Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
5
,
Octo
ber
201
9
, pp.
3872
~3
879
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v9
i
5
.
pp3872
-
38
79
3872
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Stu
dent risk iden
tific
atio
n
learning
model
usin
g
machine
learning
app
ro
ach
Susheel
amm
a K H,
K M
R
avikum
ar
S J C
Instit
u
te of
Technol
og
y
,
In
dia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
hist
or
y:
Re
cei
ved
N
ov
19
, 201
8
Re
vised
A
pr 12
, 2
01
9
Accepte
d
Apr 25
, 201
9
Sever
al
ch
allen
ges
are
associate
d
with
onli
n
e
base
d
le
arn
in
g
sy
st
ems
,
the
m
ost
important
of
whi
ch
is
the
l
ac
k
of
stu
dent
m
oti
va
ti
on
in
var
ious
cour
se
m
at
e
rials
and
for
var
ious
cour
se
activities
.
Further
,
i
t
is
i
m
porta
nt
t
o
ide
nti
f
y
s
tude
nt
who
are
a
t
risk
of
fai
l
ing
to
co
m
ple
te
th
e
cou
r
se
on
ti
m
e.
The
exi
st
ing
m
odel
s
appl
i
ed
m
ac
hine
l
ea
rnin
g
appr
oac
h
for
solving
it
.
How
eve
r,
the
se
m
odel
s
are
no
t
e
ffic
i
ent
as
th
e
y
a
re
tr
ai
ned
using
le
ga
c
y
da
t
a
and
al
so
fai
l
ed
to
addr
ess
imbala
nce
d
data
issues
for
both
tra
ini
ng
and
te
stin
g
the
cl
assifi
cation
appr
oa
ch.
Furth
er,
they
ar
e
not
e
ffi
cient
fo
r
c
la
ss
if
y
ing
n
ew
cour
ses.
For
over
coming
the
se
rese
arc
h
ch
allen
ges,
thi
s
work
pre
sente
d
a
novel
design
b
y
tra
ini
ng
the
l
ea
r
ning
m
odel
for
ide
nti
f
y
ing
r
isk
using
cur
ren
t
cour
ses.
Further
,
we
pr
ese
nt
an
XG
Boost
cl
assificat
ion
al
gor
ithm
tha
t
can
clas
sif
y
risk
for
new
cour
ses.
Expe
riments
a
re
conduc
t
ed
t
o
eva
luate
per
form
anc
e
of
proposed
m
odel
.
The
ou
tc
om
e
show
s
the
proposed
m
odel
at
t
ai
n
significan
t
per
form
anc
e
over
stat
-
of
-
a
rt
m
odel
in
te
rm
s
of
ROC,
F
-
m
ea
sure,
Prec
i
sion a
nd
R
ecal
l
.
Ke
yw
or
d
s
:
Cl
assifi
cat
ion
Im
balanced data
Ma
chine
le
a
rn
i
ng
Virtual lea
rn
i
ng e
nv
i
ronm
ent
Copyright
©
201
9
Instit
ute of
Ad
v
ance
d
Engi
ne
eri
ng
and
Sc
ie
n
ce
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Su
s
heelam
m
a K
H
,
Dep
a
rtm
ent
of
Inform
at
ion
science a
nd
E
ng
i
neer
i
ng
,
S J
C
Insti
tute of Tech
nolo
gy,
Chick
balla
pur
-
562101
,
I
ndia
.
Em
a
il
:
su
sh
eel
a.kh@
gm
ai
l.com
1.
INTROD
U
CTION
O
nl
i
ne
ba
s
e
d
l
e
a
r
ni
ng
s
y
s
t
em
ha
s
be
c
om
e
ge
ne
r
i
c
pl
a
t
f
or
m
i
n
e
du
c
a
t
i
on
a
nd
c
a
n
t
a
ke
m
a
ny
f
orm
s
,
f
r
om
l
e
a
r
ni
ng
m
a
na
ge
m
e
nt
sy
s
t
em
(
L
M
S
)
to
vi
r
t
ua
l
l
e
a
r
ni
ng
e
nv
i
r
on
m
e
nt
(
V
L
E
)
a
nd
m
a
s
s
i
ve
op
e
n
on
l
i
n
e
c
ou
r
s
e
s
(
M
O
O
C
s
)
.
I
n
M
O
O
C
s
,
s
t
ud
e
nt
s
c
a
n
l
e
a
r
n
a
ny
t
i
m
e
a
nd
f
r
o
m
a
ny
l
oc
a
t
i
on
[
1]
.
M
O
O
C
s
o
f
f
e
r
s
a
n
i
nn
o
va
t
i
ve
w
a
y
t
o
t
r
a
i
n
st
ud
e
n
t
s
,
c
ha
ng
e
t
he
s
t
a
t
e
-
of
-
a
r
t
m
et
ho
d
t
o
l
e
a
r
ni
ng
,
a
nd
a
t
t
r
a
c
t
stu
de
nt
s
f
r
om
a
rou
nd
t
he
gl
ob
e
.
T
he
w
e
l
l
-
kn
o
w
n
pl
a
t
f
or
m
s
a
r
e
C
our
s
e
r
a
,
H
a
r
va
r
d,
a
nd
E
d
x.
M
or
e
o
ve
r
,
M
O
O
C
s
ha
ve
c
on
t
r
i
bu
t
e
d
t
o
hi
gh
e
r
e
du
c
a
t
i
on
[
2]
.
I
n
M
O
O
C
s
a
nd
ot
he
r
on
l
i
ne
ba
s
e
d
l
e
a
r
ni
ng
s
y
s
t
em
s,
s
t
ud
e
nt
s
of
t
e
n
e
nr
ol
l
t
he
m
s
e
lv
e
s
t
o
do
w
nl
oa
d
m
a
t
er
i
a
l
s
a
nd
vi
de
o
s
bu
t
d
o
n
ot
f
i
ni
s
h
t
he
c
om
pl
e
te
c
ou
r
s
e
.
As
a
r
e
s
ul
t
,
t
he
t
ot
a
l
nu
m
be
r
of
a
c
t
i
vi
t
i
e
s
a
s
t
ud
e
nt
e
n
ga
ge
s
i
n
f
a
l
l
s
be
l
ow
t
he
r
e
c
om
m
e
nd
e
d
t
hr
e
s
h
ol
d
[
3]
.
T
he
r
e
f
or
e
,
t
e
a
c
he
r
s
m
ust
un
de
r
s
t
a
nd
t
he
e
ng
a
ge
m
e
nt
of
t
he
i
r
s
t
ud
e
nt
s
.
I
n
t
he
s
t
a
t
e
-
of
-
a
r
t
m
e
t
ho
d
t
o
e
du
c
a
t
i
on
,
t
r
a
i
ne
r
s
t
a
ke
va
r
i
o
us
m
e
a
s
ur
e
s
t
o
a
s
s
e
s
s
s
t
ud
e
nt
s
’
l
e
ve
l
s
of
pe
r
f
or
m
a
nc
e
,
m
ot
i
va
ti
on
,
a
n
d
e
n
ga
ge
m
e
nt
[
4]
,
s
uc
h
a
s
c
h
e
c
ki
n
g
s
t
u
de
nt
a
t
t
e
nd
a
nc
e
,
c
on
du
c
t
i
ng
e
xa
m
s
,
a
nd
m
on
it
or
i
ng
s
t
ud
y
i
ng
vi
a
C
C
TV
c
am
e
r
a
s
.
H
ow
e
ve
r
,
i
n
on
l
i
n
e
ba
s
e
d
l
e
a
r
ni
ng
s
y
s
t
em
t
he
r
e
a
r
e
no
f
a
c
e
-
to
-
f
a
c
e
m
e
et
i
ng
s
,
a
nd
i
t
i
s
c
ha
ll
e
ng
i
ng
t
o
de
t
e
rm
i
ne
s
t
ud
e
nt
e
ng
a
g
em
e
nt
l
e
ve
l
s
in
on
l
i
ne
a
c
t
i
vi
ti
e
s
s
uc
h
a
s
w
at
c
hi
n
g
vi
de
os
o
r
pa
r
t
i
c
i
pa
ti
ng
i
n
di
s
c
us
s
i
on
f
or
u
m
s
.
T
he
r
e
f
or
e
,
i
n
on
l
i
ne
ba
s
e
d
l
e
a
r
ni
ng
s
y
s
t
em
s
,
st
ud
e
nt
da
t
a
r
e
pr
e
s
e
nt
t
he
on
l
y
s
ou
r
c
e
o
f
i
nf
or
m
a
ti
on
t
hr
o
ug
h
w
hi
c
h
t
r
a
i
ne
r
s
c
a
n
a
s
s
e
s
s
s
t
ud
e
nt
pe
r
f
or
m
a
nc
e
a
nd
e
ng
a
ge
m
e
nt
.
Due
t
o
t
he
a
bs
e
nc
e
of
f
a
c
e
-
to
-
f
a
c
e
m
e
e
t
i
ng
s
,
o
nl
i
ne
ba
s
e
d
l
e
ar
ni
ng
s
y
s
t
em
s
fa
c
e
s
om
e
c
ha
l
le
ng
e
s
t
ha
t
ne
e
d
t
o
be
a
dd
r
e
s
s
e
d.
T
h
e
f
i
r
s
t
a
nd
m
os
t
im
po
r
t
a
nt
i
s
c
ou
r
s
e
d
r
o
p
o
ut
.
S
t
ud
e
nt
dr
o
po
ut
i
s
a
n
i
m
por
t
a
nt
pr
o
bl
em
a
c
r
os
s
va
r
i
o
us
l
e
ve
l
s
s
uc
h
pr
i
m
a
r
y
s
c
ho
ol
,
hi
g
he
r
s
e
c
on
da
r
y
,
g
r
a
du
a
t
i
on
l
e
ve
l
a
nd
t
he
s
c
e
na
r
i
o
i
s
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Student
risk
id
entif
ic
ation
le
arnin
g mode
l
usi
ng
m
ach
i
ne
le
ar
ni
ng
approac
h
(
Sus
heela
m
ma K H
)
3873
m
uc
h
w
or
s
e
i
n
M
O
O
C
s
.
A
s
pe
r
t
he
r
e
s
e
a
r
c
h
c
o
nd
uc
t
e
d
i
n
[
5
,
6]
,
t
he
nu
m
be
r
of
s
t
ud
e
nt
n
ot
c
om
pl
e
t
i
ng
gr
a
d
ua
t
i
on
i
n
U
S
A
i
s
2
0%
a
nd
i
n
E
ur
op
e
i
t
i
s
a
r
ou
nd
20
%
t
o
50
%
f
a
i
l
t
o
f
i
ni
s
h
t
he
i
r
s
t
ud
i
e
s
on
t
i
m
e
[
7]
.
F
or
o
nl
i
ne
or
di
s
t
a
nc
e
e
du
c
a
t
i
on
,
t
he
s
e
s
t
a
ti
s
t
ic
s
a
r
e
e
ve
n
w
or
s
e
w
i
t
h
7
8%
of
s
t
u
de
nt
s
no
t
c
om
pl
et
in
g
t
he
gr
a
d
ua
t
i
on
[
8]
.
F
ur
t
he
r
,
i
t
ge
t
s
e
ve
n
w
o
r
s
e
f
or
s
t
u
de
nt
w
ho
ge
t
s
r
e
gi
s
t
e
r
e
d
w
i
t
h
M
O
O
C
s
,
t
he
pe
r
c
e
nt
a
ge
o
f
s
t
ud
e
nt
w
h
o
e
n
r
ol
l
e
d
a
nd
s
uc
c
e
s
s
f
ul
l
y
f
i
ni
s
he
d
t
he
c
ou
r
s
e
i
s
on
l
y
5%
a
s
r
e
po
r
t
e
d
i
n
[
9]
or
15
%
a
s
r
e
po
r
t
e
d
i
n
[
10
]
.
T
he
i
s
s
ue
s
of
i
de
nt
i
f
y
i
ng
s
t
ud
e
nt
t
ha
t
a
re
e
xp
e
c
t
e
d
t
o
f
a
i
l
t
he
c
ou
r
s
e
h
a
s
be
e
n
e
xt
e
ns
i
ve
l
y
a
na
ly
z
e
d
a
c
r
os
s
va
r
i
ou
s
r
e
s
e
a
r
c
h
c
om
m
un
i
ty
i
n
r
e
c
e
nt
t
i
m
e
s
[
1
1
-
1
3]
.
I
t
w
a
s
a
l
s
o
a
m
aj
or
s
u
bj
e
c
t
of
t
he
K
D
D
'
C
U
P
20
1
5
c
om
pe
t
i
ti
o
n
t
ha
t
m
ai
nl
y
ai
m
ed
o
n
f
or
e
c
a
s
t
i
ng
s
t
ud
e
nt
w
i
t
hd
r
a
w
i
ng
f
r
om
onl
i
ne
c
ou
r
s
e
s
.
E
s
t
a
bl
i
s
hi
ng
s
tu
de
nt
,
w
ho
a
r
e
a
t
c
ha
nc
e
or
r
i
sk
of
w
i
t
hd
r
a
w
i
ng
or
f
a
i
l
i
ng
f
r
om
t
he
i
r
r
e
s
pe
ct
i
ve
c
ou
r
s
e
,
i
s
t
he
i
ni
ti
a
l
s
t
ep
t
ow
a
r
ds
pr
o
v
i
s
i
on
i
ng
t
he
m
w
i
t
h
r
em
e
di
al
(m
a
te
r
i
a
l
)
s
up
po
r
t
.
