I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
6
,
D
e
c
e
m
be
r
2025
, pp.
5240
~
5250
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
6
.pp
5240
-
5250
5240
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
A
w
e
b
-
b
ase
d
l
e
ar
n
i
n
g p
l
at
f
or
m
t
o ass
e
ss
st
u
d
e
n
t
p
e
r
f
or
m
an
c
e
u
si
n
g o
n
l
i
n
e
se
ss
i
on
ac
t
i
vi
t
y e
n
gage
m
e
n
t
S
h
as
h
ir
e
k
h
a
H
an
u
m
an
t
h
ap
p
a
1
, C
h
e
t
an
a
P
r
ak
as
h
2
1
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd E
ngi
ne
e
r
i
ng, V
i
s
ve
s
va
r
a
ya
T
e
c
hnol
ogi
c
a
l
U
ni
ve
r
s
i
t
y, M
ys
or
e
, I
ndi
a
2
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd E
ngi
ne
e
r
i
ng,
B
a
puj
i
I
ns
t
i
t
ut
e
of
E
ngi
ne
e
r
i
ng a
nd T
e
c
hnol
ogy
, D
a
va
na
ge
r
e
, I
ndi
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
M
a
r
29
,
2024
R
e
vi
s
e
d
M
a
r
15
,
2025
A
c
c
e
pt
e
d
J
un
6
,
2025
Predic
ting
student
s'
perfo
rmanc
e
and
engag
ement
is
cruc
ial
for
ac
ademic
eLearning
partners
in
colleges
and
universi
ties
as
well
as
st
udents
themselves
considering
post
-
COVID
-
19
pandemic
and
university
grant
commis
sion
(UGC)
dual
degree
regulati
on
era.
An
education
al
system'
s
data
on
students’
engagement
in
taking
courses
that
are
a
significant
component
of
an
institution
of
higher
learning
with
a
cogent
vertical
syllabus
can
be
used
to
make
predictions.
By
examining
how
closely
a
student'
s
c
ourse
-
taking
act
ions
correspond
with
the
requirements
of
the
syllabus,
o
ne
can
utilize
the
student'
s
conduct
in
the
cl
assroom
and
online
e
Learning
w
eb
tool
as a predi
ctor of
future ach
ievement.
This p
aper
presents
a study
that
uses an
eLearning
web
-
based
dataset
to
predict
students'
success
throughout
a
series
of
online
interactive
sessions.
The
dataset
records
how
students
engag
e
with
each
other
during
online
lab
work,
including
how
many
keystroke
s
they
make,
how
long
they
spend
on
each
task,
and
how
well
they
perfo
rm
on
exams
overall.
The
current
methods
lack
ac
curacy
to
assess
s
tudent
performance
and
engagement
with
high
precision.
In
addressing
this
paper
introduces
novel
multi
-
label
ensemble
learning
(MLEL)
using
X
GBoost
(XGB)
and
K
-
fold
cross
validation.
Experiment
outcome
sho
ws
the
proposed
(MLEL
-
XGB)
achieves
much
improved
outcome
than
other
existin
g model
s.
K
e
y
w
o
r
d
s
:
C
la
s
s
i
m
ba
la
n
c
e
E
-
le
a
r
ni
ng w
e
b por
ta
l
E
ns
e
m
bl
e
a
lg
or
it
hm
F
e
a
tu
r
e
i
m
por
ta
nc
e
M
a
c
hi
ne
l
e
a
r
ni
ng
M
ul
ti
-
la
be
l
c
la
s
s
if
ic
a
ti
on
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
S
ha
s
hi
r
e
kha
H
a
num
a
nt
ha
ppa
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
S
c
ie
nc
e
a
nd E
ngi
ne
e
r
in
g, V
is
ve
s
v
a
r
a
ya
T
e
c
hnol
ogi
c
a
l
U
ni
ve
r
s
it
y
M
ys
or
e
, I
ndi
a
E
m
a
il
:
s
ha
s
hi
r
e
kha
_h2k22@
r
e
di
f
f
m
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
T
he
w
id
e
s
pr
e
a
d
u
s
e
of
th
e
in
te
r
ne
t
a
nd
th
e
de
ve
lo
pm
e
nt
of
i
nf
or
m
a
ti
on
te
c
hnol
ogy
ha
ve
ha
d
a
n
im
pa
c
t
on
how
bu
s
in
e
s
s
e
s
a
nd
a
c
a
de
m
ic
s
ga
in
knowle
dge
,
m
ovi
ng
a
w
a
y
f
r
om
s
ta
nd
a
r
d
of
f
li
ne
m
ode
ls
to
vi
r
tu
a
l
one
s
th
r
ough
th
e
us
e
of
e
L
e
a
r
ni
ng
w
e
b
-
ba
s
e
d/
a
ppl
ic
a
ti
on
-
ba
s
e
d
pl
a
tf
or
m
s
[
1]
.
T
he
e
nt
ir
e
c
ur
r
ic
ul
um
ha
s
s
w
it
c
he
d
to
a
n
onl
in
e
f
or
m
a
t,
pa
r
ti
c
ul
a
r
ly
dur
in
g
th
e
C
O
V
I
D
-
19
e
pi
de
m
ic
,
e
m
pha
s
iz
in
g
th
e
im
por
ta
nc
e
of
e
L
e
a
r
ni
ng
pl
a
t
f
or
m
s
.
A
lo
ngs
id
e
,
uni
ve
r
s
it
y
g
r
a
nt
c
om
m
is
s
io
n
(
U
G
C
)
a
ll
ow
s
s
tu
de
nt
to
e
nr
ol
l
f
or
tw
o
de
gr
e
e
s
c
ons
id
e
r
in
g
one
c
onve
nt
io
na
l
of
f
li
ne
a
nd
ot
he
r
th
r
oug
h
onl
in
e
por
ta
l.
T
hus
,
it
is
im
por
ta
nt
to
a
s
s
e
s
s
th
e
s
tu
de
nt
e
nga
ge
m
e
nt
l
e
ve
l
a
nd
th
e
ir
pe
r
f
or
m
a
nc
e
by
e
m
pl
oyi
ng
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
te
c
hni
que
s
.
N
one
th
e
le
s
s
,
th
e
r
e
a
r
e
a
lo
t
of
obs
ta
c
le
s
in
th
e
w
a
y
of
de
ve
lo
pi
ng
a
le
gi
ti
m
a
te
a
nd
pr
e
c
is
e
m
ode
l
to
f
or
e
c
a
s
t
s
tu
de
nt
e
nga
ge
m
e
nt
le
ve
l
a
nd
pe
r
f
or
m
a
nc
e
[
2]
.
B
y
of
f
e
r
i
ng
in
di
vi
dua
li
z
e
d
in
f
o
r
m
a
ti
on,
a
n
e
f
f
ic
ie
nt
e
va
lu
a
ti
on t
e
c
hni
que
f
or
c
om
pr
e
he
ndi
ng s
tu
de
nt
be
ha
vi
or
vi
a
e
L
e
a
r
ni
ng pla
tf
or
m
s
tu
de
nt
e
nga
ge
m
e
nt
s
e
s
s
io
n
s
tr
e
a
m
s
c
a
n h
e
lp
i
m
pr
ove
s
tu
de
nt
s
'
a
c
a
de
m
ic
pe
r
f
or
m
a
nc
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
w
e
b
-
bas
e
d l
e
ar
ni
ng plat
fo
r
m
t
o as
s
e
s
s
s
tu
d
e
nt
pe
r
fo
r
m
anc
e
u
s
in
g onli
ne
s
e
s
s
io
n
…
(
Shas
hi
r
e
k
ha
)
5241
O
ne
of
th
e
m
a
in
pr
obl
e
m
s
of
th
e
twe
nt
y
-
f
ir
s
t
c
e
nt
ur
y
i
s
de
li
ve
r
in
g
pe
r
s
ona
li
z
e
d
c
ont
e
nt
to
s
tu
de
nt
s
in
a
n
e
-
le
a
r
ni
ng
pl
a
tf
o
r
m
ba
s
e
d
on
th
e
ir
uni
que
be
ha
vi
or
of
s
tu
de
nt
by
a
s
s
e
s
s
in
g
th
e
ir
e
nga
ge
m
e
nt
[
3]
.
T
he
us
e
of
a
da
pt
iv
e
pe
r
s
ona
li
z
a
ti
on
te
c
hni
que
s
to
c
om
pr
e
he
nd
le
a
r
ne
r
pr
of
il
e
s
ha
s
be
e
n
hi
ghl
ig
ht
e
d
[
4]
,
[
5]
.
I
n
r
e
c
e
nt
ye
a
r
s
,
m
ode
ls
f
or
a
na
ly
z
in
g s
tu
de
nt
e
nga
g
e
m
e
nt
le
ve
l
a
nd
pr
e
di
c
ti
ng
s
tu
de
nt
pe
r
f
or
m
a
nc
e
a
r
e
be
in
g
c
r
e
a
te
d
us
in
g
da
ta
m
in
in
g
a
nd
m
a
c
hi
n
e
le
a
r
ni
ng
[
6]
.
T
hr
ough
th
e
e
s
ta
bl
is
hm
e
nt
of
be
ha
vi
or
pa
tt
e
r
ns
f
r
om
e
nga
ge
m
e
nt
s
e
s
s
io
n
d
a
ta
[
7]
,
[
8]
,
da
ta
m
in
in
g
ha
s
be
e
n
ut
il
iz
e
d
f
or
e
nha
nc
in
g
pr
e
di
c
ti
ve
m
ode
l
e
f
f
ic
ie
nc
y
[
9]
a
nd
to
ga
in
va
lu
a
bl
e
in
s
ig
ht
s
f
r
om
s
tu
de
nt
e
ng
a
ge
m
e
nt
s
e
s
s
io
n
in
f
or
m
a
ti
on
of
e
L
e
a
r
ni
ng
w
e
b
-
ba
s
e
d
to
ol
s
.
T
he
a
ppr
oa
c
he
s
of
da
ta
m
in
in
g
[
10]
a
nd ma
c
hi
ne
l
e
a
r
ni
ng
[
11]
–
[
13]
s
how
gr
e
a
t
pr
om
is
e
i
n a
va
r
ie
ty
of
f
ie
ld
s
,
in
c
lu
di
ng
e
nt
e
r
pr
is
e
s
a
nd
in
f
or
m
a
ti
on
s
e
c
ur
it
y,
w
hi
c
h
in
c
lu
d
e
s
e
du
c
a
ti
on
da
ta
m
in
in
g
(
E
D
M
)
[
14]
–
[
16]
.
I
n
th
e
ne
xt
s
ub
-
s
e
c
ti
on
th
is
r
e
s
e
a
r
c
h
w
or
k
s
tu
di
e
s
th
e
va
r
io
us
r
e
c
e
nt
m
e
th
odol
ogi
e
s
de
s
ig
ne
d
to
a
n
a
ly
z
e
th
e
pe
r
f
or
m
a
nc
e
s
tu
de
nt
us
in
g
m
a
c
hi
ne
le
a
r
ni
ng
a
nd
de
e
p
le
a
r
ni
ng
ha
ve
be
e
n
s
tu
di
e
d
a
nd
hi
ghl
ig
ht
th
e
pr
obl
e
m
a
nd mot
iv
a
ti
on of
t
he
r
e
s
e
a
r
c
h w
or
k.
