I
nte
rna
t
io
na
l J
o
urna
l o
f
I
nfo
rm
a
t
ics a
nd
Co
m
m
un
ica
t
io
n T
ec
hn
o
lo
g
y
(
I
J
-
I
CT
)
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
,
p
p
.
1
6
8
2
~
1
695
I
SS
N:
2252
-
8
7
7
6
,
DOI
:
1
0
.
1
1
5
9
1
/iji
ct
.
v15
i
4
.
pp
1
6
8
2
-
1
695
1682
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ict.
ia
esco
r
e.
co
m
Dev
elo
pment of
an AI
-
ba
sed
dig
ital
startup
incubatio
n mo
del
for the
empo
wer
ment of
stude
nts
a
nd MSM
Es
t
o
wa
rds a
creativ
e ec
o
no
my
Um
a
r
B
a
kti¹,
M
a
ria
Septij
a
ntini Alie²
,
I
i
n M
a
rlia
na
³
¹
D
e
p
a
r
t
me
n
t
o
f
En
t
r
e
p
r
e
n
e
u
r
sh
i
p
,
U
n
i
v
e
r
si
t
a
s
M
i
t
r
a
I
n
d
o
n
e
si
a
,
B
a
n
d
a
r
L
a
m
p
u
n
g
,
I
n
d
o
n
e
s
i
a
²D
e
p
a
r
t
me
n
t
o
f
M
a
n
a
g
e
me
n
t
,
U
n
i
v
e
r
s
i
t
a
s
M
i
t
r
a
I
n
d
o
n
e
si
a
,
B
a
n
d
a
r
L
a
m
p
u
n
g
,
I
n
d
o
n
e
s
i
a
³D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
o
n
Te
c
h
n
o
l
o
g
y
,
U
n
i
v
e
r
si
t
a
s
M
i
t
r
a
I
n
d
o
n
e
si
a
,
B
a
n
d
a
r
La
m
p
u
n
g
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Feb
26
,
2
0
2
6
R
ev
is
ed
Au
g
15
,
2
0
2
6
Acc
ep
ted
Au
g
28
,
2
0
2
6
M
icro
,
sm
a
ll
,
a
n
d
m
e
d
i
u
m
e
n
t
e
rp
rise
s
(M
S
M
Es)
c
o
n
ti
n
u
e
t
o
fa
c
e
lo
w
d
ig
it
a
l
-
tec
h
n
o
lo
g
y
a
d
o
p
t
io
n
,
wit
h
c
o
n
v
e
n
ti
o
n
a
l
i
n
c
u
b
a
ti
o
n
p
ro
g
ra
m
s
stil
l
re
ly
in
g
o
n
g
e
n
e
ric
train
i
n
g
a
n
d
m
a
n
u
a
l
a
ss
e
ss
m
e
n
t
ra
th
e
r
th
a
n
a
u
to
m
a
ted
,
e
v
id
e
n
c
e
-
b
a
se
d
e
v
a
lu
a
ti
o
n
.
Th
is
stu
d
y
d
e
v
e
l
o
p
s
a
n
d
v
a
li
d
a
tes
a
n
AI
-
b
a
se
d
sta
rtu
p
i
n
c
u
b
a
ti
o
n
m
o
d
e
l
(AI
-
S
I
M
),
a
n
i
n
teg
ra
ted
ICT
p
latfo
rm
f
u
n
c
ti
o
n
i
n
g
a
s
a
n
i
n
telli
g
e
n
t
d
e
c
isio
n
-
su
p
p
o
rt
s
y
ste
m
fo
r
M
S
M
E
d
i
g
it
a
l
-
re
a
d
in
e
ss
c
las
sifica
ti
o
n
,
a
d
a
p
ti
v
e
train
i
n
g
re
c
o
m
m
e
n
d
a
ti
o
n
,
g
e
o
s
p
a
ti
a
l
a
n
a
ly
sis,
a
n
d
stu
d
e
n
t
-
M
S
M
E
e
x
p
e
rien
ti
a
l
lea
rn
in
g
wit
h
in
o
n
e
o
p
e
ra
ti
o
n
a
l
e
n
v
iro
n
m
e
n
t.
Da
ta
we
r
e
c
o
ll
e
c
ted
fro
m
4
,
2
7
6
M
S
M
Es
a
c
ro
ss
Lam
p
u
n
g
P
ro
v
i
n
c
e
,
In
d
o
n
e
sia
;
a
fter
strict
two
-
le
v
e
l
d
e
d
u
p
li
c
a
ti
o
n
t
o
p
re
v
e
n
t
d
a
ta
lea
k
a
g
e
,
9
6
1
u
n
i
q
u
e
re
c
o
rd
s
we
re
re
tai
n
e
d
f
o
r
m
o
d
e
li
n
g
.