G
e
ne
r
a
l
l
y
,
s
up
p
or
t
i
ve
m
eas
ur
e
s
a
r
e
c
a
r
r
i
e
d
ou
t
by
i
ns
t
r
uc
t
or
/
pr
o
f
e
s
s
or
,
w
h
o
ob
t
a
i
ns
t
he
i
nf
or
m
at
i
on
/
ou
t
c
om
e
of
f
or
e
c
a
s
t
in
g
[
11
,
12
]
.
I
n
ot
he
r
w
a
y
,
t
he
f
or
e
c
a
s
t
i
ng
m
od
e
l
m
ay
bu
i
l
d
em
a
il
m
e
s
s
a
ge
s
t
ha
t
c
omm
uni
c
a
t
e
di
r
e
c
tl
y
t
o
t
he
s
t
ud
e
nt
[
14
]
.
T
he
p
r
e
l
im
i
na
r
y
ob
je
c
t
i
ve
i
s
t
o
e
n
ha
nc
e
t
he
s
t
ud
e
nt
l
e
a
r
ni
n
g,
t
o
ke
e
p
s
t
u
d
e
nt
e
ng
a
ge
d
i
n
c
ou
r
s
e
,
a
n
d
a
i
d
t
he
m
c
om
pl
e
t
i
ng
t
he
r
e
s
e
a
r
c
h
o
r
s
t
u
dy
pr
og
r
a
m
s
.
I
n
di
s
t
a
nc
e
o
r
on
l
i
ne
c
ou
r
s
e
s
,
m
os
t
m
a
t
e
r
i
al
s
a
r
e
de
l
i
ve
r
e
d
t
hr
ou
gh
V
i
r
t
u
a
l
L
e
a
r
ni
ng
E
nv
i
r
o
nm
e
nt
(
V
L
E
)
.
I
n
V
L
E
e
a
c
h
a
c
t
i
on
a
r
e
r
e
c
or
de
r
a
n
d
s
t
or
e
d
.
A
l
o
n
g
w
i
t
h,
s
t
u
de
nt
i
nf
or
m
a
t
i
on
s
uc
h
a
s
a
s
s
e
s
sm
en
t
,
t
a
s
k
r
e
s
ul
t
s
,
a
nd
de
m
og
r
a
ph
i
c
i
nf
o
r
m
a
ti
on
,
e
t
c
.
a
r
e
a
l
s
o
ke
pt
.
T
h
e
s
e
da
t
a
a
r
e
c
l
e
a
ns
e
d
a
nd
M
L
i
s
a
pp
l
i
e
d
t
o
b
ui
l
d
a
f
or
e
c
a
s
t
i
ng
/
pr
e
di
c
t
i
ve
m
od
e
l
.
T
he
s
e
m
od
e
l
s
a
r
e
t
he
n
us
e
d
t
o
o
f
f
e
r
o
nl
i
ne
c
ou
r
s
e
pr
ov
i
de
r
t
o
f
or
e
c
a
s
t
s
t
ud
e
nt
at
-
r
i
s
k
of
c
om
pl
e
t
i
ng
i
t
on
t
i
m
e
.
A
ge
ne
r
i
c
w
a
y
of
b
ui
l
di
ng
a
pr
e
di
c
t
i
ve
m
od
e
l
i
s
t
o
t
ra
i
n
t
he
m
od
e
l
s
us
i
ng
l
e
ga
c
y
da
t
a
f
r
o
m
a
hi
s
t
or
y
or
pr
e
vi
o
us
t
a
s
k
s
ub
m
it
t
e
d
i
nf
or
m
a
ti
on
of
t
he
c
ou
r
s
e
[
1
2]
.
F
ur
t
he
r
,
i
t
i
s
a
pp
l
i
e
d
t
o
t
he
pr
e
s
e
nt
p
r
e
s
e
nt
a
t
i
on
.
H
o
w
e
ve
r
,
a
d
op
t
i
ng
t
he
s
e
m
et
ho
ds
w
i
l
l
no
t
be
e
f
f
i
c
i
e
nt
w
he
n
a
p
p
l
i
e
d
t
o
ne
w
t
y
pe
o
f
c
ou
r
s
e
s
t
ha
t
ha
s
no
hi
s
t
or
y
.
F
or
s
uc
h
c
a
s
e
,
i
t
i
s
im
po
r
t
a
nt
t
o
f
i
nd
ne
w
s
ol
ut
i
on
.
F
r
om
e
xt
e
ns
i
ve
s
ur
ve
y
c
a
r
r
i
ed
ou
t
by
M
O
O
C
s
[
15
]
a
nd
H
i
gh
e
r
E
d
uc
a
t
i
on
(
H
E
)
c
ou
r
s
e
s
[
12
]
s
ho
w
s
t
ha
t
t
he
hi
gh
e
s
t
am
ou
nt
of
dr
o
po
ut
oc
c
ur
s
d
u
r
i
ng
f
i
r
s
t
y
e
a
r
’
s
c
ou
r
s
e
s
,
a
nd
m
a
ny
s
t
ud
e
nt
dr
op
ou
t
e
ve
n
w
i
t
hi
n
a
m
on
t
h/
f
i
r
st
f
e
w
w
e
e
ks
of
t
he
c
ou
r
s
e
pr
e
s
e
nt
a
t
i
on
.
T
he
c
a
us
e
m
a
y
be
a
l
s
o
du
e
t
o
f
e
e
pa
y
m
e
nt
t
ow
a
r
d
c
o
ur
s
e
s
.
T
he
r
e
f
o
r
e
,
t
he
ob
je
c
t
i
ve
i
s
t
o
e
s
t
a
bl
i
s
h
or
f
i
nd
s
t
ud
e
nt
w
ho
a
r
e
a
t
-
r
i
s
k
of
dr
op
pi
ng
o
ut
or
f
a
i
l
i
ng
t
o
c
om
pl
e
t
e
on
t
im
e
a
s
e
a
r
ly
a
s
po
s
s
i
bl
e
.
I
t
m
ust
a
l
s
o
be
no
t
e
d
t
ha
t
t
he
s
am
e
be
ha
vi
or
or
pa
t
t
e
r
n
m
ay
no
t
be
s
a
m
e
a
c
r
os
s
di
f
f
e
r
e
nt
u
ni
ve
r
s
i
t
y/
e
du
c
a
t
i
on
i
ns
t
it
ut
i
on
or
c
ou
r
s
e
de
s
i
g
n,
r
a
pi
d
s
t
ud
e
nt
dr
op
pi
ng
o
ut
of
c
ou
r
s
e
m
a
y
a
l
s
o
a
r
i
s
e
i
n
l
a
t
e
s
t
a
ge
of
c
ou
r
s
e
[
1
3]
.
F
ur
t
he
r
,
nu
m
be
r
of
M
a
c
hi
ne
l
e
a
r
ni
ng
ba
s
e
d t
e
c
hn
i
qu
e
i
s
wi
d
e
l
y
a
pp
l
i
e
d a
nd
us
e
d a
c
r
os
s
va
r
i
ou
s
s
t
a
t
e
-
of
-
a
r
t
m
od
e
l
s
[16
-
25
]
f
or
i
de
n
t
i
f
yi
ng
r
i
s
k
of
s
t
ud
e
nt
f
a
i
l
i
ng
t
o
c
om
pl
e
te
c
our
s
e
on
t
im
e
.
T
he
ba
s
i
c
c
on
c
e
pt
i
o
n
i
s
t
o
ut
i
l
i
z
e
l
e
ga
c
y
da
t
a
t
o
l
e
a
r
n
t
he
f
or
e
c
a
s
t
in
g
m
od
e
l
s
a
nd
t
o
ut
i
l
i
z
e
t
he
s
e
a
pp
r
oa
c
he
s
t
o
pe
r
f
or
m
f
or
e
c
a
s
t
i
ng
on
c
ur
r
e
nt
c
ou
r
s
e
s
.
T
he
da
t
a
ca
n
a
i
d
t
he
c
ou
r
s
e
pr
ov
i
de
r
w
h
o
i
s
a
im
i
ng
t
o
a
dd
r
e
s
s
o
r
bu
i
l
d
po
l
i
c
i
e
s
t
o
e
nh
a
nc
e
t
he
s
t
ud
e
nt
pe
r
f
or
m
a
nc
e
(
s
t
ud
e
nt
r
e
t
e
nt
i
on
r
a
t
e
)
a
nd
s
t
ud
e
nt
dr
o
pp
i
n
g
ou
t
of
c
ou
r
s
e
s
or
f
a
i
li
ng
t
o
f
i
ni
s
h
o
n
t
im
e
.
I
n
[
16
]
,
th
e
a
pp
r
oa
c
he
s
f
or
f
i
n
di
ng
f
a
i
l
ur
e
or
s
uc
c
e
s
s
of
s
t
ud
e
nt
w
e
r
e
t
r
a
i
ne
d
us
i
ng
da
t
a
of
t
he
i
r
pr
i
or
s
t
ud
y
r
e
s
ul
t
.
I
t
c
a
n
be
s
e
e
n
t
ha
t
f
or
e
c
a
s
t
i
ng
f
a
i
l
ur
e
f
or
t
he
f
i
r
s
t
t
e
rm
of
c
ou
r
s
e
s
i
s
ve
r
y
im
po
r
t
a
nt
,
s
i
nc
e
t
he
dr
o
po
ut
r
a
t
e
i
s
ge
ne
r
a
l
l
y
hi
gh
e
r
b
ut
w
i
t
h
s
ui
t
a
bl
e
po
l
i
c
i
e
s
or
s
t
r
a
t
e
gi
e
s
(
he
l
p)
m
a
ny
s
t
ud
e
nt
c
a
n
be
s
a
ve
d
[
23
]
.
B
e
ha
vi
or
of
s
t
ud
e
nt
s
[
2
4
,
2
5
]
i
n
t
he
V
L
E
c
a
n
be
us
e
d
t
o
c
on
s
t
r
uc
t
f
o
r
e
c
a
s
t
i
ng
m
od
e
l
s
f
o
r
on
l
i
ne
c
o
ur
s
e
s
.
T
he
s
e
c
ou
l
d
b
e
ju
s
t
s
im
pl
e
sum
m
a
r
y
s
t
at
i
sti
c
s
[
19
]
.
Whe
n
ne
i
t
he
r
t
he
s
tu
de
nt
s
vi
r
t
ua
l
l
e
a
r
ni
ng
e
nv
i
r
o
nm
e
nt
a
c
t
i
vi
t
i
e
s
no
r
t
he
s
t
ud
e
nt
pr
i
o
r
s
t
ud
y
r
e
s
ul
t
s
a
r
e
a
va
i
l
a
bl
e
,
de
m
og
r
a
ph
i
c
i
nf
o
r
m
at
i
on
c
a
n
be
us
e
d
a
s
t
he
m
a
jo
r
f
ou
nd
a
t
i
on
of
i
nf
o
r
m
at
i
on
[
20
]
.
M
a
ny
a
pp
r
oa
c
he
d
[
1
1
,
2
6,
2
7
]
f
or
s
ol
vi
ng
pr
ob
l
e
m
of
c
l
a
s
s
if
i
c
a
t
i
on
w
i
t
h
pr
e
s
e
nc
e
of
im
ba
l
a
nc
e
d
da
t
a
i
n
f
or
e
c
a
s
t
i
ng
or
i
de
nt
i
f
y
i
ng
s
t
ud
e
nt
a
t
-
r
i
s
k
o
f
f
a
i
l
i
ng
.
H
ow
e
ve
r
,
t
he
y
ne
gl
e
c
t
e
d
s
t
ud
e
nt
w
ho
ha
ve
n’
t
s
h
o
w
n
a
ny
i
nt
e
r
e
s
t
i
n
pe
r
f
or
m
i
ng
t
a
s
ks
a
nd
on
l
y
f
oc
us
e
d
o
n
a
c
t
i
ve
s
t
ud
e
nt
s
.
F
or
o
ve
r
c
om
i
ng
r
e
s
e
a
r
c
h
c
ha
l
l
e
ng
e
s
t
hi
s
w
or
k,
t
hi
s
w
o
r
k
a
i
m
e
d
a
t
de
s
i
gn
i
ng
a
f
or
e
c
a
s
t
i
ng
m
od
e
l
t
ha
t
i
de
nt
i
f
y
s
t
ud
e
nt
a
t
-
r
i
s
k
o
f
f
a
i
li
ng
o
r
c
om
pl
e
ti
ng
c
ou
r
s
e
on
t
i
m
e
by
pr
e
s
e
nt
i
ng
a
no
ve
l
X
G
B
oo
s
t
c
l
a
s
s
i
fi
c
a
ti
on
m
od
e
l
.
T
he
pr
op
os
e
d
l
e
a
r
ni
ng
m
od
e
l
i
s
c
on
s
t
r
uc
t
e
d
us
i
ng
s
t
a
t
e
-
of
-
a
r
t
m
od
e
l
s
a
t
t
he
O
U
[
1
7
,
21
-
2
3]
.
I
ni
t
i
a
l
ly
,
us
i
ng
de
c
i
s
i
on
t
r
e
e
t
ha
t
i
s
t
r
a
i
ne
d
us
i
n
g
da
t
a
l
a
be
l
i
ng
s
t
ud
e
nt
be
ha
vi
or
i
n
t
he
vi
r
t
ua
l
l
e
a
r
ni
ng
e
nv
i
r
o
nm
e
nt
c
om
pl
em
e
nt
e
d
by
t
he
s
c
or
e
s
of
t
he
pa
s
t
a
s
s
e
s
sm
e
nt
s/
t
a
s
ks
[
2
1]
.