E
D
M
[
17]
ha
s
e
m
e
r
ge
d
a
ne
w
c
onc
e
pt
f
or
e
nha
nc
in
g
le
a
r
ni
n
g
s
ty
le
[
18]
,
unde
r
s
ta
ndi
ng
be
ha
vi
or
,
e
nga
ge
m
e
nt
le
ve
l
[
19]
,
a
nd
im
pr
ovi
ng
s
tu
de
nt
pe
r
f
or
m
a
nc
e
[
20]
.
T
he
E
D
M
da
ta
is
c
om
pos
e
d
of
di
f
f
e
r
e
nt
in
f
or
m
a
ti
on
[
21]
s
uc
h
a
s
a
dm
in
is
tr
a
ti
on
da
ta
,
s
tu
de
nt
s
e
s
s
io
n
s
tr
e
a
m
a
c
ti
vi
ty
[
22]
,
a
nd
s
tu
de
nt
a
c
a
d
e
m
ic
pe
r
f
or
m
a
nc
e
a
nd
e
nga
ge
m
e
nt
in
f
or
m
a
ti
on.
U
s
in
g
th
e
r
e
s
ul
ts
of
th
e
ir
m
id
te
r
m
e
xa
m
s
a
s
pr
im
a
r
y
da
ta
,
th
e
s
tu
dy
a
im
s
to
pr
e
di
c
t
unde
r
gr
a
dua
te
s
tu
de
nt
s
'
f
in
a
l
e
xa
m
m
a
r
k
s
.
T
o
ge
ne
r
a
te
pr
e
di
c
ti
ons
,
it
us
e
s
a
va
r
ie
ty
of
m
a
c
hi
ne
le
a
r
ni
ng
te
c
hni
que
s
,
in
c
lu
di
ng
a
s
r
a
ndom
f
or
e
s
t
(
R
F
)
,
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
,
n
a
ïv
e
B
a
ye
s
(
N
B
)
,
K
-
ne
a
r
e
s
t
ne
ig
hbor
s
(
K
N
N
)
,
a
nd
lo
gi
s
ti
c
r
e
gr
e
s
s
io
n
(
L
R
)
.
T
he
a
c
a
de
m
ic
a
c
c
om
pl
is
hm
e
nt
gr
a
de
s
of
1854
s
tu
de
nt
s
e
nr
ol
le
d
in
a
T
ur
ki
s
h
l
a
ngua
ge
-
I
c
our
s
e
m
a
ke
up
th
e
da
ta
s
e
t.
W
it
h
ju
s
t
th
r
e
e
pa
r
a
m
e
te
r
s
—
m
id
te
r
m
e
xa
m
gr
a
de
s
,
de
pa
r
tm
e
nt
da
ta
,
a
nd
f
a
c
ul
ty
da
ta
—
th
e
s
ugge
s
te
d
m
ode
l
w
a
s
a
bl
e
to
c
la
s
s
if
y
obj
e
c
ts
w
it
h
a
n
a
c
c
ur
a
c
y
of
be
twe
e
n
70
a
nd
75
pe
r
c
e
nt
.
E
s
ta
bl
is
hi
ng
a
le
a
r
ni
ng
a
na
ly
s
is
f
r
a
m
e
w
or
k
in
h
ig
he
r
e
duc
a
ti
on
a
nd
s
uppor
ti
ng
de
c
is
io
n
-
m
a
ki
ng
pr
oc
e
s
s
e
s
—
pa
r
ti
c
ul
a
r
ly
in
id
e
nt
if
yi
ng
h
ig
h
-
r
is
k
s
tu
de
nt
s
f
or
f
a
il
ur
e
—
a
r
e
m
a
de
pos
s
ib
le
by
th
e
f
in
di
ngs
of
th
is
s
tu
dy.
I
n
a
c
c
or
da
nc
e
w
it
h
S
a
udi
A
r
a
bi
a
n
onl
in
e
le
a
r
ni
ng
tr
a
in
in
g
r
e
gul
a
ti
ons
,
th
is
s
tu
dy
pr
e
s
e
nt
s
a
m
a
c
hi
ne
le
a
r
ni
ng
s
tr
a
te
gy
to
f
or
e
c
a
s
t
s
tu
de
nt
pe
r
f
or
m
a
nc
e
in
a
n
onl
in
e
le
a
r
ni
ng
e
nvi
r
onm
e
nt
vi
a
th
e
M
a
h
a
r
a
t
pl
a
tf
or
m
a
t
T
a
if
U
ni
ve
r
s
it
y.
H
ybr
id
opt
im
iz
a
ti
on
is
us
e
d
f
or
f
e
a
tu
r
e
e
xt
r
a
c
ti
on,
w
hi
le
th
e
S
V
M
m
e
th
od
is
u
s
e
d
f
or
pr
e
di
c
ti
on.
P
r
e
di
c
ti
ng
a
c
a
de
m
ic
s
uc
c
e
s
s
a
nd
e
va
lu
a
ti
ng
th
e
qua
li
ty
c
ont
r
ol
of
onl
in
e
tr
a
in
in
g
c
our
s
e
s
a
r
e
th
e
m
a
in
goa
ls
.
S
a
m
pl
e
vi
e
w
s
r
e
ga
r
di
ng
qua
li
ty
a
s
s
ur
a
nc
e
a
r
e
a
na
ly
z
e
d
us
in
g
de
s
c
r
ip
ti
ve
-
a
na
ly
ti
c
a
l
te
c
hni
que
s
.
B
y
br
i
dgi
ng
th
e
ga
p
be
twe
e
n
s
tu
de
nt
pe
r
f
or
m
a
nc
e
pr
e
di
c
ti
on
a
nd
onl
in
e
le
a
r
ni
ng
r
e
qui
r
e
m
e
nt
s
,
th
i
s
w
or
k
im
pr
ove
s
th
e
c
a
li
be
r
of
onl
in
e
le
a
r
ni
ng.
G
r
ubov
e
t
al
.
[
22]
s
ugge
s
te
d
a
m
ul
ti
-
out
put
hybr
id
e
ns
e
m
bl
e
m
ode
l
th
a
t
m
a
ke
s
us
e
of
in
f
or
m
a
ti
on
f
r
om
th
e
s
upe
r
s
ta
r
le
a
r
ni
ng
c
om
m
uni
c
a
ti
on
pl
a
tf
or
m
(
S
L
C
P
)
to
f
or
e
c
a
s
t
gr
a
de
s
.
I
t
pr
e
di
c
ts
m
id
te
r
m
a
nd
f
in
a
l
gr
a
d
e
s
us
in
g
th
e
X
G
B
oos
t
(
X
G
B
)
m
ode
l,
out
pe
r
f
or
m
in
g
c
om
pa
r
a
bl
e
m
ode
ls
w
it
h
a
n
a
c
c
ur
a
c
y
of
78.37%
.
F
ur
th
e
r
m
or
e
,
th
e
g
r
a
di
e
nt
boos
ti
ng
m
ode
l
out
pe
r
f
or
m
s
c
om
pa
r
a
bl
e
m
ode
ls
in
m
e
a
n
s
qua
r
e
d
e
r
r
or
w
he
n
u
s
e
d
to
f
or
e
c
a
s
t
gr
a
de
s
f
or
hom
e
w
or
k
a
nd
e
xpe
r
im
e
nt
s
.
T
hi
s
m
ul
ti
-
out
put
hybr
id
e
ns
e
m
bl
e
m
ode
l
s
he
ds
li
ght
on
how
gr
a
de
pr
e
di
c
ti
ons
c
a
n e
nh
a
nc
e
t
he
c
a
li
be
r
of
s
tu
de
nt
l
e
a
r
ni
ng a
nd t
he
e
f
f
ic
a
c
y of
t
e
a
c
he
r
i
ns
tr
uc
ti
on.
A
ld
a
lu
r
[
23]
pr
ovi
de
d
E
D
M
da
ta
s
e
t
c
ol
le
c
te
d
f
r
om
di
f
f
e
r
e
n
t
da
ta
ba
s
e
s
a
nd
e
-
le
a
r
ni
ng
s
ys
te
m
s
.
H
e
r
e
e
ns
e
m
bl
e
l
e
a
r
ni
ng c
om
bi
ni
ng mul
ti
pl
e
m
a
c
hi
ne
l
e
a
r
ni
ng
m
ode
l
m
e
c
ha
ni
s
m
i
s
c
ons
tr
uc
te
d f
or
pr
e
di
c
ti
ng
s
tu
de
nt
pe
r
f
or
m
a
nc
e
dur
in
g
th
e
c
our
s
e
.
T
he
out
c
om
e
s
how
s
e
ns
e
m
bl
e
m
ode
l
out
pe
r
f
or
m
s
ot
he
r
m
ode
l
in
te
r
m
s
of
pr
e
di
c
ti
on
a
c
c
ur
a
c
y.
S
im
il
a
r
ly
,
pr
e
di
c
ti
ng
s
tu
de
nt
pe
r
f
or
m
a
nc
e
in
onl
in
e
in
te
r
a
c
ti
ve
s
e
s
s
io
ns
us
in
g
a
da
ta
s
e
t
ga
th
e
r
e
d
f
r
om
di
gi
ta
l
e
le
c
tr
oni
c
s
e
duc
a
ti
on
a
nd
d
e
s
i
gn
s
ui
te
s
w
a
s
th
e
m
a
in
go
a
l
of
th
e
pr
oj
e
c
t.
T
he
da
ta
s
e
t
r
e
c
or
ds
te
xt
e
di
ti
ng,
ke
ys
tr
oke
s
,
a
c
ti
vi
ty
dur
a
ti
on,
e
xa
m
r
e
s
ul
ts
pe
r
s
e
s
s
io
n,
a
nd
s
tu
de
nt
in
te
r
a
c
ti
ons
dur
in
g
onl
in
e
la
b
w
or
k.
T
he
s
tu
dy
pr
e
s
e
nt
s
a
pr
e
di
c
ti
on
m
ode
l
m
a
de
up
of
86
s
ta
ti
s
ti
c
a
l
pa
r
a
m
e
te
r
s
th
a
t
c
a
n
be
br
oa
dl
y
gr
oupe
d
in
to
th
r
e
e
c
a
te
gor
ie
s
:
pe
r
ip
he
r
a
l
a
c
ti
vi
ty
c
ount
,
ti
m
in
g
s
ta
ti
s
ti
c
s
,
a
nd
a
c
ti
vi
ty
ty
pe
.
F
iv
e
w
e
ll
-
known
c
la
s
s
if
ie
r
s
a
r
e
us
e
d,
in
c
lu
di
ng
R
F
a
nd
S
V
M
,
f
or
f
e
a
tu
r
e
s
e
le
c
ti
on,
w
hi
c
h
he
lp
s
pr
e
s
e
r
ve
im
por
ta
nt
f
e
a
tu
r
e
s
.
T
he
m
ode
l'
s
goa
l
is
to
f
or
e
c
a
s
t
w
he
th
e
r
a
s
tu
de
nt
w
il
l
do
w
e
ll
o
r
poor
ly
.
T
he
m
ode
l
i
s
e
va
lu
a
te
d
in
th
r
e
e
di
f
f
e
r
e
nt
c
ir
c
um
s
ta
n
c
e
s
,
a
n
d
th
e
r
e
s
ul
ts
s
how
r
e
m
a
r
ka
bl
e
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y,
w
it
h
R
F
e
xhi
bi
ti
ng
th
e
be
s
t
pe
r
f
or
m
a
nc
e
a
t
97.4%
.