A
ra
n
d
o
m
fo
r
e
st
c
las
sifier
g
ro
u
p
e
d
M
S
M
Es
i
n
t
o
t
h
re
e
re
a
d
i
n
e
ss
c
a
teg
o
ries
(Tr
a
d
it
io
n
a
l,
S
tab
le,
Dig
i
tal
Re
a
d
y
)
u
si
n
g
two
fe
a
tu
re
c
o
n
fig
u
ra
ti
o
n
s,
a
c
h
iev
i
n
g
3
3
.
6
8
%
(
1
0
f
e
a
tu
re
s;
5
-
fo
ld
CV
3
6
.
7
5
%
)
a
n
d
3
0
.
5
7
%
a
c
c
u
ra
c
y
(7
m
a
n
u
sc
ri
p
t
-
a
li
g
n
e
d
fe
a
tu
re
s;
5
-
fo
ld
CV
3
7
.
2
3
%
),
b
o
th
n
e
a
r
th
e
3
3
.
3
%
ra
n
d
o
m
-
c
h
a
n
c
e
b
a
se
li
n
e
.
F
e
a
tu
re
-
imp
o
rtan
c
e
a
n
a
ly
sis
sh
o
we
d
p
r
e
d
ictiv
e
we
ig
h
t
sp
re
a
d
a
c
ro
ss
a
ll
se
v
e
n
fe
a
tu
re
s,
led
b
y
d
i
g
it
a
l
in
fra
stru
c
t
u
re
a
n
d
m
o
n
t
h
l
y
re
v
e
n
u
e
,
wit
h
n
o
d
o
m
in
a
n
t
p
re
d
icto
r
.
Th
e
sy
ste
m
wa
s
v
a
li
d
a
ted
th
r
o
u
g
h
fiel
d
d
e
p
lo
y
m
e
n
t,
tra
in
in
g
,
a
n
d
fo
c
u
s
g
ro
u
p
d
isc
u
ss
io
n
s
with
84
s
tu
d
e
n
t
a
n
d
M
S
M
E
p
a
rt
icip
a
n
ts,
c
o
n
firmi
n
g
it
s
p
ra
c
ti
c
a
l
v
a
lu
e
a
s
a
re
a
l
-
ti
m
e
in
c
u
b
a
ti
o
n
d
a
sh
b
o
a
rd
,
wh
il
e
th
e
c
las
sifica
ti
o
n
c
o
m
p
o
n
e
n
t
re
q
u
ires
rich
e
r
b
e
h
a
v
i
o
ra
l
a
n
d
tran
sa
c
ti
o
n
a
l
d
a
ta
t
o
imp
ro
v
e
a
c
c
u
ra
c
y
i
n
f
u
tu
re
it
e
ra
ti
o
n
s.
K
ey
w
o
r
d
s
:
AI
-
en
ab
led
d
ig
ital
in
cu
b
atio
n
Geo
s
p
atial
an
aly
tics
I
n
ter
ac
tiv
e
d
ec
is
io
n
-
s
u
p
p
o
r
t
d
ash
b
o
ar
d
M
ac
h
in
e
lear
n
in
g
class
if
icatio
n
MSM
E
d
ig
ital r
ea
d
in
ess
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Um
ar
B
ak
ti
Dep
ar
tm
en
t o
f
E
n
tr
ep
r
en
eu
r
s
h
ip
,
Un
iv
er
s
itas
Mitr
a
I
n
d
o
n
esia
B
an
d
ar
L
am
p
u
n
g
3
5
1
4
5
,
L
a
m
p
u
n
g
,
I
n
d
o
n
esia
E
m
ail:
u
m
ar
b
ak
ti@
u
m
itra
.
ac
.
i
d
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
r
ap
id
g
r
o
wth
o
f
d
i
g
ital
tech
n
o
lo
g
ies
h
as
r
esh
ap
ed
en
t
r
ep
r
en
e
u
r
ial
ec
o
s
y
s
tem
s
an
d
ac
ce
ler
ated
cr
ea
tiv
e
-
ec
o
n
o
m
y
d
e
v
elo
p
m
e
n
t
wo
r
ld
wid
e,
en
ab
lin
g
f
ir
m
s
to
lev
er
ag
e
d
ata
an
aly
tics
,
clo
u
d
co
m
p
u
tin
g
,
a
n
d
o
n
lin
e
p
latf
o
r
m
s
f
o
r
co
m
p
etitiv
en
ess
an
d
s
ca
lab
ilit
y
[
1
]
.