F
ur
t
he
r
,
[
22
]
us
e
d
de
m
og
r
a
ph
i
c
f
e
a
t
ur
e
s
f
o
r
e
nr
i
c
hi
ng
t
he
i
np
ut
da
t
a
f
or
t
r
a
i
ni
ng
m
od
e
l
.
T
he
s
i
gn
i
f
i
c
a
nt
di
s
c
ov
e
r
y
i
n
[
23
]
w
a
s
t
he
pr
om
i
ne
nc
e
of
t
h
e
e
a
r
l
y
e
s
t
a
bl
i
s
hm
e
nt
or
f
i
nd
i
n
g
of
s
t
ud
e
nt
s
a
t
r
i
s
k,
e
ve
n
pr
i
or
t
o
t
he
f
i
r
s
t
ta
s
k/
a
ss
e
s
sm
e
nt
i
n
t
he
c
ou
r
s
e
.
T
he
s
t
ud
e
nt
s
w
ho
d
o
no
t
s
u
bm
i
t
or
f
a
i
l
t
o
c
om
pl
et
e
t
he
a
s
s
e
s
sm
e
nt
a
r
e
ve
r
y
l
i
ke
l
y
t
o
f
a
i
l
or
w
i
t
hd
r
a
w
t
he
e
nt
i
r
e
c
ou
r
s
e
.
F
ur
t
he
r
,
t
he
p
r
im
a
ry
r
e
a
s
on
s
t
o
us
e
t
hi
s
a
l
g
or
i
t
hm
a
r
e
it
s
a
c
c
ur
a
c
y
,
e
f
f
i
c
i
e
nc
y
,
a
nd
f
e
a
s
i
bi
l
it
y
.
I
t
’
s
a
l
i
ne
a
r
m
od
e
l
a
nd
a
t
r
e
e
l
e
a
r
ni
ng
a
l
g
or
i
t
hm
t
ha
t
do
e
s
pa
r
a
l
l
e
l
c
om
pu
t
at
i
on
s
on
a
s
i
ng
l
e
m
a
ch
i
ne
.
I
t
a
l
s
o
h
a
s
e
xt
r
a
f
e
a
t
ur
e
s
f
o
r
d
oi
ng
c
r
os
s
va
l
i
da
t
i
on
a
nd
c
om
put
i
ng
f
e
a
t
ur
e
i
m
po
r
t
a
nc
e
.
B
e
lo
w
a
r
e
s
om
e
of
t
he
m
ai
n
f
e
a
tu
r
e
s
o
f
t
he
m
od
e
l
:
−
S
pa
r
s
i
t
y
:
I
t
a
c
ce
pt
s
s
pa
r
s
e
i
np
ut
f
or
t
r
e
e
bo
os
t
e
r
a
nd
l
i
ne
a
r
b
oo
s
t
e
r
.
−
C
us
t
om
i
z
at
i
on
:
I
t
s
up
po
r
t
s
c
us
t
om
i
ze
d
ob
je
c
t
iv
e
a
nd
e
va
l
ua
t
i
on
f
u
nc
t
i
on
s
.
−
D
M
a
t
r
i
x:
I
t
s
op
t
im
i
z
e
d
da
t
a
s
t
r
uc
t
ur
e
t
ha
t
im
pr
o
ve
s
i
t
s
pe
r
f
o
r
m
a
nc
e
a
nd
e
f
f
i
c
i
e
nc
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
8
7
2
-
3
8
7
9
3874
The
c
ontrib
ution o
f wor
k
is a
s foll
ow
s
−
P
r
e
s
e
nt
i
ng
a
n
X
G
B
o
os
t
c
l
a
s
s
i
f
i
c
a
ti
on
m
od
e
l
f
or
i
de
nt
i
f
y
i
ng
s
t
ud
e
nt
r
i
s
k
of
f
a
i
l
ur
e
.
−
T
he
X
G
B
oo
s
t
c
a
n
be
us
e
d
a
s
bo
t
h
bi
na
r
y
c
l
a
s
s
i
f
i
e
r
a
s
w
e
l
l
as
m
ul
ti
-
l
e
ve
l
c
la
s
s
i
f
i
e
r
.
−
T
he
pr
op
os
e
d
m
od
e
l
a
dd
r
e
s
s
e
s
im
ba
l
a
nc
e
d
da
t
a
i
s
s
ue
a
nd
a
ls
o
c
a
n
f
or
e
c
a
s
t
r
i
s
k
f
o
r
ne
w
c
o
ur
s
e
t
ha
t
ha
ve
no
hi
s
t
or
y
.
−
O
ur
m
od
e
l
a
t
t
ai
ns
go
od
s
pe
e
d
a
nd
a
c
c
ur
a
c
y
p
e
r
f
or
m
a
nc
e
w
h
e
n
c
om
pa
r
e
d
w
i
t
h
s
t
a
te
-
of
-
a
r
t
m
od
e
l
.
−
E
xp
e
r
i
m
e
nt
ou
t
c
om
e
s
ho
w
s
go
od
pe
r
f
or
m
a
nc
e
i
n
t
e
rm
s
of
R
O
C
,
F
-
m
e
a
s
ur
e
,
a
nd
pr
e
c
i
s
i
on
a
nd
r
e
c
a
l
l
.
T
he
pa
pe
r
i
s
o
r
ga
ni
z
e
d
a
s
f
ol
l
ow
s
:
I
n
s
e
c
t
i
on
2
t
he
p
r
o
po
s
e
d
s
t
ud
e
nt
r
i
s
k
i
de
nt
i
f
i
c
a
ti
on
m
od
e
l
us
i
ng
X
G
B
o
os
t
a
l
go
r
i
t
hm
i
s
pr
e
s
e
nt
e
d.
E
xp
e
r
im
e
nt
a
l
s
t
ud
i
e
s
a
r
e
di
s
c
us
s
e
d
i
n
s
e
c
ti
on
3.
F
i
na
l
l
y
se
c
t
i
on
4
t
he
pa
pe
r
i
s
c
on
c
l
ud
e
d
a
n
d
f
ut
ur
e
w
o
r
k
of
r
e
s
e
a
r
c
h
i
s
de
s
c
r
i
be
d.
2.
R
E
S
E
A
R
C
H
M
E
T
H
O
D
–
S
T
U
D
E
N
T
R
I
S
K
I
D
E
N
T
I
F
I
C
A
T
I
O
N
L
E
A
R
N
I
N
G
M
O
D
E
L
U
S
I
N
G
XGB
O
O
S
T
C
L
A
S
S
I
F
I
C
A
T
I
O
N
A
L
G
O
R
I
T
H
M
T
hi
s
pa
pe
r
pr
e
s
e
nt
a
n
ov
e
l
l
e
a
r
ni
ng
de
s
i
g
n
t
ha
t
us
e
da
ta
f
r
om
r
un
ni
n
g
pr
e
s
e
nt
a
t
i
on
f
or
t
r
a
i
ni
ng
f
or
e
c
a
s
t
i
ng
m
od
e
l
.
T
he
f
un
da
m
e
nt
a
l
ob
je
c
t
i
ve
s
i
s
t
o
u
s
e
t
he
i
nf
o
r
m
a
t
i
on
of
s
t
ud
e
nt
s
w
h
o
ha
ve
a
l
r
e
a
dy
c
om
pl
e
t
e
d
a
nd
s
ub
m
i
t
te
d
t
he
f
ut
ur
e
t
a
s
k
a
n
d
a
na
l
y
z
e
t
he
be
ha
vi
o
r
pa
t
t
e
r
n
of
t
he
s
t
ud
e
nt
s
w
ho
a
r
e
a
t
r
i
s
k
of
f
a
i
l
i
ng
t
o
s
ub
m
i
t
t
he
a
s
s
i
gnm
en
t
.
I
t
i
s
a
s
s
um
ed
t
ha
t
t
he
be
ha
vi
or
pa
t
t
e
r
n
of
s
t
ud
e
nt
w
ho
a
r
e
a
bo
ut
t
o
s
ub
m
it
a
r
e
e
xp
e
c
t
e
d
t
o
f
ol
l
ow
i
de
nt
i
c
a
l
be
ha
vi
or
pa
t
t
e
r
n
a
s
t
ho
s
e
w
h
o
a
l
r
e
a
dy
c
om
pl
e
te
d
a
nd
s
ub
m
i
tt
e
d
t
he
t
a
s
k
s
im
i
l
a
r
ly
,
t
he
b
e
ha
vi
or
p
a
t
t
e
r
n
w
i
ll
di
f
f
e
r
e
nt
f
or
s
t
ud
e
nt
w
ho
d
on
’
t
c
om
pl
et
e
or
s
ub
m
it
t
he
i
r
t
a
s
k.
N
um
be
r
of
m
ac
hi
ne
l
e
a
r
ni
ng
ba
s
e
d
c
l
a
s
s
i
f
ic
a
ti
on
m
od
e
l
i
s
a
va
i
l
ab
l
e
t
o
ut
i
l
i
z
e
an
d
a
t
t
a
i
n
e
f
f
i
c
i
e
nt
l
e
a
r
ni
ng
m
od
e
l
.
H
ow
e
ve
r
,
i
n
t
hi
s
w
or
k,
w
e
pr
e
s
e
nt
a
c
l
a
s
si
f
i
c
at
i
on
m
od
e
l
a
s
a
bi
na
r
y
c
l
a
s
s
if
i
c
a
t
i
on
pr
ob
l
e
m
.
H
ow
e
ve
r
,
i
t
c
a
n
w
or
k
e
v
e
n
f
or
s
ol
vi
ng
m
ul
ti
-
la
be
l
c
l
a
s
si
f
i
c
at
i
on
pr
ob
l
e
m
.
Th
a
t
i
s
,
f
or
a
gi
ve
n
da
y
(
pr
e
s
e
nt
)
,
w
hi
c
h
i
s
da
y
s
be
f
or
e
de
a
dl
i
ne
da
t
a
,
t
he
ob
je
c
t
i
ve
of
t
hi
s
w
or
k
i
s
t
o
bu
i
l
d
a
bi
na
r
y
c
l
a
s
s
i
f
ic
a
t
i
on
a
l
go
r
i
t
h
m
t
ha
t
f
or
e
c
a
s
t
w
he
t
he
r
t
he
s
t
ud
e
nt
w
i
l
l
s
ubm
i
t
t
he
a
s
s
i
gnm
e
nt
or
no
t
on
/
be
f
or
e
t
i
m
e
(
i
.
e
.
,
w
it
hi
n
t
he
f
ut
ur
e
da
y
s
)
.
I
f
=
0
,
f
or
e
c
a
s
t
i
ng
a
r
e
do
ne
on
t
he
de
a
dl
i
ne
da
y
.
O
nl
y
s
t
ud
e
nt
s
t
ha
t
a
r
e
e
nr
ol
l
e
d
i
n
c
o
ur
s
e
a
nd
ha
ve
n
’
t
f
i
ni
s
he
d
t
he
t
a
s
k
y
e
t
a
r
e
c
on
s
i
de
r
e
d
f
or
t
he
f
o
r
e
c
a
s
t
i
ng
.
2.1.
Sy
s
tem m
od
el
L
e
t
’
s
c
on
s
i
de
r
t
he
de
a
dl
i
ne
da
t
a
a
nd
t
he
da
t
e
w
he
n
t
he
f
or
e
c
a
s
t
i
ng
i
s
do
ne
,
w
hi
c
h
i
s
da
ys
pr
i
or
t
o
t
he
de
a
dl
i
ne
da
y
,
a
s
f
or
e
c
a
s
t
in
g
da
t
e
.
F
or
a
bl
e
t
o
c
on
s
t
r
uc
t
a
f
or
e
c
a
s
t
i
ng
m
od
e
l
f
or
pe
r
i
od
[
f
o
r
e
c
a
s
t
i
ng
da
t
e
;
de
a
dl
i
ne
da
t
e
]
s
uc
h
t
ha
t
de
a
d
l
i
ne
da
t
e
i
s
e
qu
a
l
t
o
f
or
e
c
a
s
t
i
ng
da
t
e
.
T
he
f
or
e
c
a
s
t
i
ng
da
t
e
an
d
de
a
dl
i
ne
da
y
c
a
n
be
e
s
t
a
bl
i
she
d
a
s
a
t
e
m
pl
a
t
e
f
or
e
c
a
s
t
i
ng
a
nd
de
a
dl
i
ne
d
a
y
s
,
r
e
s
pe
c
t
i
ve
ly
.
H
e
r
e
,
t
he
de
a
dl
i
ne
i
s
w
i
t
hi
n
t
hr
e
e
da
y
s
f
r
om
t
he
pr
e
s
e
nt
da
y
a
nd
w
e
w
a
nt
t
o
f
or
e
c
a
s
t
i
f
s
e
t
of
s
t
ud
e
nt
s
ub
m
it
t
he
i
r
a
s
s
e
s
sm
en
t
or
t
a
s
k
e
i
t
he
r
t
od
a
y
or
wi
t
hi
n
ne
xt
5
da
y
s
.
T
he
i
nf
or
m
at
i
on
f
or
t
h
e
pr
e
s
e
nt
da
y
a
r
e
i
na
c
c
e
s
s
i
bl
e
,
s
o
t
he
t
r
a
i
ni
n
g
da
t
a
w
i
l
l
c
om
e
f
r
om
t
he
da
y
s
[
p
r
e
s
e
nt
a
t
i
on
i
ni
t
i
la
i
za
e
d+
5]
=
1
0
w
i
t
h
t
he
l
a
be
l
s
of
s
u
bm
i
s
s
i
on
i
n
[
pr
e
s
e
nt
+
4;
p
r
e
s
e
nt
+
1]
=
[
9;
6]
.