H
ow
e
ve
r
,
w
he
n
da
ta
is
im
ba
la
nc
e
d
in
na
tu
r
e
th
e
s
e
m
ode
l
f
a
il
s
to
e
s
ta
bl
i
s
h
f
e
a
tu
r
e
im
pa
c
ti
ng
in
id
e
nt
if
yi
ng
th
e
e
nga
g
e
m
e
nt
le
v
e
l
a
nd
pe
r
f
or
m
a
nc
e
;
th
us
, pr
ovi
de
s
poor
c
la
s
s
if
ic
a
ti
on a
c
c
ur
a
c
ie
s
[
24]
.
T
he
f
oc
us
of
th
e
c
ur
r
e
nt
w
or
k
is
de
ve
lo
pe
d
a
nove
l
e
ns
e
m
bl
e
le
a
r
ni
ng
m
ode
l
th
a
t
is
e
f
f
ic
ie
nt
in
s
ol
vi
ng both bi
na
r
y a
nd
m
ul
ti
-
la
be
l
c
la
s
s
if
ic
a
ti
on pr
obl
e
m
in
a
tt
a
in
in
g hi
ghe
r
pr
e
di
c
ti
on
a
c
c
ur
a
c
y c
ons
id
e
r
in
g
bot
h
s
tu
de
nt
e
nga
ge
m
e
nt
a
nd
pe
r
f
or
m
a
nc
e
da
ta
s
e
t
in
onl
in
e
s
t
ude
nt
e
L
e
a
r
ni
ng
w
e
b
por
ta
l.
M
ul
ti
pl
e
m
a
c
hi
ne
le
a
r
ni
ng
m
ode
ls
a
r
e
c
om
bi
ne
d
by
e
xi
s
ti
ng
m
ode
ls
to
c
r
e
a
te
e
ns
e
m
bl
e
le
a
r
ni
ng.
N
one
th
e
le
s
s
,
th
e
s
e
m
ode
ls
w
or
k
w
e
ll
f
or
bi
na
r
y
c
la
s
s
if
ic
a
ti
on
pr
obl
e
m
s
.
H
ow
e
ve
r
,
th
e
y
pe
r
f
or
m
poor
ly
w
he
n
a
ppl
ie
d
to
m
ul
ti
-
la
be
l
c
la
s
s
if
ic
a
ti
on
pr
obl
e
m
s
th
a
t
t
a
ke
da
t
a
im
ba
la
nc
e
in
to
a
c
c
ount
[
2
5]
.
T
he
dr
a
w
b
a
c
ks
dr
iv
e
th
is
s
tu
dy'
s
e
f
f
or
ts
to
e
nha
nc
e
e
n
s
e
m
bl
e
a
ppr
oa
c
h
to
c
r
e
a
te
a
be
tt
e
r
s
tu
de
nt
pe
r
f
or
m
a
nc
e
a
nd
e
ng
a
ge
m
e
nt
pr
e
di
c
ti
on
m
ode
l.
T
hi
s
pa
pe
r
f
ir
s
t
pr
e
s
e
nt
s
an
e
L
e
a
r
ni
ng w
e
b por
ta
l
f
r
a
m
e
w
or
k e
m
pl
oyi
ng a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
t
e
c
hni
que
l
e
ve
r
a
gi
ng
nove
l
e
ns
e
m
bl
e
le
a
r
ni
ng
n
a
m
e
ly
m
ul
ti
-
la
be
l
e
n
s
e
m
bl
e
le
a
r
ni
ng
(
M
L
E
L
)
m
ode
l
to
a
s
s
e
s
s
s
tu
de
nt
e
ng
a
ge
m
e
nt
a
nd
pe
r
f
or
m
a
nc
e
.
T
he
e
ns
e
m
bl
e
m
ode
l
is
c
r
e
a
te
d
u
s
in
g
r
e
f
in
e
d
X
G
B
a
lg
or
it
hm
.
L
a
te
r
,
f
e
a
tu
r
e
e
ns
e
m
bl
e
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
6
,
D
e
c
e
m
be
r
20
25
:
5240
-
5250
5242
c
r
e
a
te
d
to
id
e
nt
if
y
th
e
us
e
f
ul
f
e
a
tu
r
e
e
m
pl
oyi
ng
K
-
f
ol
d
c
r
os
s
va
li
da
ti
on
a
nd
f
in
a
ll
y,
th
e
m
ul
ti
-
la
be
l
c
la
s
s
if
ie
r
is
c
ons
tr
uc
te
d. R
e
s
e
a
r
c
h
s
ig
ni
f
ic
a
nc
e
:
‒
T
he
w
or
k i
nt
r
oduc
e
d
M
L
E
L
le
ve
r
a
gi
ng r
e
f
in
e
d X
G
B
m
ode
l
a
n
d K
-
f
ol
d c
r
os
s
va
li
da
ti
on.
‒
T
he
w
or
k
a
na
ly
z
e
d
bot
h
pe
r
f
or
m
a
nc
e
a
nd
e
nga
ge
m
e
nt
da
ta
s
e
t.
N
o
pr
io
r
w
o
r
k
ha
s
a
na
ly
z
e
d
bot
h
da
ta
s
e
ts
t
og
e
th
e
r
. T
hi
s
s
how
s
r
obus
tn
e
s
s
of
pr
opos
e
d m
ode
l.
‒
T
he
r
e
s
ul
t
s
how
s
t
he
pr
opos
e
d m
od
e
l
a
c
hi
e
ve
s
m
uc
h be
tt
e
r
pe
r
f
or
m
a
nc
e
t
ha
n e
xi
s
ti
ng me
th
ods
.
T
hi
s
pa
p
e
r
or
ga
ni
z
e
a
s
f
ol
lo
w
s
:
s
e
c
ti
on
two,
th
e
pr
opos
e
d
w
e
b
-
ba
s
e
d
le
a
r
ni
ng
to
ol
to
a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
in
g
onl
in
e
s
e
s
s
io
n
a
c
ti
vi
ty
e
nga
g
e
m
e
nt
.
I
n
s
e
c
ti
on
th
r
e
e
,
th
e
out
c
om
e
of
pr
opos
e
d
M
L
E
L
m
ode
l
is
s
tu
di
e
d
on
di
f
f
e
r
e
nt
da
ta
s
e
t
a
nd
m
e
th
odol
ogi
e
s
.
T
he
la
s
t
s
e
c
ti
on
th
e
c
ont
r
ib
ut
io
n
of
w
or
k
is
pr
ovi
de
d, f
ol
lo
w
e
d by f
ut
ur
e
e
nha
nc
e
m
e
nt
.
2.
P
R
O
P
O
S
E
D
M
E
T
H
O
D
I
n
th
is
s
e
c
ti
on,
a
w
e
b
-
ba
s
e
d
le
a
r
ni
ng
to
ol
de
s
ig
ne
d
i
s
in
tr
oduc
e
d
to
e
va
lu
a
te
s
tu
de
nt
pe
r
f
or
m
a
nc
e
th
r
ough
th
e
ir
e
nga
ge
m
e
nt
in
onl
in
e
s
e
s
s
io
ns
,
a
s
il
lu
s
tr
a
te
d
in
F
ig
u
r
e
1.
T
o
e
nha
nc
e
th
e
a
c
c
ur
a
c
y
of
pe
r
f
or
m
a
nc
e
a
s
s
e
s
s
m
e
nt
a
nd
ga
uge
e
ng
a
ge
m
e
nt
le
ve
l
s
,
w
e
pr
opos
e
a
nove
l
e
ns
e
m
bl
e
le
a
r
ni
ng
m
ode
l
th
a
t
r
e
f
in
e
s
X
G
B
th
r
ough
K
-
f
ol
d
c
r
os
s
-
va
li
da
ti
on
(
C
V
)
.
B
y
in
te
gr
a
ti
ng
th
e
s
e
te
c
hni
que
s
,
w
e
a
im
to
pr
ovi
de
a
r
obus
t
a
nd
c
om
pr
e
he
ns
iv
e
a
ppr
oa
c
h
f
or
a
na
ly
z
in
g
s
t
ude
nt
e
nga
ge
m
e
nt
a
nd
pe
r
f
or
m
a
nc
e
in
onl
in
e
le
a
r
ni
ng e
nvi
r
onm
e
nt
s
.
F
ig
ur
e
1
.
e
L
e
a
r
ni
ng w
e
b por
ta
l
to
a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
a
nd e
nga
ge
m
e
nt
us
in
g
M
L
E
L
X
G
B
m
ode
l
2.1. Re
f
in
e
d
X
G
B
oos
t
m
od
e
l
T
he
X
G
B
tr
e
e
a
lg
or
it
hm
r
e
pr
e
s
e
nt
s
a
n
e
nha
nc
e
d
ve
r
s
io
n
of
th
e
pr
e
vi
ous
gr
a
di
e
nt
-
boos
ti
ng
a
ppr
oa
c
h
[
25]
.
I
t
in
vol
ve
s
th
e
a
ggr
e
ga
ti
on
of
le
s
s
e
f
f
e
c
ti
ve
c
la
s
s
if
ie
r
s
to
f
or
m
a
r
obus
t
c
la
s
s
if
ie
r
,
r
e
s
ul
ti
ng
in
im
pr
ove
d
c
la
s
s
if
ic
a
ti
on
r
e
s
ul
t
s
.
L
e
t
us
c
ons
id
e
r
a
d
a
ta
s
e
t
d
e
not
e
d
a
s
E
,
th
a
t
r
e
pr
e
s
e
nt
s
a
n
ongoing
s
tr
e
a
m
of
le
a
r
ni
ng
s
e
s
s
io
n
in
f
or
m
a
ti
on.
T
hi
s
da
ta
s
e
t
c
ons
is
ts
of
o
e
xa
m
pl
e
s
,
w
he
r
e
e
a
c
h
s
a
m
pl
e
is
r
e
pr
e
s
e
nt
e
d
by
a
pa
ir
(
y
j
,
z
j
)
.
H
e
r
e
,
y
j
r
e
pr
e
s
e
nt
s
a
ve
c
to
r
of
n
f
e
a
tu
r
e
s
,
a
nd
z
j
r
e
pr
e
s
e
nt
s
a
la
be
l
a
s
s
oc
ia
te
d
w
it
h
th
e
e
xa
m
pl
e
.
T
h
e
va
r
ia
bl
e
z
̂
j
is
ut
il
iz
e
d t
o de
not
e
t
he
e
xpe
c
t
e
d r
e
s
ul
t
ge
ne
r
a
te
d by the
a
ppr
oa
c
h i
n
(
1)
.
z
̂
j
=
∑
L
l
−
1
g
l
(
y
j
)
,
g
l
∈
G
(
1)
T
he
te
r
m
g
l
r
e
f
e
r
s
to
a
n
in
d
e
pe
nde
nt
r
e
gr
e
s
s
io
n
-
tr
e
e
,
w
hi
le
g
l
(
y
j
)
de
not
e
s
th
e
c
or
r
e
s
ponding
pr
e
di
c
ti
on
r
e
s
ul
ts
ge
n
e
r
a
te
d
by
th
e
l
th
tr
e
e
f
or
th
e
j
th
s
a
m
pl
e
a
s
s
ho
w
n
in
(
2)
.