Mi
cr
o
,
s
m
all,
an
d
m
ed
i
u
m
e
n
ter
p
r
is
es
(
MSM
E
s
)
ar
e
v
ital
to
ec
o
n
o
m
ic
g
r
o
wth
an
d
em
p
lo
y
m
en
t,
y
et
m
an
y
f
a
ce
p
er
s
is
ten
t
b
ar
r
ier
s
to
d
ig
i
tal
tr
an
s
f
o
r
m
atio
n
,
in
clu
d
in
g
lim
ited
tech
n
o
lo
g
ic
al
ca
p
ab
ilit
y
,
o
r
g
an
izatio
n
al
r
ea
d
in
ess
,
d
ig
ital
s
k
ills
,
an
d
ac
ce
s
s
to
d
a
ta
-
dr
iv
en
d
ec
is
io
n
s
u
p
p
o
r
t
[
2
]
-
[
4
]
,
w
h
ich
co
n
s
tr
ain
th
eir
c
o
m
p
etitiv
en
ess
in
em
er
g
in
g
ec
o
n
o
m
ies
wh
er
e
tech
n
o
lo
g
ica
l
r
eso
u
r
ce
s
r
em
ain
u
n
ev
e
n
ly
d
is
tr
ib
u
ted
[
5
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Dev
elo
p
men
t o
f a
n
A
I
-
b
a
s
ed
d
ig
ita
l sta
r
tu
p
in
cu
b
a
tio
n
mo
d
e
l fo
r
th
e
emp
o
w
ermen
t o
f
… (
Uma
r
B
a
kti
)
1683
Fro
m
an
I
C
T
p
er
s
p
ec
tiv
e,
th
e
s
e
co
n
s
tr
ain
ts
r
ef
lect
s
y
s
tem
s
-
lev
el
g
ap
s
:
MSM
E
s
an
d
th
eir
in
cu
b
ato
r
s
ty
p
ically
lack
i
n
teg
r
ated
in
tellig
en
t
d
ec
is
io
n
-
s
u
p
p
o
r
t
in
f
r
astru
ctu
r
e
t
o
s
en
s
e
d
ig
ital
r
ea
d
in
ess
,
r
o
u
te
p
er
s
o
n
alize
d
i
n
ter
v
en
tio
n
s
,
a
n
d
v
is
u
alize
s
p
atial/b
eh
av
io
r
a
l
p
atter
n
s
in
r
ea
l
tim
e.
B
u
s
in
ess
in
cu
b
atio
n
is
a
r
ec
o
g
n
ized
s
tr
ateg
y
f
o
r
MSM
E
d
ev
elo
p
m
e
n
t
t
h
r
o
u
g
h
m
e
n
t
o
r
in
g
,
n
etwo
r
k
in
g
,
an
d
ca
p
ac
i
ty
b
u
ild
in
g
[
6
]
,
[
7
]
,
an
d
v
ir
tu
al/d
ig
ital
in
cu
b
atio
n
m
o
d
els
h
av
e
s
in
ce
ex
ten
d
ed
t
h
ese
s
er
v
ices
th
r
o
u
g
h
d
ig
ital
p
latf
o
r
m
s
[
8
]
,
[
9
]
.
AI
h
as
also
b
ec
o
m
e
an
im
p
o
r
tan
t
en
ab
ler
f
o
r
o
p
p
o
r
tu
n
ity
r
ec
o
g
n
itio
n
,
s
eg
m
en
tatio
n
,
p
r
e
d
ictiv
e
an
aly
tics
,
an
d
b
u
s
in
ess
d
ec
is
io
n
-
m
ak
in
g
[
1
0
]
-
[
1
2
]
.
Nev
er
th
eless
,
m
o
s
t in
cu
b
atio
n
p
r
o
g
r
am
s
s
till
r
ely
o
n
g
en
er
ic
tr
ain
in
g
an
d
m
an
u
al
ass
ess
m
en
t,
with
f
e
w
o
p
er
atin
g
as
u
n
if
ied
I
C
T
p
latf
o
r
m
s
co
u
p
lin
g
class
i
f
icatio
n
,
g
e
o
s
p
atial
in
tellig
en
ce
,
an
d
r
ec
o
m
m
e
n
d
a
tio
n
in
to
o
n
e
d
ec
is
io
n
-
s
u
p
p
o
r
t
lo
o
p
.
Prio
r
s
tu
d
ies
p
r
o
p
o
s
ed
d
ig
ital
-
ec
o
s
y
s
tem
f
r
am
ewo
r
k
s
f
o
r
MSM
E
tr
an
s
f
o
r
m
atio
n
[
1
3
]
a
n
d
ap
p
lied
AI
/d
ata
an
aly
tics
to
im
p
r
o
v
e
o
p
er
atio
n
al
an
d
m
ar
k
etin
g
p
er
f
o
r
m
an
ce
[
1
4
]
,
[
1
5
]
.