I
t
s
ho
w
s
t
he
vi
r
t
ua
l
vi
e
w
o
f
t
h
e
da
y
s
f
o
r
t
r
a
i
n
i
ng
a
n
d
t
e
s
t
i
ng
da
t
a
,
da
y
=
0
de
pi
c
t
s
t
he
pr
e
s
e
nt
da
y
,
ne
ga
t
i
ve
ke
y
s
s
ho
w
s
t
o
k
no
w
n
i
nf
o
r
m
at
i
on
a
nd
po
s
i
t
i
ve
ke
y
s
t
o
ne
w
/
u
nk
no
w
n
da
t
a
.
T
hi
s
a
i
ds
,
t
ha
t
w
e
ha
ve
m
or
e
da
y
s
va
c
a
nt
w
he
n
a
p
pl
y
i
ng
t
he
f
or
e
c
a
s
t
i
ng
m
od
e
l
,
s
om
e
pr
e
vi
ou
s
/
ol
de
r
da
y
s
c
a
nn
ot
be
ut
i
li
z
e
d
a
s
t
he
y
w
e
r
e
no
t
pr
e
s
e
nt
i
n
t
r
a
i
ni
ng
s
t
a
ge
.
2.2.
Window
tr
ad
e
off and
f
e
atur
e sele
ction
m
odel
for lear
nin
g
B
a
s
e
d
on
s
y
s
t
em
m
od
el
de
s
c
r
i
be
d,
us
i
ng
l
on
g
-
t
e
r
m
hi
s
t
or
y
m
e
a
ns
t
he
w
i
ndow
s
a
m
pl
i
ng
f
or
l
a
be
l
s
i
s
gr
o
w
i
ng
.
T
he
m
or
e
da
y
s
pr
i
or
t
o
t
he
de
a
dl
i
ne
da
t
e
,
t
he
m
ore
da
y
s
i
s
r
e
qu
i
r
e
d
f
o
r
t
r
a
i
ni
ng
l
a
be
l
s
.
T
he
c
on
di
t
i
on
f
or
t
he
pr
e
s
e
nt
da
y
be
i
ng
0
t
o
5
da
y
s
pr
i
or
t
o
t
he
de
a
dl
i
ne
da
t
e
.
F
or
da
y
s
pr
i
or
t
o
t
he
de
a
dl
i
ne
,
t
he
s
i
z
e
of
t
he
w
i
nd
ow
f
or
b
ot
h
t
r
a
i
ni
ng
a
nd
t
e
s
t
i
ng
l
a
be
l
s
w
i
ll
be
+
1
.
T
he
da
t
a
a
va
i
la
bl
e
f
or
e
f
f
i
c
i
e
nt
l
e
a
r
ni
ng
i
s
c
om
po
s
e
d
of
i
nf
or
m
at
i
on
s
uc
h
a
s
a
c
t
i
vi
t
ie
s
a
nd
de
m
og
r
a
ph
i
c
s
i
n
t
h
e
vi
r
t
ua
l
le
a
r
ni
ng
e
n
vi
r
on
m
e
nt
.
F
or
e
xt
e
ns
i
ve
a
na
l
y
s
i
s
t
hi
s
w
or
k
c
a
r
r
i
e
d
o
ut
,
i
t
ca
n
be
s
e
e
n
t
he
de
m
og
r
a
ph
i
c
da
t
a
i
s
s
t
a
ti
c
i
n
na
t
ur
e
,
i
t
is
im
po
r
t
a
nt
t
o
c
a
r
r
y
ou
t
t
r
a
ns
f
o
r
m
a
ti
on
of
t
he
s
e
i
nf
or
m
a
ti
on
,
s
uc
h
a
s
s
t
a
n
da
r
di
z
a
t
i
on
f
o
r
nu
m
e
r
i
c
al
da
ta
a
nd
ve
c
t
or
i
z
a
t
i
on
of
c
a
t
e
go
r
i
c
a
l
da
t
a
.
S
im
il
a
r
l
y
,
t
he
vi
r
t
ua
l
l
e
a
rni
ng
e
n
vi
r
o
nm
e
nt
da
t
a
a
r
e
ge
ne
r
a
l
l
y
a
r
e
c
om
po
s
e
d
of
ve
r
y
r
i
c
h
i
n
f
or
m
a
ti
on
s
uc
h
a
s
da
i
l
y
c
li
c
k
e
ve
nt
s
c
l
us
t
e
r
e
d
by
pr
e
c
i
s
e
a
c
t
i
on
,
i
.
e
.
,
s
t
ud
e
nt
ha
s
vi
e
w
e
d
15
t
im
e
s
a
pa
r
t
i
c
ul
a
r
do
c
um
e
nt
or
p
r
e
s
e
nt
a
t
i
on
r
e
s
e
a
r
c
h
m
at
e
r
ia
l
.
A
l
l
t
he
e
ve
nt
s
/
a
c
ti
on
s
a
r
e
c
l
us
t
e
r
e
d
i
nt
o
a
c
t
i
on
s
t
y
pe
s
s
uc
h
a
s
v
i
de
o,
r
e
s
o
ur
c
e
s
,
bl
og
s
,
e
t
c
.
F
o
r
a
gi
ve
n
da
y
(
pr
e
s
e
nt
)
w
he
n
t
he
a
l
go
r
i
t
hm
(
m
od
e
l
)
i
s
l
e
a
r
ne
d,
t
he
vi
r
t
ua
l
l
e
a
r
ni
n
g
e
nv
i
r
o
nm
e
nt
fe
a
t
ur
e
s
a
r
e
a
l
i
gn
e
d
i
n
r
e
ve
r
s
e
w
i
t
h
r
e
s
pe
c
t
t
o
t
im
e
on
pa
r
t
i
c
ul
a
r
da
y
s
,
i
.
e
.
,
da
y
0
i
s
t
he
pr
e
s
e
nt
d
a
y
,
da
y
1
i
s
de
pi
c
t
e
d
a
s
y
e
s
t
er
da
y
a
nd
s
o
o
n
.
T
he
ol
de
s
t
da
y
ut
i
li
z
e
d
f
or
t
r
a
i
ni
ng
i
s
t
he
da
y
t
ha
t
t
he
c
ou
r
s
e
i
s
in
i
t
i
al
i
z
e
d.
I
n
ad
di
t
i
on
t
o
vi
r
t
ua
l
le
a
r
ni
ng
e
nv
i
r
on
m
e
nt
da
il
y
c
ou
nt
s
,
i
t
’
s
li
ke
l
y
t
o
ob
t
a
i
n
va
r
i
ou
s
s
t
a
ti
s
ti
c
a
l
i
nf
or
m
a
ti
on
of
s
t
ud
e
nt
be
ha
vi
or
p
a
t
t
e
r
n
i
n
t
he
vi
r
t
ua
l
l
e
a
r
ni
ng
e
nv
i
r
o
nm
e
nt
,
suc
h
a
s
t
he
ho
w
l
on
g
(
da
y
s
)
a
s
t
ud
e
nt
i
s
a
ct
i
ve
i
n
t
he
vi
r
t
ua
l
l
e
a
r
ni
ng
e
nv
i
r
on
m
e
nt
(
i
.
e
.
,
w
he
n
a
pe
r
s
o
n
(
s
t
ud
e
nt
)
ha
s
l
a
s
t
a
c
c
e
ss
e
d
or
ha
s
l
og
ge
d
i
n)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Student
risk
id
entif
ic
ation
le
arnin
g mode
l
usi
ng
m
ach
i
ne
le
ar
ni
ng
approac
h
(
Sus
heela
m
ma K H
)
3875
2.3.
Ad
dres
sing
I
mba
l
an
ced
d
ata
pr
ob
le
m in
cl
as
sific
ati
on
T
he
M
a
c
hi
ne
L
e
a
r
ni
ng
(
M
L
)
a
l
go
r
i
t
hm
s
a
re
ge
ne
r
a
l
l
y
m
od
e
l
l
e
d
t
o
l
e
a
r
n
o
bj
e
c
t
i
ve
pa
r
am
e
t
e
r
f
r
om
da
t
a
w
he
n
t
he
c
l
a
s
s
e
s
i
n
t
he
tr
a
i
ni
ng
i
nf
o
r
m
a
t
i
on
a
r
e
ba
l
a
nc
e
d.
H
o
w
e
ve
r
,
c
on
s
i
de
r
i
n
g
r
e
a
l
-
w
or
l
d
e
n
vi
r
o
nm
e
nt
,
t
he
da
t
a
a
r
e
ge
ne
r
a
l
l
y
im
ba
l
an
c
e
d
(
i
.
e
.
,
s
om
e
c
l
a
s
s
e
s
da
t
a
w
i
l
l
ha
ve
s
i
gn
i
f
i
c
a
nt
l
y
l
e
s
s
da
t
a
t
ha
n
ot
he
r
c
l
a
s
s
e
s
)
.
A
s
a
r
e
s
ul
t
,
t
he
s
t
a
te
-
of
-
a
r
t
a
l
go
r
i
t
hm
[
28
-
3
2]
pe
r
f
or
m
s
ve
r
y
po
or
i
n
i
de
nt
i
f
y
i
ng
pr
o
ba
bi
l
i
ty
of
r
i
s
k
of
f
a
i
lu
r
e
of
s
t
ud
e
nt
t
ha
t
ha
s
be
e
n
m
od
e
l
l
ed
s
o
f
a
r
[
1
4]
.
F
or
a
d
dr
e
s
s
i
ng
t
he
pr
ob
l
e
m
of
im
ba
la
nc
e
d
da
t
a
t
he
f
ol
l
ow
i
ng
t
w
o
s
t
a
ge
s
m
us
t
be
c
on
s
i
de
r
e
d
s
uc
h
a
s
:
a
)
A
l
go
r
i
t
h
m
s
t
a
ge
:
on
-
c
l
a
s
s
or
l
i
ne
a
r
c
l
a
s
s
i
f
ic
a
ti
on
m
od
e
l
s
,
c
os
t
-
s
en
s
i
t
i
ve
le
a
r
ni
ng
,
a
nd
va
r
i
ou
s
ki
n
d
o
f
e
ns
em
bl
e
al
go
r
i
t
hm
m
od
e
l
a
r
e
s
om
e
of
th
e
de
s
i
gn
s
t
ha
t
a
r
e
ge
ne
r
a
l
ly
us
e
d;
b)
D
a
t
a
s
t
ag
e
:
by
a
pp
l
y
i
ng
s
am
pl
i
ng
w
i
nd
ow
f
or
m
od
i
f
yi
ng
t
he
c
l
a
s
s
la
be
l
di
s
t
r
i
bu
t
io
n
i
n
s
uc
h
w
a
y
t
he
t
r
a
i
ni
ng
da
t
a
b
e
c
om
e
s
m
or
e
ba
l
a
nc
e
d.
T
he
ke
y
f
u
nc
t
i
on
a
l
i
ty
of
c
os
t
-
s
e
ns
i
t
i
ve
ba
s
e
d
l
e
a
r
ni
ng
m
od
e
l
i
s
t
o
pe
na
l
i
z
e
t
he
c
os
t
pa
r
a
m
e
t
e
r
e
r
r
or
o
n
m
a
r
gi
na
l
c
l
a
s
s
va
r
i
a
bl
e
du
r
i
ng
t
r
a
i
ni
ng
s
t
a
ge
,
w
hi
c
h
i
s
do
ne
by
us
i
ng
a
c
os
t
m
a
t
r
i
x.
H
ow
e
ve
r
,
f
or
a
t
t
a
i
ni
ng
f
i
ne
-
g
r
a
i
ne
d
bi
na
r
y
c
l
a
s
s
i
f
i
ca
t
i
on
m
od
e
l
i
t
i
s
be
t
t
e
r
t
o
f
i
x
t
he
w
e
i
gh
t
pa
r
a
m
et
e
r
f
or
m
i
no
r
i
t
y
c
la
s
s
es
(
i
.
e
.
,
by
c
on
s
i
de
r
i
ng
w
e
i
gh
t
of
m
a
jo
r
i
t
y
c
la
s
s
w
i
ll
be
1)
.
F
ur
t
he
r
,
n
um
be
r
of
a
pp
r
oa
c
he
d
[
1
1
,
26
]
,
a
nd
[
2
7]
f
or
s
ol
vi
ng
pr
ob
l
e
m
of
cl
a
s
s
i
f
i
ca
t
i
on
w
i
t
h
pr
e
s
e
nc
e
of
im
ba
la
nc
e
d
da
t
a
i
n
f
or
e
c
a
s
t
i
ng
or
i
de
nt
i
f
y
i
ng
s
t
ud
e
nt
a
t
-
r
i
s
k
of
f
a
i
li
ng
.
H
o
w
e
ve
r
,
t
he
y
ne
gl
e
c
te
d
s
t
ud
e
nt
w
h
o
ha
ve
n’
t
s
ho
w
n
a
ny
i
nt
e
r
e
s
t
i
n
pe
r
f
or
m
i
ng
t
a
s
ks
a
nd
on
l
y
f
oc
us
e
d
o
n
a
c
t
i
ve
s
t
ud
e
nt
s
.
2.4.
