o
u
r
ca
s
e
,
en
s
em
b
le
s
tack
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
C
o
mp
a
r
in
g
g
lo
b
a
l a
n
d
va
r
iety
-
s
p
ec
ific e
n
s
emb
le
mo
d
els fo
r
a
vo
ca
d
o
ma
tu
r
ity
…
(
C
h
r
is
tia
n
Ova
lle)
2309
also
r
ed
u
ce
d
in
te
r
-
ca
m
p
ai
g
n
v
ar
iab
ilit
y
an
d
en
h
a
n
ce
d
m
o
d
el
r
o
b
u
s
tn
ess
f
o
r
u
n
d
er
r
ep
r
esen
ted
m
atu
r
ity
ca
teg
o
r
ies an
im
p
o
r
tan
t f
ac
to
r
f
o
r
o
p
er
atio
n
al
d
ep
lo
y
m
en
t in
h
eter
o
g
en
e
o
u
s
p
r
o
d
u
ctio
n
co
n
tex
ts
.
Fig
u
r
e
9
.
St
r
o
n
g
av
o
ca
d
o
R
OC
cu
r
v
e
ch
ar
t
f
o
r
ea
c
h
m
atu
r
it
y
th
r
esh
o
ld
Fig
u
r
e
1
0
.
Hass
av
o
ca
d
o
c
o
n
f
u
s
io
n
m
atr
ix
W
h
en
co
m
p
ar
e
d
to
p
r
ev
io
u
s
PLS
-
DA
m
o
d
els
ap
p
lied
to
av
o
ca
d
o
[
4
5
]
,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
ac
h
iev
ed
a
m
o
r
e
b
alan
ce
d
a
n
d
r
eliab
le
class
if
icatio
n
ac
r
o
s
s
all
ca
teg
o
r
ies.
W
h
er
ea
s
th
e
r
ef
er
e
n
ce
d
s
tu
d
y
r
ep
o
r
ted
lim
ited
p
er
f
o
r
m
an
ce
f
o
r
in
ter
m
ed
iate
m
at
u
r
ity
(
F1
-
s
co
r
e
=0
.
3
2
)
,
th
e
m
o
d
el
o
b
tain
ed
F1
v
alu
es
ex
ce
ed
in
g
0
.
7
8
,
al
o
n
g
with
h
i
g
h
MCC
an
d
AUC
s
co
r
es
ac
r
o
s
s
clas
s
es.
T
h
ese
r
esu
lt
s
in
d
icate
th
at
ad
v
an
ce
d
ML
alg
o
r
ith
m
s
p
ar
ticu
lar
ly
wh
en
co
m
b
in
ed
th
r
o
u
g
h
s
tack
in
g
a
r
e
m
o
r
e
ca
p
ab
le
o
f
m
an
ag
in
g
o
v
e
r
lap
p
in
g
s
p
ec
tr
al
p
atter
n
s
an
d
th
e
in
h
er
en
t c
o
m
p
lex
ity
o
f
th
e
av
o
ca
d
o
r
ip
en
in
g
p
r
o
ce
s
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.