I
n
th
e
c
ont
e
xt
of
th
is
s
tu
dy,
it
is
obs
e
r
ve
d
th
a
t
f
or
e
a
c
h
tr
e
e
,
de
not
e
d
a
s
g
(
y
)
,
th
e
r
e
e
xi
s
ts
s
om
e
a
gr
e
e
m
e
nt
r
e
ga
r
di
ng
th
e
le
a
f
-
w
e
ig
ht
,
r
e
pr
e
s
e
nt
e
d
by
x
,
a
nd
th
e
s
tr
uc
tu
r
e
va
r
ia
bl
e
,
de
not
e
d
a
s
t
.
T
he
r
e
gr
e
s
s
io
n
-
tr
e
e
,
de
not
e
d
a
s
g
l
,
a
lo
ngs
id
e
it
s
c
or
r
e
s
ponding f
unc
ti
on c
a
n both be
a
c
qui
r
e
d by mi
ni
m
iz
in
g t
he
obj
e
c
ti
ve
-
f
unc
ti
on
pr
e
s
e
nt
e
d i
n (
3)
.
G
=
{
g
(
y
)
=
x
t
(
y
)
}
(
2)
O
=
∑
o
j
=
1
m
(
z
j
,
z
̂
j
)
+
∑
L
l
=
1
β
(
g
l
)
(
3)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
w
e
b
-
bas
e
d l
e
ar
ni
ng plat
fo
r
m
t
o as
s
e
s
s
s
tu
d
e
nt
pe
r
fo
r
m
anc
e
u
s
in
g onli
ne
s
e
s
s
io
n
…
(
Shas
hi
r
e
k
ha
)
5243
I
n
th
is
s
tu
dy,
th
e
va
r
ia
bl
e
m
is
de
f
in
e
d
to
be
th
e
in
it
ia
l
tr
a
in
in
g
-
lo
s
s
f
unc
ti
on,
w
hi
c
h
is
ut
il
iz
e
d
to
qua
nt
if
y t
he
di
f
f
e
r
e
nc
e
be
twe
e
n t
he
pr
e
di
c
te
d outc
om
e
, de
not
e
d
a
s
z
̂
j
, a
lo
ng w
it
h
t
he
a
c
tu
a
l
out
c
om
e
, de
not
e
d
a
s
z
j
.
T
o
m
it
ig
a
te
th
e
is
s
ue
of
ove
r
-
f
it
ti
ng,
r
e
s
e
a
r
c
he
r
s
of
te
n
e
m
pl
oy
a
va
r
ia
bl
e
de
not
e
d
a
s
β
to
pe
na
li
z
e
th
e
c
om
pl
e
x
na
tu
r
e
of
a
pr
e
di
c
ti
ve
a
ppr
oa
c
h.
T
hi
s
a
ppr
oa
c
h
s
e
e
ks
to
a
c
hi
e
ve
a
ba
la
n
c
e
b
e
twe
e
n
a
ppr
oa
c
h
c
om
pl
e
xi
ty
a
nd
ge
ne
r
a
li
z
a
ti
on
pe
r
f
or
m
a
nc
e
.
B
y
in
t
r
oduc
in
g
a
pe
na
lt
y
te
r
m
,
th
e
a
ppr
oa
c
h'
s
a
bi
li
ty
to
f
it
noi
s
e
or
ir
r
e
le
va
nt
f
e
a
tu
r
e
s
in
th
e
d
a
ta
i
s
r
e
duc
e
d,
th
us
im
pr
ovi
ng
it
s
a
bi
li
ty
to
m
a
ke
a
c
c
ur
a
te
pr
e
di
c
ti
ons
on
uns
e
e
n
da
ta
. T
he
e
v
a
lu
a
ti
on of
β
is
gi
ve
n a
s
s
how
n i
n (
4)
.
β
(
g
l
)
=
δ
U
+
1
2
μ
‖
x
‖
2
(
4)
T
he
r
e
gul
a
r
iz
a
ti
on
-
va
r
ia
bl
e
i
s
de
not
e
d
by
δ
a
nd
μ
,
w
hi
le
th
e
le
a
f
'
s
-
s
iz
e
is
r
e
pr
e
s
e
nt
e
d
by
U
.
A
ddi
ti
ona
ll
y,
th
e
r
a
nki
ng
f
or
va
r
io
us
le
a
ve
s
is
de
not
e
d
by
x
.
T
he
c
ons
tr
uc
ti
on
of
a
ny
e
n
s
e
m
bl
e
-
tr
e
e
is
a
c
hi
e
ve
d by me
a
ns
of
a
s
um
m
a
ti
on me
th
od. T
he
a
nt
ic
ip
a
te
d r
e
s
ul
ts
f
or
t
he
j
th
s
a
m
pl
e
dur
in
g t
he
u
th
it
e
r
a
ti
on,
de
not
e
d
a
s
z
̂
j
(
u
)
,
ne
c
e
s
s
it
a
t
e
s
th
e
in
c
lu
s
io
n
of
g
u
in
or
de
r
to
m
in
im
iz
e
th
e
s
pe
c
if
ie
d
f
unc
ti
on
a
s
s
how
n
in
(
5)
.
T
he
a
f
or
e
m
e
nt
io
ne
d
e
qua
ti
on
c
a
n
be
r
e
duc
e
d
by
e
m
pl
oyi
ng
th
e
te
c
hni
que
of
r
e
m
ovi
ng
th
e
s
ta
bl
e
va
r
ia
bl
e
us
in
g t
he
s
e
c
ond
-
or
de
r
T
a
yl
or
'
s
e
xpa
ndi
ng, whic
h c
a
n b
e
e
xpr
e
s
s
e
d i
n (
6)
.
O
(
u
)
=
∑
o
j
=
1
m
(
z
j
,
z
̂
j
(
u
−
1
)
+
g
u
(
y
j
)
)
+
β
(
g
l
)
(
5)
O
(
u
)
=
∑
o
j
=
1
[
h
j
g
j
(
y
j
)
+
1
2
i
j
g
u
(
y
j
)
2
]
+
β
(
g
l
)
(
6)
T
he
va
r
ia
bl
e
h
j
is
us
e
d
to
de
not
e
th
e
in
it
ia
l
o
r
de
r
-
gr
a
di
e
nt
w
it
h
r
e
ga
r
d
to
m
,
a
nd
it
s
de
f
in
e
d
is
gi
ve
n
a
s
s
ho
w
n
in
(
7)
.
T
h
e
va
r
ia
bl
e
i
j
is
u
s
e
d
to
d
e
not
e
th
e
ne
xt
or
de
r
-
gr
a
di
e
nt
w
it
h
r
e
ga
r
d
to
m
,
a
nd
it
s
de
f
in
e
d
is
gi
ve
n
a
s
s
how
n
in
(
8)
.
H
e
nc
e
,
th
e
obj
e
c
ti
ve
-
f
unc
ti
on
pa
r
a
m
e
te
r
s
of
th
e
pr
e
di
c
ti
ng
a
ppr
oa
c
h
a
r
e
m
a
th
e
m
a
ti
c
a
ll
y r
e
pr
e
s
e
nt
e
d by the
s
ubs
e
qu
e
nt
e
qua
ti
on a
s
i
n (
9
)
.
h
j
=
∂
z
̂
j
(
u
−
1
)
m
(
z
j
,
z
̂
j
(
u
−
1
)
)
(
7)
i
j
=
∂
z
̂
j
(
u
−
1
)
2
m
(
z
j
,
z
̂
j
(
u
−
1
)
)
(
8)
O
(
u
)
=
∑
o
j
=
1
[
h
j
g
j
(
y
j
)
+
1
2
i
j
g
u
(
y
j
)
2
]
+
δ
U
+
1
2
μ
∑
U
k
=
1
x
k
2
(
9)
T
he
f
or
m
ul
a
m
e
nt
io
ne
d
a
bov
e
is
r
e
pr
e
s
e
nt
e
d
in
it
s
s
im
pl
e
s
t
f
or
m
a
s
s
how
n
in
(
10)
.
T
he
s
a
m
pl
e
c
ol
le
c
ti
on
of
le
a
f
k
,
de
not
e
d
a
s
J
k
,
is
r
e
pr
e
s
e
nt
e
d
in
th
e
f
ol
lo
w
in
g
m
a
nne
r
a
s
s
how
n
in
(
11)
.
T
he
tr
e
e
-
s
iz
e
,
de
not
e
d
by
r
,
is
a
s
s
um
e
d
to
be
f
ix
e
d.
I
n
or
de
r
to
de
te
r
m
in
e
th
e
id
e
a
l
w
e
ig
ht
s
,
x
k
∗
,
f
or
le
a
f
j
,
th
e
f
o
ll
ow
in
g
e
qua
ti
on
is
e
m
pl
oye
d
a
s
s
how
n
in
(
12)
.
T
he
id
e
a
l
w
e
ig
ht
va
lu
e
s
f
or
e
a
c
h
tr
e
e
-
s
iz
e
a
r
e
th
e
n
de
r
iv
e
d
in
(
13)
.
T
he
va
r
ia
bl
e
H
k
is
de
not
e
d i
n t
he
f
ol
lo
w
in
g m
a
nne
r
a
s
s
how
n i
n (
1
4)
. T
he
va
r
ia
bl
e
I
k
is
de
not
e
d i
n (
15)
.
O
(
u
)
=
∑
U
j
=
1
[
(
∑
j
∈
J
k
h
j
)
x
j
1
2
(
∑
j
∈
J
k
i
j
+
μ
)
x
k
2
]
+
δ
U
(
10)
J
k
=
{
r
(
y
j
=
k
)
}
(
11)
x
k
∗
=
H
k
I
k
+
μ
(
12)
O
∗
=
1
2
∑
U
k
−
1
H
k
2
I
k
+
μ
+
δ
U
(
13)
H
k
=
∑
j
∈
J
k
h
j
(
14)
I
k
=
∑
j
∈
J
k
i
j
(
15)
T
he
O
∗
m
e
tr
ic
is
ut
il
iz
e
d
to
e
va
lu
a
te
th
e
c
ha
r
a
c
te
r
is
ti
c
s
of
tr
e
e
r
,
w
it
h
a
lo
w
e
r
va
lu
e
in
di
c
a
ti
ng
a
m
or
e
f
a
vor
a
bl
e
tr
e
e
or
ga
ni
z
a
ti
on.
W
hi
le
X
G
B
a
ppr
oa
c
h
a
r
e
kn
ow
n
f
or
th
e
ir
e
f
f
ic
ie
nc
y
in
a
c
hi
e
vi
ng
e
xc
e
ll
e
nt
f
or
e
c
a
s
ti
ng
a
c
c
ur
a
c
y,
it
is
im
por
ta
nt
to
not
e
th
a
t
th
e
y
m
a
y
e
n
c
ount
e
r
c
ha
ll
e
nge
s
in
s
c
e
na
r
io
s
w
h
e
r
e
f
e
a
tu
r
e
s
e
le
c
ti
on
is
in
a
de
qua
te
or
w
he
ne
ve
r
de
a
li
ng
w
it
h
im
ba
la
nc
e
d
in
f
or
m
a
ti
on
c
ons
id
e
r
in
g
m
ul
ti
-
la
be
l
c
la
s
s
if
ic
a
ti
on
pr
obl
e
m
.