Forec
as
tin
g u
sing
X
GB
oo
s
t l
earning m
od
e
l
F
or
t
r
a
i
ni
ng
t
h
e
l
e
a
r
ni
ng
a
l
go
r
i
t
hm
s
a
nd
f
or
e
va
l
ua
t
i
on
of
o
u
r
m
od
e
l
,
t
hi
s
w
or
k
c
o
nd
uc
t
e
d
s
ur
ve
y
of
va
r
i
ou
s
e
xi
t
i
ng
m
a
c
hi
ne
l
e
a
r
ni
ng
ba
s
e
d
c
l
a
s
s
i
f
i
c
a
t
i
on
m
od
e
l
s
s
uc
h
a
s
l
og
i
s
t
ic
r
e
gr
e
s
s
i
on
,
N
a
i
ve
B
ay
e
s
,
S
up
p
or
t
V
e
c
t
or
M
a
c
hi
n
e
,
T
r
e
e
B
o
os
t
i
ng
,
R
a
nd
om
For
e
s
t
a
nd
s
o
o
n
.
H
ow
e
ve
r
,
t
he
s
e
m
od
e
l
s
a
r
e
no
t
e
f
f
i
c
i
e
nt
w
he
n
t
he
d
a
t
a
i
s
l
i
ne
a
r
l
y n
o
n
-
s
e
pa
r
a
bl
e
.
A
s
a
r
e
s
ul
t
,
i
nc
ur
de
gr
a
da
t
i
on
i
n
a
c
c
ur
a
c
y
of
c
l
a
s
s
i
f
ic
a
ti
on
p
e
r
f
or
m
a
nc
e
.
F
ur
t
he
r
,
ve
r
y
f
e
w
a
l
g
o
r
i
t
hm
pr
ov
i
s
i
o
n
p
r
o
ba
bi
l
i
s
t
ic
f
or
e
c
a
s
t
i
ng
.
A
s
t
hi
s
a
i
ds
i
n
o
r
de
r
i
n
g
s
t
u
de
nt
s
ba
s
e
d
o
n
t
he
i
r
l
i
ke
li
ne
s
s
t
o
f
a
i
l
,
a
nd
t
he
n
us
e
t
he
r
e
s
ou
r
c
e
s
c
on
s
t
r
a
i
nt
.
F
or
ov
e
r
c
om
in
g
r
e
s
e
a
r
c
h
c
ha
l
l
e
ng
e
s
,
t
hi
s
w
o
r
k
pr
e
s
e
n
t
X
G
B
o
os
t
c
l
a
s
s
i
f
i
c
at
i
on
a
l
go
r
i
t
hm
f
or
i
de
nt
i
f
y
i
ng
s
t
ud
e
nt
t
ha
t
f
a
i
l
t
o
c
om
pl
et
e
t
he
c
ou
r
s
e
o
n
t
im
e
.
X
G
B
o
os
t
i
s
r
o
bu
s
t
a
n
d
s
c
a
l
a
b
l
e
M
L
m
od
e
l
f
or
t
r
e
e
bo
os
t
i
ng
t
ha
t
is
de
ve
l
op
e
d
i
n
[
35
]
.
G
r
a
di
e
nt
bo
os
t
i
ng
i
s
t
h
e
ba
s
e
l
e
a
r
ni
ng
m
od
e
l
s
of
X
G
B
oo
s
t
c
l
a
s
s
i
f
i
c
a
ti
on
a
l
go
r
i
t
hm
,
w
hi
c
h
c
om
bi
ne
s
w
e
a
ke
r
ba
s
e
l
e
a
r
ni
ng
m
od
e
l
i
nt
o
a
s
t
r
on
g
e
r
l
e
a
r
ni
ng
m
od
e
l
i
n
a
n
i
t
e
r
a
t
i
ve
m
a
nn
e
r
[
36
]
.
F
or
e
a
c
h
i
t
e
r
a
ti
on
s
t
e
p
of
gr
a
di
e
nt
bo
os
t
i
ng
,
t
he
r
e
s
i
du
a
l
w
i
l
l
be
ut
i
li
z
e
d
t
o
op
t
im
i
z
e
t
he
pr
e
ce
di
ng
f
or
e
c
a
s
t
e
r
t
ha
t
t
he
de
f
i
ne
d l
os
s
pa
r
am
e
t
e
r
c
an
be
op
t
im
i
ze
d.
F
ur
t
he
r
,
fo
r
e
n
ha
nc
i
ng
pe
r
f
o
r
m
a
nc
e
,
re
gu
l
a
r
i
z
a
ti
on
f
un
c
t
i
on
i
s
a
dd
e
d
t
o
l
os
s
f
un
c
t
i
on
t
o
ob
t
a
in
t
he
ob
je
c
t
i
ve
ou
t
c
om
e
i
n
X
G
B
o
os
t
.
M
ea
s
ur
i
ng
t
he
a
l
go
r
i
t
hm
pe
r
f
o
r
m
a
nc
e
c
an
be
ob
t
a
i
ne
d
a
s
f
ol
l
ow
s
(
)
=
(
)
+
(
)
(1)
T
he
va
r
i
a
bl
e
t
r
a
i
ne
d
f
r
om
i
np
ut
da
t
a
s
e
t
a
r
e
de
pi
c
t
e
d
a
s
,
i
s
t
he
t
r
a
i
ni
ng
l
os
s
pa
r
a
m
et
e
r
,
s
uc
h
a
s
l
og
i
s
t
i
c
or
s
qu
a
r
e
l
os
s
,
w
hi
c
h
m
e
a
s
ur
e
ho
w
g
oo
d
t
he
a
l
go
r
i
t
hm
f
i
t
s
f
or
pa
r
t
i
c
ul
a
r
t
r
a
i
ni
ng
da
t
a
s
e
t
,
a
nd
i
s
t
he
r
e
gu
l
a
r
i
z
a
t
i
on
pa
r
a
m
et
e
r
,
t
ha
t
m
e
a
s
ur
e
t
he
c
om
pl
e
xi
t
y
of
t
he
a
l
go
r
i
t
hm
.
N
a
i
ve
r
m
od
e
l
i
s
e
xp
e
c
t
e
d
t
o
ha
ve
be
t
t
e
r
ou
t
c
om
e
a
ga
i
ns
t
ov
e
r
f
i
t
t
i
ng
a
s
t
he
ba
s
e
m
od
e
l
i
s
de
c
is
i
on
t
r
e
e
.
T
he
ou
t
c
om
e
of
a
l
go
r
i
t
hm
̅
i
s
a
ve
r
a
ge
d
or
vo
t
e
d
c
ol
l
e
c
t
i
on
o
f
of
t
r
e
e
s
,
w
hi
c
h
c
a
n
be
e
xp
r
e
s
s
e
d
a
s
f
o
l
l
ow
s
̅
=
∑
(
)
,
∈
.
=
1
(2)
O
b
je
c
t
i
ve
pa
r
a
m
e
te
r
a
t
ℎ
i
ns
t
a
nc
e
i
t
e
r
a
ti
on
c
a
n
be
e
s
t
a
bl
i
s
he
d
i
nt
o
(
)
=
∑
(
,
̅
)
+
∑
(
)
=
1
,
=
1
(3)
W
he
r
e
i
s
t
he
a
m
ou
nt
of
f
o
r
e
c
a
s
t
i
ng
.
I
n
t
hi
s
w
or
k,
t
he
̅
(
)
c
a
n
be
e
xp
r
e
s
s
e
d
a
s
f
ol
l
ow
s
̅
(
)
=
∑
(
)
=
=
1
̅
(
−
1
)
+
(
)
.
(4)
A
s
de
s
c
r
i
be
d
i
n
[
35
]
,
t
he
r
e
g
u
l
a
r
i
z
a
ti
on
pa
r
a
m
e
te
r
(
)
f
or
a
ny
D
T
i
s
de
s
c
r
i
be
d
a
s
f
ol
l
ow
s
(
)
=
+
1
2
∑
2
,
=
1
(5)
W
he
r
e
de
pi
c
t
s
c
om
pl
e
xi
ty
of
e
a
c
h
a
nd
e
ve
r
y
l
e
a
f
,
i
s
a
va
r
i
a
bl
e
t
o
i
nc
r
e
a
se
pe
na
l
t
y
,
a
nd
i
s
t
he
am
ou
nt
of
l
e
a
v
e
s
i
n
a
D
e
c
i
s
i
on
T
r
e
e
(
D
T
)
,
a
nd
i
s
t
he
ve
c
to
r
o
f
s
c
or
e
s
on
e
a
c
h
l
e
a
f
.
T
he
n,
i
t
us
e
s
s
e
c
on
d
-
or
de
r
T
a
y
l
or
e
xp
a
ns
i
o
n,
r
a
t
h
e
r
t
ha
n
f
i
r
s
t
-
o
r
de
r
i
n
g
r
a
di
e
nt
bo
os
t
i
ng
,
w
hi
c
h
i
s
us
e
d
a
s
a
l
os
s
pa
r
a
m
et
e
r
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
8
7
2
-
3
8
7
9
3876
X
G
B
o
os
t
.
F
ur
t
he
r
,
i
t
i
s
c
on
s
i
de
r
e
d
t
ha
t
t
he
lo
s
s
pa
r
a
m
et
e
r
is
m
e
a
n
s
qu
a
r
e
e
r
r
or
,
t
he
ob
je
c
t
i
ve
s
t
r
at
e
gy
c
an
be
e
xp
r
e
s
s
e
d
a
s
f
o
l
l
ow
s
(
)
≈
∑
[
ℎ
(
)
+
(
)
2
]
+
+
=
1
1
2
∑
2
,
=
1
(6)
B
y
r
em
ov
i
ng
t
he
c
o
ns
t
a
nt
t
he
(
.
)
i
s
a
f
un
c
t
i
on
t
ha
t
a
s
s
i
gn
da
t
a
po
i
nt
w
i
t
h
r
e
s
pe
c
t
t
o
l
e
a
f
,
ℎ
i
s
t
he
f
i
r
s
t
de
r
i
va
t
i
ve
of
m
ea
n
s
qu
a
r
e
e
r
r
or
l
os
s
f
u
nc
t
i
on
a
nd
i
s
t
he
s
e
c
on
d
de
r
i
va
t
i
ve
of
m
e
an
s
qu
a
r
e
e
r
r
o
r
l
os
s
f
un
c
t
i
on
.
I
n
a
b
ov
e
e
q
ua
t
i
on
(
6)
,
t
he
l
os
s
f
u
nc
t
i
on
i
s
c
om
pu
t
e
d
by
s
um
m
i
ng
t
he
l
os
s
of
e
a
c
h
da
t
a
f
e
a
tu
r
e
s
.
T
hi
s
i
s
do
ne
a
s
e
a
c
h
da
t
a
f
e
a
t
ur
e
c
o
r
r
e
s
p
on
d
s
t
o
on
l
y
o
ne
l
e
a
f
no
de
a
t
a
t
im
e
.
T
he
r
e
f
or
e
,
t
he
l
os
s
f
u
nc
t
i
on
i
s
c
om
pu
t
e
d
a
s
a
s
um
of
l
os
s
of
e
a
c
h
a
nd
e
ve
r
y
l
e
a
f
no
de
a
s
f
o
l
l
ow
s
(
)
≈
∑
[
(
∑
ℎ
∈
)
+
1
2
(
∑
+
∈
)
2
]
+
.
=
1
(7)
A
c
c
or
di
ng
t
o
a
bo
ve
e
qu
a
t
i
on
(
7)
,
c
a
n
be
c
o
m
pu
t
e
d
a
s
f
ol
lo
w
s
=
∑
ℎ
∈
,
(8)
S
im
i
l
a
r
ly
us
i
ng
E
q.
(
7)
,
c
a
n
b
e
c
om
pu
t
e
d
a
s
f
ol
l
ow
s
=
∑
∈
(9)
W
he
r
e
de
pi
c
ts
e
nt
i
r
e
da
t
a
fe
a
t
ur
e
s
e
t
s
i
n
l
e
a
f
no
de
.
T
he
r
e
f
or
e
,
t
he
o
pt
im
i
z
a
ti
on
of
o
bj
e
c
t
i
ve
s
t
r
a
t
e
gy
ca
n
be
op
t
im
i
z
e
d
i
nt
o
a
pr
ob
l
e
m
of
e
s
t
a
bl
i
s
hi
ng
m
i
ni
m
um
of
a
qu
a
dr
a
t
i
c
f
un
c
t
i
on
.
I
n
ot
he
r
w
a
y
,
po
s
t
c
om
pl
e
ti
on
s
pl
i
t
t
i
ng
of
c
e
r
t
a
i
n
no
de
i
n
D
T
,
t
he
o
pt
im
iz
a
t
i
on
of
a
l
go
r
i
t
hm
ou
t
c
om
e
ca
n
be
c
om
pu
t
e
d
us
i
n
g
ob
je
c
t
i
ve
s
t
r
a
te
gy
.
I
f
t
he
D
T
a
l
go
r
i
t
hm
ou
t
c
om
e
i
s
im
pr
ov
e
d
p
os
t
c
o
m
pl
et
i
on
of
s
pl
i
t
ti
ng
t
h
e
no
de
,
t
hi
s
op
t
im
i
z
at
io
n
w
i
l
l
be
us
ed
.
O
r
e
l
s
e
,
t
he
s
pl
i
t
ti
ng
pr
oc
e
s
s
w
i
l
l
be
t
e
rm
in
a
t
e
d.
A
l
on
g
w
i
t
h,
w
he
n
pe
r
f
or
m
i
ng
op
t
im
i
z
at
i
on
,
th
e
ob
je
c
t
i
ve
s
t
r
a
t
e
gy
,
a
f
or
e
c
a
s
t
i
ng
c
l
a
s
s
i
f
ic
a
ti
on
m
od
e
l
c
a
n
be
t
r
a
i
ne
d
a
ga
i
ns
t
ov
e
r
f
i
t
t
i
ng
du
e
t
o
r
e
gu
l
a
r
i
z
a
t
i
on
.
T
he
pr
op
os
e
d
c
l
a
s
s
i
f
i
e
r
at
t
ai
n
a
s
i
gn
i
f
i
c
a
nt
c
l
a
s
s
i
f
i
ca
ti
on
pe
r
f
or
m
a
nc
e
w
he
n
c
om
pa
r
e
d
w
i
t
h
s
t
a
t
e
-
of
-
a
r
t
m
od
e
l
w
hi
c
h
i
s
e
x
p
e
r
im
e
nt
a
ll
y
pr
ov
e
d
i
n
ne
xt
s
e
c
t
i
on
be
l
ow
.
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
S
T
hi
s
s
e
c
t
i
on
e
va
l
ua
t
e
s
pe
r
f
or
m
a
nc
e
e
va
l
ua
ti
on
o
f
pr
op
os
e
d
s
t
ud
e
nt
r
i
s
k
i
de
nt
i
f
i
c
a
ti
on
l
e
ar
ni
ng
m
od
e
l
ov
e
r
s
t
a
t
e
-
of
-
a
r
t
m
od
e
l
s
[
29
,
33
,
34
]
.
F
or
e
x
p
e
r
im
e
nt
a
na
l
y
sis
va
r
i
ou
s
e
x
pe
r
im
e
nt
a
r
e
c
on
s
id
e
r
e
d
a
n
d
da
t
a
us
e
d
f
or
e
x
pe
r
i
m
e
nt
s
a
r
e
pu
bl
i
c
a
l
ly
a
va
i
la
bl
e
.
Th
e
e
xp
e
r
i
m
e
nt
i
s
c
on
du
c
t
e
d
u
s
i
ng
w
i
nd
ow
s
10
o
pe
r
a
t
i
ng
s
y
s
t
em
,
I
nt
e
l
I
-
5
c
l
a
s
s
64
bi
t
pr
oc
e
s
s
o
r
,
16
G
B
R
A
M
,
4G
B
N
vi
di
a
C
U
D
A
e
na
bl
e
d
G
P
U
.
F
o
r
e
x
pe
r
im
e
nt
a
na
ly
s
i
s
t
hi
s
w
or
k
us
e
d
p
ub
l
i
cl
y
a
va
i
l
a
bl
e
da
t
a
s
e
t
ob
t
a
i
ne
d
f
r
om
O
U
L
A
D
[
33
,
34
]
w
h
i
c
h
c
om
po
s
e
d
of
di
f
f
e
r
e
nt
c
ou
r
s
e
s
w
i
t
h
s
t
ud
e
nt
e
n
r
ol
lm
e
nt
a
r
ou
n
d
1
20
0
t
o
25
00
.
T
he
ob
je
c
t
i
ve
of
t
hi
s
w
or
k
i
s
t
o
f
o
r
e
c
a
s
t
t
he
s
ub
m
i
s
si
on
of
f
i
r
s
t
a
s
s
e
s
sm
e
nt
of
a
pa
r
t
i
c
ul
a
r
c
ou
r
s
e
w
i
t
hi
n
de
a
dl
i
ne
t
im
e
a
r
ou
nd
2
0
t
o
3
0
da
y
s
.
T
he
c
ou
r
s
e
i
s
c
om
po
s
e
d
of
w
i
de
va
r
i
e
t
y
of
f
i
e
l
ds
s
uc
h
a
s
m
a
t
hs,
hi
s
t
or
y
,
e
ng
i
ne
e
r
i
ng
a
n
d
s
o
on
.
F
o
r
c
om
pl
e
t
i
ng
t
he
c
ou
r
s
e
,
t
he
s
t
ud
e
nt
ha
s
t
o
a
t
t
ai
n s
om
e
m
in
im
um
s
c
or
e
s
f
or
a
gi
ve
n t
a
s
k
or
a
s
s
e
s
s
m
e
nt
a
nd
t
he
n
pa
s
s
t
he
fi
na
l
e
xa
m
.
T
he
p
r
o
po
s
e
d s
t
u
de
nt
r
i
s
k
l
e
a
r
ni
ng
m
od
e
l
f
or
f
o
r
e
c
a
s
t
i
ng
dr
op
ou
t
h
a
s
be
e
n
a
im
e
d
a
t
a
t
t
ai
ni
ng
f
ol
l
ow
i
n
g
ob
je
c
t
i
ve
s
,
F
i
r
s
t
l
y
,
c
a
r
r
y
ou
t
a
na
l
y
s
i
s
da
i
ly
us
i
ng
M
L
a
l
g
or
i
t
hm
t
o
e
va
lu
a
t
e
c
l
a
s
s
i
f
i
ca
ti
on
m
od
e
l
.
S
e
c
on
dl
y
,
a
na
l
y
s
e
s
a
nd
i
de
nt
i
f
i
e
s
t
he
e
f
f
e
c
t
s
a
nd
p
r
o
bl
em
s
of
im
ba
la
nc
e
d
da
t
a
.
T
h
e
n,
c
om
pa
r
e
ou
r
p
r
o
po
s
e
d
m
od
e
l
o
ve
r
s
t
a
t
e
-
of
-
a
r
t
m
od
e
l
tr
a
i
ne
d
us
i
ng
l
e
ga
c
y
d
a
t
a
.
F
ou
r
t
hl
y
,
e
xp
e
r
i
m
e
nt
i
s
c
on
d
uc
t
e
d
f
or
di
f
f
e
r
e
nt
a
nd
c
ou
r
s
e
s
a
nd
e
v
a
l
ua
t
e
pe
r
f
or
m
a
nc
e
a
t
t
ai
ne
d
by
pr
op
os
e
d
m
od
e
l
ov
e
r
e
xi
s
t
i
ng
m
od
e
l
i
n
t
e
r
m
s
of
pr
e
c
i
s
i
on
,
r
e
c
a
l
l
,
F
-
m
e
a
s
ur
e
,
a
nd
R
O
C
.
E
xp
e
r
i
m
e
nt
a
r
e
c
on
d
uc
t
e
d
t
o
e
va
l
ua
t
e
R
O
C
pe
r
f
or
m
a
nc
e
at
t
a
i
ne
d
by
pr
op
os
e
d
X
G
B
oo
s
t
ov
e
r
e
xi
t
i
ng
S
up
po
r
t
V
e
c
t
or
M
a
c
hi
ne
(
S
V
M
)
a
s
s
ho
w
n
i
n
F
i
g.
1.
T
he
ou
t
c
o
m
e
s
ho
w
s
X
G
B
oo
s
t
a
t
t
a
i
n
a
n
R
O
C
pe
r
f
or
m
a
nc
e
im
pr
ov
e
m
e
nt
o
f
35
.
33
%
ov
e
r
S
V
M
.
F
ur
t
he
r
,
F
i
gu
r
e
2
s
ho
w
s
F
-
m
e
a
s
ur
e
a
t
t
ai
ne
d
by
bo
t
h
pr
op
os
e
d
a
n
d
e
xi
s
t
i
ng
m
od
e
l
.
T
he
ov
e
r
a
l
l
r
e
s
ul
t
a
t
ta
i
ne
d
s
h
ow
s
p
r
op
os
e
d
l
e
a
r
ni
ng
m
od
e
l
im
pr
ov
e
s
F
-
m
e
a
s
ur
e
s
c
or
e
by
24
.
45
%
,
26
.
65
%
,
a
n
d
18
.
96
%
o
ve
r
e
xi
s
t
i
ng
l
e
a
r
ni
ng
m
od
e
l
.
A
n
a
ve
r
a
ge
i
m
pr
ov
e
m
e
nt
of
2
3.
3
5%
i
s
a
t
t
a
in
e
d
by
pr
o
po
s
e
d
l
e
a
r
ni
ng
m
od
e
l
ov
e
r
e
xi
s
t
i
ng
m
od
e
l
.
3.1.
ROC per
f
oramcne e
va
lu
at
i
on
T
hi
s
s
e
c
t
i
on
e
v
a
l
ua
t
e
d
R
O
C
p
e
r
f
or
m
a
nc
e
a
t
ta
i
ne
d
by
P
r
op
o
s
e
d
X
G
B
oo
s
t
(
P
X
G
B
)
m
od
e
l
ov
e
r
e
xi
t
i
ng
c
l
a
s
s
i
f
ic
a
ti
on
m
od
e
l
.
E
xp
e
r
im
e
nt
s
a
r
e
c
ond
uc
t
e
d
c
on
s
i
de
r
i
ng
di
f
f
e
r
e
nt
de
a
dl
i
ne
da
y
s
.
E
xp
e
r
im
en
t
s
a
r
e
c
on
d
uc
t
e
d
f
or
di
f
f
e
r
e
nt
c
o
ur
s
e
[
27
,
28
]
a
n
d
R
O
C
pe
r
f
o
r
m
a
nc
e
i
s
a
ve
r
a
g
e
d
a
nd
r
e
s
ul
t
i
s
no
t
e
d
a
s
s
h
o
w
n
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Student
risk
id
entif
ic
ation
le
arnin
g mode
l
usi
ng
m
ach
i
ne
le
ar
ni
ng
approac
h
(
Sus
heela
m
ma K H
)
3877
F
i
gu
r
e
1.
A
n
a
ve
r
a
ge
R
O
C
pe
r
f
or
m
a
nc
e
im
pr
o
ve
m
e
nt
of
35
.
96
9%
i
s
a
tt
a
i
ne
d
by
P
X
G
B
ov
e
r
e
xi
t
i
ng
m
od
e
l
c
on
s
i
de
r
i
n
g
va
r
i
ou
s
de
a
dl
i
ne
da
t
e
s
c
e
na
r
i
os
.
F
r
om
f
i
gu
r
e
i
t
c
a
n
be
s
e
e
n
a
s
de
a
dl
i
ne
da
y
s
i
nc
r
e
a
s
e
s
t
he
PXG
B
m
od
e
l
at
t
ai
n
hi
gh
e
r
a
c
c
ur
a
c
y
i
n
i
de
nt
i
f
y
i
ng
r
i
s
k.
H
ow
e
v
e
r
,
t
he
a
c
c
ur
a
c
y
o
f
e
xi
s
t
i
ng
m
od
e
l
de
gr
a
de
s
r
a
pi
dl
y
.
T
he
ov
e
r
a
l
l
r
e
s
ul
t
a
tt
ai
ne
d
th
e
e
f
f
i
c
i
e
nc
y
of
P
X
G
B
m
ode
l
c
on
s
i
de
r
i
n
g
f
or
e
c
a
s
t
i
ng
f
or
di
f
f
e
r
e
nt
d
e
a
dl
i
ne
da
y
s
s
c
e
na
r
i
os
.
3.2.
F
-
me
as
ure
pe
rfo
r
amcne
evalua
tio
n
T
hi
s
s
e
c
t
i
on
e
va
l
ua
t
e
d
F
-
m
ea
s
ur
e
pe
r
f
o
r
m
an
c
e
a
t
t
a
i
ne
d
by
P
r
op
os
e
d
X
G
B
oo
s
t
(
P
X
G
B
)
m
od
e
l
ov
e
r
e
xi
t
i
ng
c
l
a
s
si
f
ic
a
t
i
on
m
od
e
l
.
E
xp
e
r
im
e
nt
s
a
r
e
c
on
du
c
t
e
d
c
on
s
i
de
r
i
ng
t
op
K
f
o
r
e
c
a
s
t
in
g.
E
x
pe
r
i
m
e
nt
s
a
r
e
c
on
d
uc
t
e
d
f
or
di
f
f
e
r
e
nt
c
o
ur
s
e
s
[
27
,
2
8]
a
nd
F
-
m
e
a
s
ur
e
pe
r
f
or
m
a
nc
e
i
s
a
ve
r
a
ge
d
a
nd
r
e
s
ul
t
i
s
no
t
e
d
a
s
s
ho
w
n
i
n
F
i
gu
r
e
2.
T
he
o
ut
c
om
e
a
tt
a
i
ne
d
s
ho
w
s
P
X
G
B
a
t
t
a
i
n
F
-
m
ea
s
ur
e
pe
r
f
or
m
a
nc
e
im
pro
ve
m
e
nt
of
2
4
.
45
%
,
26
.
65
%
,
a
nd
1
8.
9
6%
.
A
n
a
ve
r
a
ge
F
-
m
e
a
s
ur
e
pe
r
f
o
r
m
a
nc
e
im
pr
ov
e
m
e
nt
of
23
.
35
%
i
s
a
t
t
a
i
ne
d
by
P
X
G
B
ov
e
r
e
xi
t
i
ng
m
od
e
l
c
on
s
i
de
r
i
n
g
va
r
i
ou
s
t
op
s
c
e
na
r
i
os
.
3.3.
Preci
si
on
pe
rf
oramcne e
valuat
i
on
T
hi
s
s
e
c
t
i
on
ev
a
l
ua
t
e
d
pr
e
c
i
s
i
on
pe
r
f
or
m
a
nc
e
a
tt
a
i
ne
d
by
P
r
op
os
e
d
X
G
B
oo
s
t
(
P
X
G
B
)
m
od
e
l
ov
e
r
e
xi
t
i
ng
c
l
a
s
si
f
ic
a
t
i
on
m
od
e
l
.