I
n
s
uc
h
c
a
s
e
s
,
th
e
pr
e
di
c
ti
on
a
c
c
ur
a
c
y
o
f
X
G
B
a
ppr
oa
c
he
s
m
a
y
e
xpe
r
ie
nc
e
a
de
c
r
e
a
s
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
6
,
D
e
c
e
m
be
r
20
25
:
5240
-
5250
5244
in
pe
r
f
or
m
a
nc
e
.
T
h
e
s
ub
s
e
que
nt
s
ub
s
e
c
ti
on
w
il
l
f
oc
us
on
th
e
m
ode
li
ng
of
a
n
e
f
f
ic
ie
nt
f
e
a
tu
r
e
s
e
l
e
c
ti
on
te
c
hni
que
t
hr
oughout t
he
t
r
a
in
in
g i
nf
or
m
a
ti
on, a
s
a
m
e
a
ns
of
s
o
lv
in
g t
he
m
ul
ti
-
la
be
l
c
la
s
s
if
ic
a
ti
on pr
obl
e
m
.
2.2. M
u
lt
i
-
la
b
e
l
e
n
s
e
m
b
le
l
e
ar
n
in
g m
od
e
l
I
n
th
is
s
tu
dy,
w
e
pr
opos
e
a
m
odi
f
ic
a
ti
on
f
or
th
e
f
e
a
tu
r
e
-
s
e
le
c
ti
on
m
e
th
od
of
th
e
c
onve
nt
io
na
l
X
G
B
a
ppr
oa
c
h
in
or
de
r
to
pe
r
f
or
m
c
la
s
s
if
ic
a
ti
on
on
m
ul
ti
-
la
be
l
da
ta
s
e
t.
O
ur
a
im
is
to
e
nha
n
c
e
th
e
f
e
a
tu
r
e
-
im
por
ta
nc
e
r
e
s
ul
t,
th
e
r
e
by
le
a
di
ng
to
w
a
r
ds
a
n
e
nha
nc
e
d
pr
e
d
ic
ti
on
a
ppr
oa
c
h.
I
n
th
is
w
or
k
a
nove
l
K
-
f
ol
d
c
r
os
s
va
li
da
ti
on
a
ppr
oa
c
h
is
in
tr
oduc
e
d
by
m
odi
f
yi
ng
th
e
be
lo
w
s
ta
nda
r
d
K
-
f
ol
d
c
r
os
s
va
li
da
ti
on
c
ons
id
e
r
in
g
gr
id
s
iz
e
of
l
a
s
s
how
n i
n (
16)
.
CV
(
σ
)
=
1
M
∑
K
k
=
1
∑
j
∈
G
−
k
P
(
b
j
,
g
̂
σ
−
k
(
j
)
(
y
j
,
σ
)
)
(
16)
N
e
ve
r
th
e
le
s
s
,
it
is
im
por
ta
nt
to
not
e
th
a
t
th
e
a
f
or
e
m
e
nt
io
ne
d
e
q
ua
ti
on
doe
s
not
e
xpl
ic
it
ly
s
pe
c
if
y
th
e
pa
r
ti
c
ul
a
r
f
e
a
tu
r
e
(
s
)
th
a
t
e
xe
r
t
in
f
lu
e
nc
e
on
th
e
c
or
r
e
c
tn
e
s
s
of
th
e
pr
e
di
c
ti
ve
a
ppr
oa
c
h.
T
hi
s
s
tu
dy
a
im
s
to
in
ve
s
ti
ga
te
th
e
im
pl
e
m
e
nt
a
ti
on
of
a
r
obu
s
t
C
V
te
c
hni
que
c
oup
le
d
w
it
h
a
n
e
f
f
ic
ie
nt
f
e
a
tu
r
e
-
s
e
le
c
ti
on
m
e
th
od
to
e
nha
nc
e
pr
e
di
c
ti
ve
a
c
c
ur
a
c
y.
T
h
e
pr
opos
e
d
a
ppr
oa
c
h
i
s
de
s
ig
ne
d
to
in
c
or
por
a
te
f
e
a
tu
r
e
s
th
a
t
ha
ve
a
s
ig
ni
f
ic
a
nt
e
f
f
e
c
t
on t
he
pr
e
di
c
ti
ve
pe
r
f
or
m
a
nc
e
. T
hi
s
i
s
e
va
lu
a
t
e
d us
in
g (
17)
.
CV
(
σ
)
=
1
SM
∑
S
s
=
1
∑
K
k
=
1
∑
j
∈
G
−
k
P
(
b
j
,
g
̂
σ
−
k
(
j
)
(
y
j
,
σ
)
)
(
17)
I
n
(
17)
,
th
e
p
r
oc
e
s
s
of
s
e
le
c
ti
ng
th
e
opt
im
a
l
va
lu
e
f
or
σ
̂
in
or
de
r
to
opt
im
iz
e
th
e
le
a
r
ne
r
'
s
pr
e
di
c
ti
ve
a
ppr
oa
c
h
is
a
c
hi
e
ve
d
th
r
ough
th
e
f
ol
lo
w
in
g
a
s
gi
ve
n
(
18)
.
I
n
(
18)
,
M
r
e
pr
e
s
e
nt
s
th
e
s
iz
e
of
th
e
tr
a
in
in
g
-
s
e
t
unde
r
c
ons
id
e
r
a
ti
on.
T
h
e
f
unc
ti
on
P
(
∙
)
is
us
e
d
f
or
de
f
in
in
g
th
e
lo
s
s
-
f
unc
ti
on,
w
hi
le
th
e
f
unc
ti
on
g
̂
σ
−
k
(
j
)
(
∙
)
is
e
m
pl
oye
d
f
or
c
a
lc
ul
a
ti
ng
th
e
c
oe
f
f
ic
ie
nt
s
.
T
h
e
s
e
le
c
ti
on
of
e
f
f
e
c
ti
ve
f
e
a
tu
r
e
s
in
de
v
e
lo
pi
ng
a
le
a
r
ne
r
'
s
pe
r
f
or
m
a
nc
e
pr
e
di
c
ti
ve
a
ppr
oa
c
h
is
a
c
c
om
pl
i
s
he
d us
in
g t
he
t
e
c
hni
que
of
r
a
nki
ng
r
(
∙
)
, a
s
i
ndi
c
a
te
d
in
(
19)
.
σ
̂
=
CV
s
(
σ
)
(
18)
r
(
a
)
=
{
0
if
n
j
is
not
s
e
l
e
c
t
e
d
1
if
n
j
is
s
e
l
e
c
t
e
d
as
o
p
t
i
m
a
l
p
r
e
di
c
t
i
o
n
m
o
de
l
j
=
1
,
2
,
3
,
…
,
n
(
19)
T
he
f
ol
lo
w
in
g
e
qua
ti
on
is
us
e
d
f
or
c
on
s
tr
uc
ti
on
of
th
e
f
e
a
tu
r
e
-
s
ubs
e
t
a
s
s
ho
w
n
in
(
20)
.
T
he
opt
im
a
l
f
e
a
tu
r
e
s
,
w
hi
c
h
a
c
hi
e
ve
s
th
e
hi
ghe
s
t
r
a
nk,
is
de
te
r
m
in
e
d
by
c
ons
id
e
r
in
g
m
ul
ti
pl
e
in
s
ta
nc
e
s
of
K
−
fo
l
d
s
.
T
hi
s
is
e
va
lu
a
te
d
us
in
g
(
21)
.
N
e
xt
,
th
e
c
a
lc
ul
a
ti
on
is
pe
r
f
o
r
m
e
d
to
de
te
r
m
in
e
th
e
f
r
e
que
nc
y
of
a
s
pe
c
if
ic
f
e
a
tu
r
e
be
in
g
c
hos
e
n
w
it
hi
n
K
f
e
a
tu
r
e
-
s
ubs
e
ts
th
a
t
ha
ve
th
e
hi
gh
e
s
t
r
a
nk.
T
he
f
in
a
li
z
e
d
f
e
a
tu
r
e
-
s
ubs
e
t
c
a
n
be
de
r
iv
e
d
us
in
g
th
e
(
22)
.
T
h
e
f
unc
ti
on
f
s
(
∙
)
r
e
pr
e
s
e
nt
s
a
s
c
e
na
r
io
in
w
hi
c
h
th
e
n
th
f
e
a
tu
r
e
i
s
e
it
he
r
c
hos
e
n
or
not
. T
hi
s
i
s
s
c
ie
nt
if
ic
a
ll
y de
not
e
d a
s
(
23)
.
F
s
=
{
r
(
n
1
)
,
r
(
n
1
)
,
…
,
r
(
n
n
)
}
(
20)
F
s
k
=
{
r
(
n
1
)
,
r
(
n
1
)
,
…
,
r
(
n
n
)
}
(
21)
F
s
f
i
n
a
l
=
{
f
s
(
p
1
)
,
f
s
(
n
2
)
,
…
,
f
s
(
n
n
)
}
(
22)
F
s
(
a
)
=
{
0
if
q
j
is
c
ho
s
e
n
l
e
s
s
e
r
t
ha
n
K
2
t
im
e
s
j
=
1
,
2
,
3
,
…
,
n
1
if
q
j
is
c
ho
s
e
n
g
r
e
a
t
e
r
or
e
q
ua
l
to
K
2
t
im
e
s
,
j
=
1
,
2
,
3
,
…
,
n
(
23)
T
he
pr
e
c
e
di
ng
e
qua
ti
on
is
ut
il
iz
e
d
to
ge
ne
r
a
te
a
gr
oup
of
n
′
c
hos
e
n
f
e
a
tu
r
e
s
,
in
w
hi
c
h
n
th
r
e
pr
e
s
e
nt
s
th
e
f
r
e
que
nc
y of
f
e
a
tu
r
e
s
e
le
c
ti
on. T
he
t
r
a
in
in
g s
e
t
us
e
d i
n t
hi
s
s
tu
dy i
s
a
gr
oup tha
t
ha
s
be
e
n c
a
r
e
f
ul
ly
c
hos
e
n
to
in
c
lu
de
onl
y
r
e
le
v
a
nt
f
e
a
tu
r
e
s
.
T
hi
s
a
ppr
oa
c
h
a
im
s
to
c
on
s
tr
uc
t
a
le
a
r
ne
r
'
s
pr
e
di
c
ti
ve
a
ppr
oa
c
h
th
a
t
is
bot
h
e
f
f
ic
ie
nt
a
nd
a
c
c
ur
a
te
.
T
o
m
it
ig
a
te
th
e
im
pa
c
t
of
va
r
ia
bi
li
ty
th
r
oughout
th
e
tr
a
in
in
g
s
ta
ge
,
th
e
K
−
fo
l
d
s
te
c
hni
que
is
e
m
pl
oye
d
by
it
e
r
a
ti
ve
ly
r
e
pe
a
ti
ng
th
e
s
te
ps
S
nu
m
e
r
ous
ti
m
e
s
.
A
ddi
ti
ona
ll
y,
w
it
h
th
e
goa
l
to
m
in
im
iz
e
va
r
ia
bi
li
ty
,
a
pa
r
ti
c
ul
a
r
gr
oup
of
f
e
a
tu
r
e
s
w
il
l
be
c
hos
e
n
ba
s
e
d
on
th
e
e
qu
a
ti
ons
out
li
ne
d.