E
xp
e
r
im
e
nt
s
a
r
e
c
on
du
c
t
e
d
c
on
s
i
de
r
i
ng
t
op
K
f
o
r
e
c
a
s
t
in
g.
E
x
pe
r
i
m
e
nt
s
a
r
e
c
on
d
uc
t
e
d
f
or
di
f
f
e
r
e
nt
c
o
ur
s
e
s
[
2
7
,
2
8]
a
nd
pr
e
c
i
s
i
on
pe
r
f
or
m
a
nc
e
i
s
a
ve
r
a
ge
d
a
nd
r
e
s
ul
t
i
s
no
t
e
d
a
s
s
h
ow
n
i
n
F
i
gu
r
e
3.
T
he
ou
t
c
om
e
at
t
ai
ne
d
s
ho
w
s
P
X
G
B
a
tt
a
i
n
pr
e
c
i
s
io
n
pe
r
f
o
r
m
a
nc
e
im
pr
ov
em
e
nt
of
34
.
6
6%
,
4
3
.
57
%
,
a
nd
2
3.
9
6%
.
A
n
a
ve
r
a
ge
pr
e
c
i
s
i
on
pe
r
f
or
m
a
nc
e
im
pr
ov
em
e
nt
of
34
.
0
6%
i
s
a
tt
a
i
ne
d
by
P
X
G
B
ov
e
r
e
xi
t
i
ng
m
od
e
l
c
on
s
i
de
r
i
ng
va
r
i
o
us
t
o
p
s
c
e
na
r
i
os
.
3.4.
Reca
ll
per
f
oramcne e
va
lu
at
i
on
T
hi
s
s
e
c
t
i
on
e
v
a
l
ua
t
e
d r
e
c
a
l
l
pe
r
f
or
m
a
nc
e
a
t
ta
i
ne
d b
y
P
r
o
po
s
e
d X
G
B
oo
s
t
(
P
X
G
B
)
m
od
e
l
ov
e
r
e
xi
t
i
ng
c
l
a
s
s
i
f
ic
a
ti
on
m
od
e
l
.
E
xp
e
r
im
e
nt
s
a
r
e
c
on
du
c
t
e
d
c
on
s
i
de
r
i
ng
t
op
K
f
or
e
c
a
s
t
i
ng
.
E
xp
e
r
i
m
e
nt
s
a
r
e
c
on
du
c
t
e
d
f
o
r
di
f
f
e
r
e
nt
c
o
ur
s
e
s
[
27
,
2
8]
a
n
d
r
e
c
a
l
l
pe
r
f
or
m
a
nc
e
i
s
a
ve
r
a
ge
d
a
n
d
r
e
s
ul
t
i
s
no
t
e
d
a
s
s
ho
w
n
i
n
F
i
gu
r
e
4
.
T
he
ou
t
c
om
e
a
tt
a
i
ne
d
s
ho
w
s
P
X
G
B
a
t
ta
i
n
r
e
c
al
l
pe
r
f
or
m
a
nc
e
im
pr
ov
em
e
nt
of
10
.
0
8%
,
4.
97
%
,
a
n
d
7.
0
2
%
.
A
n
a
ve
r
a
ge
F
-
m
e
as
ur
e
pe
r
f
or
m
a
nc
e
im
pr
ov
e
m
en
t
of
7.
35
%
i
s
a
t
t
ai
ne
d
by
P
X
G
B
ov
e
r
e
xi
t
i
ng
m
od
e
l
c
on
s
i
de
r
i
n
g
va
r
i
ou
s
t
o
p
s
c
e
na
r
i
os
.
F
i
gu
r
e
1.
R
O
C
pe
r
f
or
m
a
nc
e
f
or
va
r
i
e
d
nu
m
be
r
of
de
a
dl
i
ne
da
y
s
Figure
1
.
F
-
m
easur
e
p
e
rfo
rm
a
nce
for varie
d
To
p K
Figure
2
.
Pr
eci
sion pe
rfor
m
ance for
va
ried T
op K
Figure
3
.
Reca
ll
perform
ance f
or
var
ie
d
T
op
K
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
8
7
2
-
3
8
7
9
3878
4.
C
O
N
C
L
U
S
I
O
N
This
m
anu
script
introduc
ed
a
novel
desig
n
f
or
early
fin
ding
of
stu
de
nt
w
ho
are
at
risk
of
fail
in
g
or
com
pleti
ng
the
course
on
ti
m
e
without
us
i
ng
le
gacy
data.
The
pr
opos
e
d
m
od
el
us
es
the
sign
ific
a
nce
f
act
or
of
first
ta
sk
bein
g
i
m
po
rta
nt
factor
in
the
pr
ogress
of
co
urse
work.
The
best
way
is
to
extract
the
stud
e
nt
beh
a
viou
r
who
al
read
y
s
ubm
i
tt
ed
their
ta
s
k
and
le
ar
n
it
s
pa
tt
ern
.
This
w
ork
def
i
nes
t
he
pro
blem
as
a
bin
a
ry
cl
assifi
cat
ion
t
ask wit
h object
ive to lea
rn
a
nd
f
oreca
st dail
y usin
g forecas
ti
ng
window.
T
he pr
opos
e
d
m
od
el
is
evaluate
d
us
in
g
pu
bl
i
c
l
y
a
va
i
l
a
bl
e
O
U
L
A
D
da
t
a
s
e
t
.
T
h
e
ou
t
c
om
e
s
ho
w
s
t
he
pr
op
os
e
d
m
od
e
l
p
r
e
di
c
t
s
a
c
c
ur
a
t
e
l
y
e
ve
n
f
o
r
e
a
r
l
y
da
y
(
i
.
e
.
f
or
0
a
nd
1
da
y
s
)
,
a
l
s
o
pr
e
di
c
t
s
e
f
f
i
c
i
e
nt
l
y
f
or
l
a
te
r
da
y
s
o
f
c
o
ur
s
e
c
om
pl
e
t
i
on
,
a
nd
a
t
t
ai
n
be
t
t
e
r
ou
t
c
om
e
t
ha
n
tr
a
i
ni
ng
us
i
ng
l
e
ga
c
y
da
t
a
.
F
r
om
ov
e
r
a
l
l
e
xp
e
r
im
e
nt
a
na
l
y
si
s,
i
t
c
a
n
be
s
e
e
n
f
e
a
t
ur
e
s
e
l
e
c
t
i
on
V
L
E
i
s
im
po
r
t
a
nt
f
or
f
or
e
c
a
s
t
i
ng
s
t
ud
e
nt
a
t
r
i
s
k
of
f
a
i
l
i
ng
.
T
he
pr
op
os
e
d
X
G
B
oo
s
t
ba
s
e
d
c
l
a
s
s
i
f
i
ca
t
i
on
m
od
e
l
a
t
ta
i
ns
go
od
r
e
c
a
l
l
,
pr
e
c
i
s
i
on
,
RO
C
a
nd
F
-
m
eas
ur
e
pe
r
f
or
m
a
nc
e
.
A
n
a
ve
r
a
ge
R
O
C
pe
r
f
or
m
a
nc
e
im
pr
ov
em
e
nt
of
35
.
97
%
i
s
a
t
t
ai
ne
d
by
P
X
G
B
ov
e
r
e
xi
s
t
i
ng
m
od
e
l
.
F
ur
t
he
r
,
a
n
a
ve
r
a
ge
F
-
m
e
a
s
ur
e
pe
r
f
or
m
a
nc
e
im
pr
ov
em
e
nt
of
2
3.
35
%
i
s
a
t
t
a
i
ne
d
by
P
X
G
B
m
od
e
l
ov
e
r
e
xi
s
t
i
ng
m
ode
l
.
T
he
n,
a
n
av
e
r
a
g
e
pr
e
c
i
s
i
on
a
nd
r
e
c
a
l
l
pe
r
f
or
m
a
nc
e
im
pr
ov
e
m
e
nt
of
34
.
06
%
,
a
nd
7.
3
5%
i
s
at
t
ai
ne
d
by
P
X
G
B
m
od
e
l
ov
e
r
e
xi
s
t
i
ng
m
od
e
l
,
r
e
s
pe
c
t
i
ve
l
y
.
T
he
ov
e
r
a
l
l
r
e
s
ul
t
a
tt
a
i
ne
d
s
ho
w
s
t
he
pr
o
po
s
e
d
m
od
e
l
e
f
f
i
c
i
en
c
y
of
P
X
G
B
m
od
e
l
c
on
s
i
de
r
i
n
g
f
or
e
c
a
s
t
i
ng
f
o
r
di
f
f
e
r
e
nt
de
a
dl
i
ne
da
y
s
a
nd
T
o
p
s
c
e
na
r
i
os
.
T
h
e
f
ut
ur
e
w
or
k
w
e
w
ou
l
d
c
o
n
s
i
de
r
e
xp
e
r
i
m
e
nt
a
na
l
y
s
i
s
c
on
s
i
de
r
in
g
di
f
f
e
r
e
nt
da
t
a
s
e
t
a
nd
a
l
s
o
c
on
s
i
de
r
e
nh
a
nc
i
ng
f
or
e
c
a
s
t
i
ng
m
od
e
l
.
REFERE
NCE
S
[1]
Fetty
Fitr
i
y
an
ti
L
ubis,
Yus
epRos
m
ansy
ah,
Suhono
H.
Supangkat
,
“
Topi
c
discove
r
y
of
onli
ne
cour
se
rev
ie
ws
using
LDA wit
h
le
v
eraging
rev
i
ews he
l
pfulne
ss
”,
In
te
rn
ati
onal Journal of
Elec
tri
cal
and
Computer
Enginee
ring (
IJE
CE)
Vol.
9
,
No.
1,
pp
.
426
-
438
,
pp
.
42
6
-
438
,
2019
.
[2]
M.
I.
Olazaba
l
a
ga,
C.
C
.
Garri
do,
and
G.
U.
Ruiz
,
“
Resea
r
ch
on
MO
OC
s:
tre
nds
and
m
et
hodologi
es,
”
Porta
Linguarum
,
pp.
87
–
98,
2016
.
[3]
C.
Ye
and
G.
B
iswas,
“
Ea
rl
y
pr
edi
c
ti
on
of
student
dropout
and
per
form
anc
e
in
MO
OCs
using
h
ighe
r
gra
nul
ari
t
y
te
m
pora
l inform
at
ion
,
”
Journal
o
f
Learning
Anal
y
ti
cs
,
vol
.
1
,
no
.
3
,
pp
.
169
–
172
,
2
014.
[4]
J.
Mutahi,
A.
Kinai,
N.
Bor
e,
A.
Diri
y
e
,
and
K.
W
el
demari
a
m
,
“
Stud
y
ing
e
ngage
m
ent
and
per
form
anc
e
wit
h
le
arn
ing
technol
og
y
in
an
Afri
c
an
class
room
,
”
in
Proceedi
ngs
of
Sevent
h
In
te
r
nati
onal
Learni
ng
Anal
y
ti
cs
&
Knowle
dge
Conf
ere
nce
, p
p
.
148
–
152,
Van
couve
r
,
BC,
C
ana
d
a, Ma
rch
2017
.
[5]
Pete
r
J.
Quinn.
“
Drop
-
out
and
complet
ion
in
hig
her
educat
ion
in
eur
ope
among
student
s
from
under
-
rep
res
ent
e
d
groups.
”
Technic
al
report,
Europ
ean
Comm
ission
,
Oct
2013.
[6]
H.
Vos
senste
y
n
,
A.
Kottmann,
B.
Jongbloed
,
a
nd
F.
“
Kaise
r.
Drop
-
out
and
c
om
ple
ti
on
in
h
i
gher
educat
ion
i
n
eur
ope executi
v
e
sum
m
ary
”
.
T
echnic
a
l
rep
o
rt, E
u
rope
an
Com
m
ission,
2015
.
[7]
G.
Kena
,
J.
W
.
X.
R.
A.
Mus
u
-
Gill
ette
,
L
aur
en
a
nd
Robinson,
J.
Zha
ng,
S.
W
il
k
i
n
son
-
Flic
ker
,
A.
Barmer,
and
E
.
D.
V.
Vel
ez
.
“
The conditi
on
o
f
edu
c
at
ion
”
2015
.
T
echnic
a
l
Repor
t
20
15
-
144,
NCES,
Ma
y
2015.
[8]
O.
Sim
pson.
22
%
-
“
ca
n
we
do
bet
t
er
?
”
In The
CW
P Re
te
nti
on
Li
teratu
r
e
R
evi
e
w,
47,
2010.
[9]
K.
Jordan.
Mooc
complet
ion
ra
te
s:
The
da
ta.
ht
tp:
/
/www
.
katy
jo
rda
n.
com/MO
OCproject.
html
,
2015.
Acc
essed
:
2017
-
10
-
10.
[10]
D.
Kolle
r
,
A.
Ng,
C.
Do,
and
Z.
Ch
en.
“
Retenti
on
and
in
tention
in
m
assive
open
on
li
n
e
c
ourses:
In
dept
h
.
”
EDUCA
US
E,
htt
p://ww
w.e
duc
ause
.
edu
/e
ro
/a
rt
i
cl
e
/re
t
ent
ion
-
and
-
int
ention
-
m
assive
-
op
en
-
onl
ine
-
cour
ses
-
dept
h
-
0
,
Jun 2013.
[11]
S.
M.
Ja
y
apr
akas
h,
E.
W
.
Moody
,
E.
J.
M.
La
u
ri
a,
J.
R.
Reg
an,
a
nd
J.
D.
Baron.
“
Ea
rl
y
Al
ert
of
Aca
demicall
y
A
t
-
Risk Stude
nts:
An Open
Sourc
e Anal
y
tics Ini
t
ia
t
i
ve.
”
Journal
o
f
L
earning
Ana
ly
t
ics
,
vol
1
,
no
.
1
,
p
p.
6
-
47
,
2
014
.
[12]
A.
W
ol_,
Z
.
Z
dra
hal,
D.
Her
r
m
annova
,
J.
Ku
zi
l
ek,
and
M.
Hlosta.
“
Deve
lo
ping
pre
di
ct
iv
e
m
odel
s
for
ear
l
y
det
e
ct
ion
of
at
-
r
i
sk stude
nts on
di
stanc
e
l
ea
rning m
odule
s.
”
In
Mac
hine
Learning
and
Learning
Anal
yt
i
cs
workshop
at
LAK14
,
24
-
28
Marc
h
2014,
In
dia
n
apo
li
s, India
na,
US
A,
4
,
201
4.
[13]
H.
He
and
E. A.
Garc
ia.
“
Learni
n
g
from
imbala
nc
ed
data.
”
IE
EE
T
rans
.
on
Knowl.
and
Data
Eng
.,
vol
21
,
no
.
9
,
p
p
.
1263
-
1284,
Sep
2009.
[14]
M.
Kloft,
F.
Sti
ehl
er
,
Z
.
Zhe
ng
,
and
N.
Pinkwar
t.
“
Predicting
m
ooc
dropout
ove
r
wee
ks
using
m
ac
hine
le
a
rnin
g
m
et
hods.
”
In
Pr
oce
ed
ings o
f
the
EMNLP
2014
Workshop on
Anal
ysis
of
Lar
ge
Scal
e
Soc
ial
Int
erac
ti
on
in MO
OC
s
,
pp.
60
-
65
,
2014
.
[15]
C.
T
a
y
lor
,
K.
Vee
ramac
h
ane
ni
,
and
U.
O'
R
ei
l
l
y
.
“
Li
k
ely
to
s
top
?
pr
edi
c
ti
ng
stopout
in
m
assive
op
en
onl
ine
cour
ses
”
.
CoRR
,
abs/1408.
3382
,
2014.
[16]
Md.
Arm
anur
Rahman1,
J.
Hos
sen2,
Venka
ta
ses
hai
ah
C3
,
CK
Ho4,
Ta
n
Kim
Ge
ok5,
Aziza
Sultana6,
Jesm
ee
n
M.
Z.
H.7
,
Ferdous
Hos
sain8,
“
A
Surve
y
of
Ma
ch
ine
L
ea
rn
ing
Techni
ques
for
Se
l
f
-
tuni
ng
Hadoop
Perform
anc
e
,
”
Inte
rnational
Jo
urna
l
of El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
(
IJE
CE)
,
Vol.
8,
No.
3
,
pp
.
1854
-
1
862
,
2018
.
[17]
Hlosta
M.
et
al.,
2014.
“
Modelli
ng
student
on
li
n
e
beh
avi
our
in
a
virt
ual
l
ea
rning
envi
ronm
ent
.
In
dia
napo
li
s
,
”
LAK
2014.
[18]
Pande
y
M.
&
S
har
m
a
V.
K
.
,
“
A
Dec
ision
Tree
Algorit
hm
Pertaining
to
the
Stu
dent
Perform
ance.
Anal
y
sis
and
Predic
ti
on
.
”
In
te
r
nati
onal Journal
of
Comput
er
Ap
pli
cations
,
vo
l
61
,
no
.
13
,
pp
.
1
-
5
,
2013
.
[19]
Rom
ero
,
C.
,
Ló
pez
,
M.
,
Luna,
J.
&
Ventur
a
,
S.,
.
“
Predic
ti
ng
student
s'
fin
al
per
f
orm
anc
e
from
par
ticipat
ion
in
o
n
-
li
ne
discussion
f
orum
s.
”
Computers
&
Educ
aton
,
I
ss
ue
68,
pp
.
458
-
472
,
2013
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Student
risk
id
entif
ic
ation
le
arnin
g mode
l
usi
ng
m
ach
i
ne
le
ar
ni
ng
approac
h
(
Sus
heela
m
ma K H
)
3879
[20]
W
la
dis,
C
.
,
Hac
he
y
,
A
.
C
.
&
C
onwa
y
,
K.
,
2014
.
“
An
inve
st
iga
t
i
on
of
cou
rse
-
le
v
el
f
ac
to
rs
as
pr
e
dic
tors
of
onli
n
e
STEM
cour
se
ou
tc
om
es.
”
Comput
ers
&
Educ
ati
on
,
Iss
ue
77
,
pp
.
14
5
-
150.
[21]
W
olff,
A.,
Zdrahal
,
Z.,
Nikolov
,
A.
&
Pantu
ce
k
,
M.,
2013.
“
Im
proving
retent
io
n:
pre
di
ct
ing
at
-
risk
student
s
b
y
ana
l
y
sing
beh
aviour i
n
a
v
irt
ua
l l
ea
rning
environ
m
ent
.
”
s.l., LAK 2013.
[22]
W
olff,
A.,
Zdrahal
,
Z.,
Herrm
an
nova,
D.
&
Knot
h,
P.,
2013
.
“
Pred
ic
ti
ng
student
p
erf
orm
anc
e
from
combined
data
source
s.
In:
A.
P
eña
-
A
y
ala,
ed.
E
duca
t
iona
l
Data
Mining:
Appli
cations
and
Tre
nds
.
”
V
erlag
:
Spring
er
Int
ern
ational
Publishing,
pp
.
1
75
-
202.
[23]
W
olff,
A.
e
t
al
.
,
2014.
“
Deve
lop
ing
pre
di
ct
iv
e
m
odel
s
for
e
arly
d
et
e
ct
io
n
o
f
at
-
ris
k
student
s
on
di
stanc
e
learni
n
g
m
odule
s.
”
Ind
ianapoli
s,
LAK 20
14.
[24]
W
ang,
Rui
&
C
hen,
Fang
li
n
&
Chen,
Zhe
n
y
u
&
Li
,
Tianxing
&
Hara
ri,
Gabri
el
l
a
&
T
ignor,
St
efanie
&
Zhou,
Xia
&
Ben
-
Z
ee
v
,
Dr
or
&
T
.
C
ampbell,
Andrew.
“
Student
Life:
As
sess
ing
m
ent
al
he
al
th
,
a
ca
d
emic
p
erf
orm
anc
e
and
beha
vior
al
tre
n
ds
of
colleg
e
s
tude
nts
using
sm
art
phones.
”
Ub
iComp
2014
-
Proce
ed
ings
of
the
2014
ACM
Inte
rnational
Jo
i
nt
Conf
ere
nce o
n
Pe
rvasi
ve and U
biqui
tous Computi
ng
.
10.
1145
/
2632048.
263205
4,
2014
.
[25]
W
ang,
Rui
&
C
hen,
Fang
li
n
&
Chen,
Zhe
n
y
u
&
Li
,
Tianxing
&
Hara
ri,
Gabri
el
l
a
&
T
ignor,
St
efanie
&
Zhou,
Xia
&
Ben
-
Z
ee
v
,
Dror
&
T.
C
ampbell,
Andrew.
“
Student
Li
f
e:
Us
ing
Sm
art
phones
to
As
sess
Menta
l
He
al
th
an
d
Aca
demic
Perfor
m
anc
e
of
Coll
eg
e
Studen
ts.
”
7
-
3
3.
10
.
1007/978
-
3
-
319
-
5139
4
-
2_2,
2017
.
[26]
J.
He,
J.
B
aile
y
,
B.
I.
Rubinstei
n,
and
R
.
Zh
an
g.
“
Ide
nti
f
y
i
ng
at
-
risk
studen
ts
in
m
assive
open
onli
ne
cour
ses,
”
In
AAAI
,
1749
-
1
755,
2015
.
[27]
N.
Tha
i
-
Nghe,
A.
Busche,
and
L.
Schm
idt
-
Th
iem
e.
“
Im
proving
ac
ad
emic
per
fo
r
m
anc
e
pre
d
ic
t
io
n
b
y
d
ealing
wi
t
h
cl
ass
imbal
anc
e
.
”
In
Nin
th
In
te
r
nati
onal
Con
fe
r
enc
e
on
Int
el
l
ig
ent
S
yste
ms
Design
and
App
lica
ti
ons,
ISDA
200
9
,
Pis
a,
It
aly
,
Nove
m
ber
30
-
Dec
em
ber
2,
2009,
878
-
883,
2009
.
[28]
C.
Ye
and
G.
B
iswas.
“
Ea
rl
y
pr
edi
c
ti
on
of
stud
ent
dropout
and
per
form
anc
e
in
m
oocs
u
sing
hi
gher
gra
nul
ar
i
t
y
te
m
pora
l inform
at
ion
.
”
Journal
o
f
Learning
Anal
y
ti
cs
,
vol
.
1
,
no
.
3
,
pp
.
169
-
172
,
2
014.
[
2
9]
H
.
L
a
k
k
a
r
a
j
u
,
E
.
A
g
u
i
a
r
,
C
.
S
h
a
n
,
D
.
M
i
l
l
e
r
,
N
.
B
h
a
n
p
u
r
i
,
R
.
G
h
a
n
i
,
a
n
d
K
.
L
.
A
d
d
i
s
o
n
.
“
A
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
f
r
a
m
e
w
o
r
k
t
o
i
d
en
t
i
f
y
s
t
u
d
e
n
t
s
a
t
r
i
s
k
o
f
a
d
v
e
r
s
e
a
c
a
d
e
m
i
c
o
u
t
c
o
m
e
s
.
”
1909
-
1
9
1
8
,
2
0
1
5
.
[30]
E.
Aguiar
,
H.
L
a
kkar
aj
u
,
N.
Bhanpuri
,
D.
Mill
er
,
B.
Yuhas,
and
K.
L.
Addison.
“
W
ho,
when,
and
why
:
A
m
ac
hine
le
arn
ing
appr
oa
c
h
to
priori
tizing
student
s
at
risk
of
not
gra
duat
in
g
high
school
on
ti
m
e.
”
In
LAK
'15,
93
-
102,
New
York,
NY
,
US
A,
2015.
ACM
.
[31]
J.
Bai
nbridge,
J.
Meli
tski,
A.
Zahradni
k,
E
.
La
ur
ia
a,
S.
M.
Ja
y
a
pra
kash,
and
J.
Baron.
“
Us
ing
Le
arn
ing
Anal
y
t
ics
to
Predic
t
At
-
Risk
Student
s
in
Online
Grad
uat
e
Publi
c
Affai
rs
and
Adm
ini
strat
ion
Educat
ion
.
”
The
J
P
A
E
Me
ss
enge
r
,
vo
l.
21
,
no
.
2
,
pp
.
24
7
-
262,
2015
.
[32]
S.
Jiang,
M.
W
arsc
haue
r
,
A.
E.
W
il
li
ams
,
D.
OD
owd,
and
K.
S
che
nke
.
“
Predic
t
ing
m
ooc
per
form
anc
e
with
wee
k
1
beha
vior
.
”
In
E
DM14
,
273
-
275,
2014.
[
3
3]
K
u
z
i
l
e
k
,
M
.
H
lo
s
t
a
,
a
n
d
Z
.
Z
dr
a
h
a
l
.
“
O
p
e
n
u
ni
v
e
r
s
i
t
y
l
e
a
r
n
i
n
g
a
n
a
l
y
t
i
c
s
d
a
t
a
s
e
t
.
”
In
D
a
t
a
l
i
ter
a
c
y
f
o
r
L
e
a
rni
n
g
A
n
a
l
y
t
i
c
s
w
o
r
k
s
ho
p
a
t
L
A
K
1
6
,
2
6
t
h
A
p
r
i
l
2
0
1
6
,
E
d
i
n
b
u
r
g
h
,
U
K
,
9
,
2
0
1
6
.
[
3
4]
K
u
z
i
l
e
k
,
J
a
k
u
b
;
H
l
o
s
t
a
,
M
a
r
t
i
n
a
n
d
Z
d
r
á
h
a
l
,
Z
d
e
n
e
k
.
(
2
0
1
7
)
.
“
Op
e
n
U
n
i
v
e
r
s
i
t
y
L
e
a
r
n
i
n
g
A
n
a
l
y
t
i
c
s
d
a
t
a
s
e
t
.
”
S
c
i
e
n
t
i
f
i
c
D
a
t
a
.
4
.
1
7
0
1
7
1
.
1
0
.
1
0
3
8
/
s
d
a
t
a
.
2
0
1
7
.
1
7
1
.
[35]
T.
Q.
Ch
en,
C.
Guestrin,
“
XG
Boost:
a
s
ca
l
abl
e
tree
boosting
sy
stem”
,
Proc
eedings
of
th
e
22
nd
ACM
SIGK
DD
Inte
rnational
Co
nfe
renc
e
on
Kno
wle
dge
Discov
ery
and
Data
Min
i
ng
,
pp
.
785
-
794
,
2016.
[36]
J.
H.
Friedman
,
“
Gree
d
y
func
t
i
on
appr
oximat
io
n:
a
gr
adient
bo
osting
m
ac
hin
e,”
Annal
s
o
f
stat
isti
cs
,
pp
.
1189
-
1232,
2001
.
Evaluation Warning : The document was created with Spire.PDF for Python.