T
he
da
ta
s
e
t
E
is
r
e
duc
e
d
to
E
′
by
s
e
le
c
ti
ve
ly
r
e
ta
in
in
g
f
e
a
tu
r
e
s
ba
s
e
d
on
th
e
c
r
it
e
r
ia
de
f
in
e
d
in
(
23
)
.
T
he
r
e
s
ul
ti
ng
da
ta
s
e
t,
de
not
e
d
a
s
E
′
,
is
obt
a
in
e
d
by
in
c
lu
di
ng
onl
y
th
e
in
s
ta
nc
e
s
f
r
om
E
th
a
t
s
a
ti
s
f
y
th
e
c
ondi
ti
ons
s
pe
c
if
ie
d by
n
′
. T
he
K
-
f
ol
ds
m
e
th
od i
s
a
na
ly
z
e
d i
n t
he
s
a
m
e
w
a
y
a
s
(
17)
. T
he
E
P
M
da
ta
s
e
t,
de
not
e
d
a
s
E
′
(
−
k
)
,
is
ut
il
iz
e
d f
or
tr
a
in
in
g pu
r
pos
e
s
by e
li
m
in
a
ti
ng t
he
k
th
s
e
c
ti
on. T
he
r
e
m
a
in
in
g s
e
c
ti
on,
E
′
(
k
)
, i
s
r
e
s
e
r
ve
d a
s
t
he
te
s
ti
ng
in
f
or
m
a
ti
on
s
e
t.
T
hi
s
pr
oc
e
s
s
is
r
e
pe
a
te
d
f
or
e
a
c
h
va
lu
e
of
k
f
r
om
1
to
K
.
T
he
f
ol
lo
w
in
g
pha
s
e
s
a
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
w
e
b
-
bas
e
d l
e
ar
ni
ng plat
fo
r
m
t
o as
s
e
s
s
s
tu
d
e
nt
pe
r
fo
r
m
anc
e
u
s
in
g onli
ne
s
e
s
s
io
n
…
(
Shas
hi
r
e
k
ha
)
5245
e
xe
c
ut
e
d
in
a
c
ont
in
ua
l
w
a
y,
th
r
ough
a
pr
e
de
te
r
m
in
e
d
s
iz
e
de
not
e
d
a
s
S
.
I
n
or
de
r
to
opt
im
iz
e
va
r
ia
bl
e
s
,
w
e
it
e
r
a
te
over
a
gr
id
of
s
iz
e
L
,
de
not
e
d
by
th
e
va
r
ia
bl
e
l
,
r
a
ngi
ng
f
r
om
1
to
L
.
T
he
de
ve
lo
pm
e
nt
of
a
pr
e
di
c
ti
on
a
ppr
oa
c
h i
s
a
c
hi
e
v
e
d t
hr
ough the
ut
il
iz
a
ti
on of
t
he
(
24)
.
g
̂
σ
l
=
g
̂
(
E
′
(
−
k
)
;
σ
l
)
(
24)
I
n
or
de
r
to
c
a
lc
ul
a
te
th
e
e
r
r
or
,
us
e
th
e
lo
s
s
-
f
unc
ti
on
e
qua
ti
ons
gi
ve
n
be
lo
w
f
or
a
va
r
ie
ty
of
l
va
lu
e
s
a
nd
im
pl
e
m
e
nt
g
̂
σ
l
to
th
e
tr
a
in
in
g
da
ta
s
e
t
E
′
(
−
k
)
a
s
s
how
n
in
(
25)
.
C
a
lc
ul
a
te
th
e
K
−
fo
l
d
CV
-
e
r
r
or
w
it
h
va
r
yi
ng
va
lu
e
s
of
th
e
opt
im
iz
a
ti
on
va
r
ia
bl
e
L
a
s
s
how
n
in
(
26)
.
U
ti
li
z
in
g
a
s
e
que
nt
ia
l
C
V
m
e
th
od,
th
e
c
a
lc
ul
a
ti
on
of
C
V
e
r
r
or
is
a
s
s
how
n
in
(
27)
.
U
s
e
th
e
s
ub
s
e
que
n
t
e
qua
ti
on
to
de
te
r
m
in
e
th
e
opt
im
um
va
lu
e
f
or
th
e
opt
im
iz
a
ti
on va
r
ia
bl
e
f
or
a
gi
ve
n r
a
nge
of
l
va
lu
e
s
a
s
s
how
n
i
n (
28)
.
E
σ
l
=
P
(
b
j
,
g
̂
(
E
′
(
−
k
)
;
σ
l
)
)
(
25)
CV
(
g
̂
;
σ
l
)
=
1
M
∑
K
k
=
1
∑
j
∈
E
′
(
−
k
)
P
(
b
j
,
g
̂
(
E
′
(
−
k
)
;
σ
l
)
)
(
26)
CV
S
(
g
̂
;
σ
l
)
=
1
KS
∑
S
s
=
1
∑
K
k
=
1
∑
j
∈
E
′
(
−
k
)
P
(
b
j
,
g
̂
(
E
′
(
−
k
)
;
σ
l
)
)
(
27)
σ
̂
=
CV
S
(
g
̂
;
σ
l
)
(
28)
T
he
f
in
a
l
pr
e
di
c
ti
on
a
ppr
oa
c
h
i
s
obt
a
in
e
d
by
c
onf
ig
ur
in
g
th
e
opt
im
um
va
lu
e
s
of
im
pr
ovi
ng
pa
r
a
m
e
te
r
s
a
s
a
n
obj
e
c
ti
ve
-
f
unc
ti
on
a
nd
th
e
n
m
in
im
iz
in
g
it
us
i
ng
a
gr
a
di
e
nt
-
de
c
e
nt
a
ppr
oa
c
h.
T
o
r
e
duc
e
th
e
a
m
ount
of
unpr
e
di
c
ta
bi
li
ty
in
th
e
pr
e
di
c
ti
on
a
ppr
oa
c
h
w
hi
le
ta
ki
ng
in
to
a
c
c
ount
di
s
ti
nc
t
f
ol
ds
,
s
ta
ge
1
of
tr
a
in
in
g
in
vol
ve
s
bui
ld
in
g
K
-
f
ol
ds
a
nd
it
e
r
a
ti
ng
on
th
e
m
S
ti
m
e
s
.
S
ta
ge
2
in
c
or
por
a
te
s
a
s
e
le
c
t
gr
oup
of
f
e
a
tu
r
e
s
in
to
th
e
f
in
a
l
f
o
r
e
c
a
s
ti
ng
a
ppr
oa
c
h,
he
lp
in
g
to
lo
w
e
r
v
a
r
ia
ti
on
in
th
e
pr
oc
e
s
s
.
T
hus
,
w
he
n
c
om
pa
r
e
d
to
s
ta
te
-
of
-
th
e
-
a
r
t
e
ns
e
m
bl
e
a
nd
M
L
-
ba
s
e
d
s
tu
de
nt
pe
r
f
or
m
a
nc
e
a
nd
e
nga
ge
m
e
nt
pr
e
di
c
ti
ve
a
ppr
oa
c
he
s
,
th
e
pr
e
s
e
nt
e
d
M
L
E
L
-
ba
s
e
d
s
tu
de
nt
p
e
r
f
or
m
a
nc
e
a
nd
e
nga
g
e
m
e
nt
pr
e
di
c
ti
ve
a
ppr
oa
c
h
gr
e
a
tl
y
e
nha
n
c
e
s
a
c
c
ur
a
c
y a
s
pr
ove
d i
n
f
ol
lo
w
in
g
s
e
c
ti
on.
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
T
he
s
e
c
ti
on
s
tu
di
e
s
th
e
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
a
c
hi
e
ve
d
u
s
i
ng
pr
opos
e
d
M
L
E
L
us
in
g
X
G
B
ove
r
ot
he
r
e
xi
s
ti
ng
a
ppr
oa
c
he
s
li
ke
X
G
B
-
ba
s
e
d,
a
nd
m
ul
ti
-
s
pl
it
opt
i
m
iz
a
ti
on
(
M
S
O
)
,
R
F
-
e
ns
e
m
bl
e
,
a
nd
two
-
la
ye
r
e
ns
e
m
bl
e
(
T
L
E
)
.
E
xpe
r
im
e
nt
s
a
r
e
c
onduc
te
d
on
m
ul
ti
-
la
be
l
pe
r
f
or
m
a
nc
e
a
nd
e
nga
ge
m
e
nt
da
ta
s
e
t.
M
or
e
de
ta
il
s
of
da
ta
s
e
t
c
a
n
be
obt
a
in
e
d.
T
h
e
f
ol
lo
w
in
g
m
e
tr
ic
s
a
r
e
us
e
d
f
or
va
li
da
ti
ng
m
ode
ls
.
T
he
a
c
c
ur
a
c
y
is
c
om
put
e
d
a
s
s
how
n
in
(
29)
.
W
he
r
e
T
P
de
f
in
e
s
tr
ue
pos
it
iv
e
,
F
P
de
f
in
e
s
f
a
ls
e
pos
it
iv
e
,
T
N
de
f
in
e
s
tr
ue
ne
ga
ti
ve
,
a
nd
F
N
de
f
in
e
s
f
a
ls
e
ne
ga
ti
ve
.
T
he
r
e
c
a
ll
is
c
om
put
e
d
a
s
s
how
n
in
(
30)
.
T
he
pr
e
c
is
io
n
is
c
om
put
e
d
a
s
s
how
n i
n (
31)
. T
he
F
1
-
s
c
or
e
i
s
c
om
put
e
d a
s
s
how
n i
n (
32)
.
Ac
c
ur
a
c
y
=
TP
+
TN
TP
+
FP
+
TN
+
FN
(
29)
R
e
c
a
l
l
=
TP
TP
+
FN
(
30)
P
r
e
c
is
io
n
=
TP
TP
+
FP
(
31)
F1
−
s
c
o
r
e
=
2
×
P
r
e
ci
s
io
n
×
S
e
n
s
iti
v
ity
P
r
e
ci
s
io
n
×
S
e
n
s
iti
v
ity
(
32)
3.1.
S
t
u
d
e
n
t
p
e
r
f
or
m
an
c
e
an
al
ys
is
T
he
s
tu
de
nt
pe
r
f
or
m
a
nc
e
a
na
ly
s
is
is
done
us
in
g
pe
r
f
or
m
a
nc
e
-
or
ie
nt
e
d
da
ta
s
e
t
obt
a
in
e
d.
V
a
r
io
us
m
ode
ls
li
ke
X
G
B
-
ba
s
e
d,
a
nd
M
S
O
,
R
F
-
e
ns
e
m
bl
e
,
a
nd
T
L
E
a
r
e
us
e
d
to
c
om
pa
r
e
w
it
h
c
la
s
s
if
ic
a
ti
on
out
c
om
e
of
pr
opos
e
d
M
L
E
L
-
X
G
B
.
T
he
F
ig
ur
e
2
s
how
s
th
e
a
c
c
ur
a
c
y
out
c
om
e
a
c
hi
e
ve
d
us
in
g
di
f
f
e
r
e
nt
m
ode
l
s
.
T
he
r
e
s
ul
t
s
how
s
th
e
pr
opos
e
d
M
L
E
L
-
X
G
B
a
c
hi
e
ve
s
a
n
a
c
c
ur
a
c
y
of
0.9995
a
nd
th
e
ne
xt
be
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d us
in
g
R
F
-
e
ns
e
m
bl
e
w
it
h 0.974, XG
B
w
it
h 0.971, T
L
E
w
it
h 0.85
,
a
nd M
S
O
w
it
h 0.65.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
6
,
D
e
c
e
m
be
r
20
25
:
5240
-
5250
5246
F
ig
ur
e
2. A
c
c
ur
a
c
y outc
om
e
s
t
o a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
i
ng s
e
s
s
io
n a
c
ti
vi
ty
F
ig
ur
e
3
s
how
s
th
e
p
r
e
c
is
io
n
out
c
om
e
a
c
hi
e
ve
d
us
in
g
di
f
f
e
r
e
nt
m
ode
ls
.
T
he
r
e
s
ul
t
s
how
s
th
e
pr
opos
e
d
M
L
E
L
-
X
G
B
a
c
hi
e
ve
s
a
n
a
c
c
ur
a
c
y
of
0.9989
a
nd
th
e
ne
xt
b
e
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d
us
in
g
RF
-
e
ns
e
m
bl
e
w
it
h
0.974,
X
G
B
w
it
h
0.97,
M
S
O
w
it
h
0.867
,
a
nd
T
L
E
w
it
h
0.86.
F
ig
ur
e
4
s
how
s
th
e
r
e
c
a
ll
out
c
om
e
a
c
hi
e
ve
d
u
s
in
g
di
f
f
e
r
e
nt
m
ode
ls
;
th
e
r
e
s
ul
t
s
how
s
th
e
pr
opos
e
d
M
L
E
L
-
X
G
B
a
c
hi
e
ve
s
a
n
a
c
c
ur
a
c
y
of
1 a
nd t
he
ne
xt
be
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d us
in
g R
F
-
e
ns
e
m
bl
e
w
it
h 0.974, X
G
B
w
it
h 0.974,
T
L
E
w
it
h
0.94
,
a
nd
M
S
O
w
it
h 0.857.
F
ig
ur
e
3. P
r
e
c
is
io
n outc
om
e
s
t
o a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
i
ng s
e
s
s
io
n a
c
ti
vi
ty
F
ig
ur
e
4. R
e
c
a
ll
out
c
om
e
s
t
o a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
in
g
s
e
s
s
io
n a
c
ti
vi
ty
F
ig
ur
e
5
s
how
s
th
e
F
1
-
s
c
or
e
out
c
om
e
a
c
hi
e
ve
d
us
in
g
di
f
f
e
r
e
nt
m
ode
ls
.
T
he
r
e
s
ul
t
s
how
s
th
e
pr
opos
e
d
M
L
E
L
-
X
G
B
a
c
hi
e
ve
s
a
n
a
c
c
ur
a
c
y
of
0.9991
a
nd
th
e
ne
xt
b
e
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d
us
in
g
RF
-
e
ns
e
m
bl
e
w
it
h
0.974,
X
G
B
w
it
h
0.97,
T
L
E
w
it
h
0.9
,
a
nd
M
S
O
w
it
h
0.857.
T
a
bl
e
1
s
how
s
th
e
0
0.2
0.4
0.6
0.8
1
1.2
R
F
-
E
ns
e
mble
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
A
c
c
u
r
a
c
y
P
r
e
di
c
i
t
v
e
m
ode
l
A
ccu
racy
(perform
an
ce)
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
0.75
0.8
0.85
0.9
0.95
1
1.05
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
P
r
e
c
i
s
i
on
v
a
l
u
e
P
r
e
di
c
t
i
on
m
ode
l
s
Preci
si
on
(perform
an
ce)
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
0.75
0.8
0.85
0.9
0.95
1
1.05
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
R
e
c
a
ll
P
r
e
di
c
t
i
on
m
ode
l
s
Recal
l
(perform
an
ce)
R
F
-
E
n
s
e
m
b
l
e
M
S
O
T
L
E
XGB
M
L
E
L
-
XGB
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
w
e
b
-
bas
e
d l
e
ar
ni
ng plat
fo
r
m
t
o as
s
e
s
s
s
tu
d
e
nt
pe
r
fo
r
m
anc
e
u
s
in
g onli
ne
s
e
s
s
io
n
…
(
Shas
hi
r
e
k
ha
)
5247
c
om
pa
r
a
ti
ve
s
tu
dy
of
a
ll
m
e
tr
ic
s
of
di
f
f
e
r
e
nt
c
la
s
s
if
ic
a
ti
on
m
ode
l
te
s
ti
ng
on
pe
r
f
or
m
a
nc
e
r
e
la
te
d
da
ta
s
e
t.
T
he
ove
r
a
ll
r
e
s
ul
ts
a
c
hi
e
ve
d
s
how
th
e
pr
opos
e
d
m
ode
l
a
c
hi
e
v
e
s
m
uc
h
hi
ghe
r
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F1
-
s
c
or
e
i
n c
om
pa
r
is
on w
it
h X
G
B
-
ba
s
e
d, M
S
O
, R
F
-
ba
s
e
d
,
a
nd
T
L
E
.
F
ig
ur
e
5. F
1
-
s
c
or
e
out
c
om
e
s
t
o a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
in
g s
e
s
s
io
n a
c
ti
vi
ty
T
a
bl
e
1. C
om
pa
r
a
ti
ve
s
tu
dy f
or
pe
r
f
or
m
a
nc
e
da
ta
s
e
t
P
r
e
di
c
t
i
ve
m
ode
l
A
c
c
ur
a
c
y
P
r
e
c
i
s
i
on
R
e
c
a
l
l
F
1
-
s
c
or
e
RF
-
e
ns
e
m
bl
e
0.974
0.974
0.974
0.974
M
S
O
0.8
0.917
0.353
0.857
T
L
E
0.86
0.86
0.95
0.9
XGB
0.967
0.972
0.97
0.971
M
L
E
L
-
XGB
0.9992
0.999
0.9992
0.9991
3.2.
S
t
u
d
e
n
t
e
n
gage
m
e
n
t
an
al
ys
is
T
he
s
tu
de
nt
e
nga
ge
m
e
nt
a
n
a
ly
s
is
i
s
done
us
in
g
e
nga
ge
m
e
n
t
-
or
ie
nt
e
d
da
ta
s
e
t
obt
a
in
e
d.
V
a
r
io
us
m
ode
ls
li
ke
X
G
B
-
ba
s
e
d,
M
S
O
,
R
F
-
ba
s
e
d
,
a
nd
T
L
E
a
r
e
us
e
d
to
c
om
pa
r
e
w
it
h
c
l
a
s
s
if
ic
a
ti
on
out
c
om
e
of
pr
opos
e
d
M
L
E
L
-
X
G
B
.
T
he
F
ig
ur
e
6 s
how
s
th
e
a
c
c
ur
a
c
y
out
c
o
m
e
a
c
hi
e
ve
d
us
in
g
di
f
f
e
r
e
nt
m
ode
ls
;
th
e
r
e
s
ul
t
s
how
s
th
e
pr
opos
e
d
M
L
E
L
-
X
G
B
a
c
hi
e
ve
s
a
n
a
c
c
ur
a
c
y
of
0.9
992
a
nd
th
e
ne
xt
be
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d
u
s
in
g
RF
-
e
ns
e
m
bl
e
w
it
h 0.974, XG
B
w
it
h 0.967, T
L
E
w
it
h 0.86
,
a
nd M
S
O
w
it
h 0.8.
F
ig
ur
e
6. A
c
c
ur
a
c
y outc
om
e
s
t
o a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
i
ng s
e
s
s
io
n a
c
ti
vi
ty
F
ig
ur
e
7
s
how
s
th
e
pr
e
c
is
io
n
out
c
om
e
a
c
hi
e
ve
d
us
in
g
di
f
f
e
r
e
nt
m
ode
ls
.
T
he
r
e
s
ul
t
s
how
s
th
e
pr
opos
e
d
M
L
E
L
-
X
G
B
a
c
hi
e
ve
s
a
n
a
c
c
ur
a
c
y
of
0.999
a
nd
th
e
n
e
xt
be
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d
us
in
g
R
F
-
e
ns
e
m
bl
e
w
it
h
0.974,
X
G
B
w
it
h
0.972,
M
S
O
w
it
h
0.917
,
a
nd
T
L
E
w
it
h
0.86.
F
ig
ur
e
8
s
how
s
th
e
r
e
c
a
ll
out
c
om
e
a
c
hi
e
ve
d
u
s
in
g
di
f
f
e
r
e
nt
m
ode
ls
;
th
e
r
e
s
ul
t
s
how
s
th
e
pr
opos
e
d
M
L
E
L
-
X
G
B
a
c
hi
e
ve
s
a
n
a
c
c
ur
a
c
y
of
0.9992
a
nd t
he
ne
xt
be
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d us
in
g R
F
-
e
ns
e
m
bl
e
w
it
h 0.9
74, XG
B
w
it
h 0.97,
T
L
E
w
it
h 0.95
,
a
nd M
S
O
w
it
h 0.353.
0.75
0.8
0.85
0.9
0.95
1
1.05
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
F1
-
s
c
or
e
P
r
e
di
c
i
t
v
e
m
ode
l
F
-
sco
re
(per
form
an
ce)
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
0
0.2
0.4
0.6
0.8
1
1.2
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
A
c
c
u
r
a
c
y
P
r
e
di
c
i
t
v
e
m
ode
l
A
ccu
racy
(en
gag
em
en
t)
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
6
,
D
e
c
e
m
be
r
20
25
:
5240
-
5250
5248
F
ig
ur
e
7. P
r
e
c
is
io
n outc
om
e
s
t
o a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
i
ng s
e
s
s
io
n a
c
ti
vi
ty
F
ig
ur
e
8. R
e
c
a
ll
out
c
om
e
s
t
o a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
in
g
s
e
s
s
io
n a
c
ti
vi
ty
F
ig
ur
e
9
s
how
s
th
e
F
1
-
s
c
or
e
out
c
om
e
a
c
hi
e
ve
d
u
s
in
g
di
f
f
e
r
e
nt
m
ode
ls
.
T
he
r
e
s
ul
t
s
how
s
th
e
pr
opos
e
d
M
L
E
L
-
X
G
B
a
c
hi
e
ve
s
a
n
a
c
c
ur
a
c
y
of
0.9991
a
nd
th
e
ne
xt
b
e
s
t
r
e
s
ul
t
is
a
c
hi
e
ve
d
us
in
g
RF
-
e
ns
e
m
bl
e
w
it
h
0.974,
X
G
B
w
it
h
0.971,
T
L
E
w
it
h
0.9
,
a
nd
M
S
O
w
it
h
0.857.
T
a
bl
e
2
s
how
s
th
e
c
om
pa
r
a
ti
ve
s
tu
dy
of
a
ll
m
e
tr
ic
s
of
di
f
f
e
r
e
nt
c
la
s
s
if
ic
a
ti
on
m
ode
l
te
s
ti
ng
on
pe
r
f
or
m
a
nc
e
r
e
la
te
d
da
ta
s
e
t.
T
he
ove
r
a
ll
r
e
s
ul
ts
a
c
hi
e
ve
d
s
how
th
e
pr
opos
e
d
m
ode
l
a
c
hi
e
v
e
s
m
uc
h
hi
ghe
r
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F1
-
s
c
or
e
i
n c
om
pa
r
is
on w
it
h X
G
B
-
ba
s
e
d, M
S
O
, R
F
-
ba
s
e
d
,
a
nd
T
L
E
.
F
ig
ur
e
9. F
1
-
s
c
or
e
out
c
om
e
s
t
o a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
us
in
g s
e
s
s
io
n a
c
ti
vi
ty
T
a
bl
e
2. C
om
pa
r
a
ti
ve
s
tu
dy f
or
e
nga
ge
m
e
nt
da
ta
s
e
t
P
r
e
di
c
t
i
ve
m
ode
l
A
c
c
ur
a
c
y
P
r
e
c
i
s
i
on
R
e
c
a
l
l
F
1
-
s
c
or
e
RF
-
e
ns
e
m
bl
e
0.974
0.974
0.974
0.974
M
S
O
0.8
0.917
0.353
0.857
T
L
E
0.86
0.86
0.95
0.9
XGB
0.967
0.972
0.97
0.971
M
L
E
L
-
XGB
0.9992
0.999
0.9992
0.9991
0.75
0.8
0.85
0.9
0.95
1
1.05
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
P
r
e
c
i
s
i
on
P
r
e
di
c
t
i
on
m
ode
l
s
Preci
si
on
(en
gagem
en
t)
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
0
0.2
0.4
0.6
0.8
1
1.2
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
R
e
c
a
l
l
P
r
e
di
c
t
i
on
m
ode
l
s
Rec
al
l
(e
n
ga
ge
m
en
t)
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
0.75
0.8
0.85
0.9
0.95
1
1.05
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
F1
-
S
c
o
r
e
P
r
e
di
c
i
t
v
e
m
ode
l
F1
-
score
(
e
n
g
a
g
e
m
e
n
t
)
R
F
-
E
ns
e
m
bl
e
M
S
O
T
L
E
XGB
M
L
E
L
-
X
G
B
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
w
e
b
-
bas
e
d l
e
ar
ni
ng plat
fo
r
m
t
o as
s
e
s
s
s
tu
d
e
nt
pe
r
fo
r
m
anc
e
u
s
in
g onli
ne
s
e
s
s
io
n
…
(
Shas
hi
r
e
k
ha
)
5249
4.
C
O
N
C
L
U
S
I
O
N
T
hi
s
pa
pe
r
in
tr
oduc
e
d
a
nove
l
e
ns
e
m
bl
e
m
ode
l
na
m
e
ly
M
L
E
L
-
X
G
B
to
a
c
c
ur
a
te
ly
a
s
s
e
s
s
s
tu
de
nt
pe
r
f
or
m
a
nc
e
a
nd
e
nga
ge
m
e
nt
le
ve
l
f
or
de
s
ig
ni
ng
e
f
f
e
c
ti
ve
e
L
e
a
r
ni
ng
-
ba
s
e
d
w
e
b
-
ba
s
e
d
to
ol
.
T
he
pr
opos
e
d
m
ode
l
is
s
tu
di
e
d
us
in
g
a
pe
r
f
o
r
m
a
nc
e
-
or
ie
nt
e
d
da
ta
s
e
t;
th
e
r
e
s
ul
ts
a
c
hi
e
ve
d
a
r
e
ve
r
y
pr
om
is
in
g.
F
ur
th
e
r
,
th
e
pr
opos
e
d
m
ode
l
is
va
li
da
te
d
us
in
g
a
n e
nga
ge
m
e
nt
-
or
ie
nt
e
d da
ta
s
e
t;
a
ve
r
y
good
pe
r
f
or
m
a
nc
e
is
a
c
hi
e
v
e
d.
C
ons
id
e
r
in
g
bot
h
th
e
da
ta
s
e
t
th
e
pr
opo
s
e
d
M
L
E
L
-
X
G
B
m
o
de
l
a
c
hi
e
ve
s
m
u
c
h
be
tt
e
r
pe
r
f
or
m
a
nc
e
th
a
n
X
G
B
-
ba
s
e
d,
M
S
O
,
R
F
-
e
ns
e
m
bl
e
,
a
nd
T
L
E
.
T
he
r
e
s
ul
t
s
how
s
r
obus
tn
e
s
s
of
pr
opos
e
d
c
la
s
s
if
ie
r
in
ha
ndl
in
g
m
ul
ti
-
la
be
l
c
la
s
s
if
ic
a
ti
on
da
ta
s
e
t.
T
he
f
ut
ur
e
w
or
k
w
oul
d
f
oc
us
on
a
na
ly
z
in
g
m
or
e
di
f
f
e
r
e
nt
ki
nd
of
da
ta
s
e
t
to
a
s
s
e
s
s
t
he
p
e
r
f
or
m
a
nc
e
of
pr
opos
e
d c
la
s
s
if
ie
r
.
A
C
K
N
O
WL
E
D
G
M
E
N
T
S
W
e
w
oul
d
li
ke
to
e
xt
e
nd
our
he
a
r
tf
e
lt
th
a
nks
to
our
gui
de
f
or
hi
s
unw
a
ve
r
in
g
gui
da
nc
e
,
in
va
lu
a
bl
e
in
s
ig
ht
s
, a
nd e
nc
our
a
ge
m
e
nt
t
hr
oughout t
he
r
e
s
e
a
r
c
h pr
oc
e
s
s
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
N
o f
undi
ng i
s
r
a
is
e
d f
or
t
hi
s
r
e
s
e
a
r
c
h.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
S
ha
s
hi
r
e
kha
H
a
num
a
nt
ha
ppa
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
he
ta
na
P
r
a
ka
s
h
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
T
he
a
ut
hor
s
s
t
a
te
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
D
a
ta
s
e
t
is
ut
il
iz
e
d i
n t
hi
s
r
e
s
e
a
r
c
h i
n r
e
f
e
r
e
nc
e
[
23]
.
R
E
F
E
R
E
N
C
E
S
[
1]
B
.
F
l
a
na
ga
n,
R
.
M
a
j
um
da
r
,
a
nd
H
.
O
ga
t
a
,
“
E
a
r
l
y
-
w
a
r
ni
ng
pr
e
di
c
t
i
on
of
s
t
ude
nt
pe
r
f
or
m
a
nc
e
a
nd
e
nga
ge
m
e
nt
i
n
ope
n
book
a
s
s
e
s
s
m
e
nt
by
r
e
a
di
ng
be
ha
vi
or
a
na
l
ys
i
s
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
E
duc
at
i
onal
T
e
c
hnol
ogy
i
n
H
i
ghe
r
E
duc
at
i
on
,
vol
.
19,
no.
1,
2022, doi
:
10.1186/
s
41239
-
022
-
00348
-
4.
[
2]
A
.
A
l
-
Z
a
w
qa
r
i
,
D
.
P
e
um
a
ns
,
a
nd
G
.
V
a
nde
r
s
t
e
e
n,
“
A
f
l
e
xi
bl
e
f
e
a
t
ur
e
s
e
l
e
c
t
i
on
a
ppr
oa
c
h
f
or
pr
e
di
c
t
i
ng
s
t
ude
nt
s
’
a
c
a
de
m
i
c
pe
r
f
or
m
a
nc
e
i
n onl
i
ne
c
our
s
e
s
,”
C
om
put
e
r
s
and E
duc
at
i
on:
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
, vol
. 3, 2022, doi
:
10.1016/
j
.c
a
e
a
i
.2022.100103.
[
3]
G
.
B
.
B
r
a
hi
m
,
“
P
r
e
di
c
t
i
ng
s
t
ude
nt
pe
r
f
o
r
m
a
nc
e
f
r
om
on
l
i
ne
e
nga
ge
m
e
nt
a
c
t
i
vi
t
i
e
s
us
i
ng
nove
l
s
t
a
t
i
s
t
i
c
a
l
f
e
a
t
ur
e
s
,”
A
r
abi
an
J
our
nal
f
or
Sc
i
e
nc
e
and E
ngi
ne
e
r
i
ng
, vol
. 47, no. 8, pp. 10225
–
10243, 2022, d
oi
:
10.1007/
s
13369
-
021
-
06548
-
w.
[
4]
K
.
J
a
w
a
d,
M
.
A
.
S
ha
h,
a
nd
M
.
T
a
hi
r
,
“
S
t
ude
nt
s
’
a
c
a
de
m
i
c
pe
r
f
or
m
a
nc
e
a
nd
e
nga
ge
m
e
nt
pr
e
di
c
t
i
on
i
n
a
vi
r
t
ua
l
l
e
a
r
ni
ng
e
nvi
r
onm
e
nt
us
i
ng r
a
ndom
f
or
e
s
t
w
i
t
h da
t
a
ba
l
a
nc
i
ng,”
Sus
t
ai
nabi
l
i
t
y
, vol
. 14, no. 22, 2022, doi
:
10.3390/
s
u142214795.
[
5]
M
.
Y
a
ğc
ı
,
“
E
duc
a
t
i
ona
l
da
t
a
m
i
ni
ng:
pr
e
di
c
t
i
on
of
s
t
ude
nt
s
’
a
c
a
de
m
i
c
pe
r
f
or
m
a
nc
e
us
i
ng
m
a
c
hi
ne
l
e
a
r
ni
ng
a
l
gor
i
t
hm
s
,”
Sm
a
r
t
L
e
ar
ni
ng E
nv
i
r
onm
e
nt
s
, vol
. 9, no. 1, 2022, doi
:
10.1186/
s
40561
-
022
-
00192
-
z.
[
6]
H
.
X
ue
a
nd
Y
.
N
i
u,
“
M
ul
t
i
-
out
put
ba
s
e
d
hybr
i
d
i
nt
e
gr
a
t
e
d
m
ode
l
s
f
or
s
t
ude
nt
pe
r
f
or
m
a
nc
e
pr
e
di
c
t
i
on,”
A
ppl
i
e
d
Sc
i
e
nc
e
s
,
vol
.
13,
no. 9, 2023, doi
:
10.3390/
a
pp13095384.
[
7]
M
.
N
.
I
nj
a
da
t
,
A
.
M
ouba
ye
d,
A
.
B
.
N
a
s
s
i
f
,
a
nd
A
.
S
ha
m
i
,
“
S
ys
t
e
m
a
t
i
c
e
ns
e
m
bl
e
m
ode
l
s
e
l
e
c
t
i
on
a
ppr
oa
c
h
f
or
e
duc
a
t
i
ona
l
da
t
a
m
i
ni
ng,”
K
now
l
e
dge
-
B
as
e
d Sy
s
t
e
m
s
, vol
. 200, 2020, doi
:
10.1016/
j
.knos
ys
.202
0.105992.
[
8]
S
.
C
.
M
.
S
unda
r
a
r
a
j
a
n,
G
.
U
.
M
a
he
s
w
a
r
i
,
P
.
K
a
ur
,
a
nd
A
.
K
a
us
hi
k,
“
H
e
a
r
t
di
s
e
a
s
e
s
di
a
gnos
i
s
us
i
ng
c
h
a
ot
i
c
H
a
r
r
i
s
H
a
w
k
opt
i
m
i
z
a
t
i
on
w
i
t
h
E
-
C
N
N
f
o
r
I
oM
T
f
r
a
m
e
w
or
k,”
I
nf
or
m
at
i
on
T
e
c
hnol
ogy
and
C
ont
r
ol
,
vol
.
52,
no.
2,
pp.
500
–
514,
2023,
doi
:
10.5755/
j
01.i
t
c
.52.2.32778.
Evaluation Warning : The document was created with Spire.PDF for Python.