R
an
d
o
m
f
o
r
est
in
p
ar
ticu
lar
h
as
p
er
f
o
r
m
ed
well
in
class
if
y
in
g
d
i
g
ital
m
atu
r
ity
[
1
6
]
,
[
1
7
]
,
an
d
g
eo
s
p
atial
an
al
y
tics
h
as
id
en
tifie
d
s
p
atial
MSM
E
p
atter
n
s
[
1
8
]
,
[
1
9
]
.
Ho
wev
er
,
th
ese
s
tu
d
ies g
en
er
ally
tr
ea
t c
lass
if
icat
io
n
,
in
cu
b
atio
n
,
r
ec
o
m
m
en
d
atio
n
,
g
eo
s
p
atial
an
aly
tics
,
an
d
e
n
tr
ep
r
en
eu
r
s
h
ip
e
d
u
ca
ti
o
n
as
s
ep
ar
ate,
s
tan
d
alo
n
e
d
o
m
ai
n
s
.
Few
in
teg
r
ate
au
to
m
ated
r
ea
d
in
ess
ass
ess
m
e
n
t,
ad
ap
tiv
e
r
ec
o
m
m
en
d
atio
n
,
g
eo
s
p
atial
in
tellig
en
ce
,
d
ash
b
o
ar
d
s
,
an
d
s
tu
d
en
t
–
MSM
E
co
llab
o
r
atio
n
in
to
o
n
e
AI
-
d
r
iv
en
in
cu
b
atio
n
f
r
am
e
wo
r
k
th
at
is
also
tr
an
s
p
ar
en
t
ab
o
u
t
th
e
em
p
ir
i
ca
l
lim
its
o
f
its
p
r
ed
ictiv
e
co
m
p
o
n
en
ts
.
T
o
ad
d
r
ess
th
is
g
ap
,
th
is
s
t
u
d
y
p
r
o
p
o
s
es
an
AI
-
b
ased
s
tar
tu
p
in
cu
b
atio
n
m
o
d
el
(
AI
-
SIM
)
,
an
in
teg
r
ated
d
ig
ital
in
cu
b
atio
n
f
r
am
ewo
r
k
co
m
b
in
in
g
m
ac
h
in
e
lear
n
in
g
class
if
icatio
n
,
g
eo
s
p
atial
an
aly
s
is
,
ad
ap
tiv
e
r
ec
o
m
m
en
d
at
io
n
,
a
n
d
d
ash
b
o
a
r
d
v
is
u
aliza
tio
n
in
o
n
e
p
latf
o
r
m
.
AI
-
SIM
class
if
ies
MSM
E
s
b
y
d
ig
ital
r
ea
d
in
ess
u
s
in
g
r
a
n
d
o
m
f
o
r
est
[
1
6
]
,
[
2
0
]
,
g
e
n
er
ates
p
er
s
o
n
a
lized
in
cu
b
atio
n
/tra
in
in
g
r
ec
o
m
m
en
d
atio
n
s
[
2
1
]
,
[
1
7
]
,
a
n
d
s
u
p
p
o
r
ts
r
ea
l
-
tim
e
m
o
n
ito
r
in
g
v
ia
web
-
b
ased
d
ash
b
o
ar
d
s
[
2
2
]
,
wh
ile
also
lin
k
in
g
u
n
iv
er
s
ity
s
tu
d
en
ts
with
MSM
E
s
f
o
r
ex
p
er
ien
tial
lear
n
in
g
th
at
s
tr
en
g
th
e
n
s
r
eg
io
n
al
d
ig
ital
ec
o
s
y
s
tem
s
[
2
3
]
,
[
2
4
]
.
T
h
e
f
r
a
m
ewo
r
k
was
ev
alu
ated
u
s
in
g
4
,
2
7
6
r
a
w
MSM
E
r
ec
o
r
d
s
f
r
o
m
L
a
m
p
u
n
g
Pr
o
v
in
ce
,
I
n
d
o
n
esia;
a
r
ig
o
r
o
u
s
two
-
lev
el
d
ed
u
p
licatio
n
p
r
o
ce
d
u
r
e
to
p
r
ev
en
t
d
ata
leak
a
g
e
y
ield
ed
9
6
1
u
n
iq
u
e
an
aly
tical
o
b
s
er
v
atio
n
s
as
th
e
b
asis
f
o
r
ev
alu
atio
n
.
T
h
is
s
tu
d
y
th
e
r
ef
o
r
e
aim
s
to
d
ev
elo
p
a
n
d
v
alid
ate
AI
-
SI
M
as
an
i
n
teg
r
ated
p
latf
o
r
m
ca
p
ab
le
o
f
ass
es
s
in
g
MSM
E
d
ig
ital
r
ea
d
in
ess
,
g
en
er
atin
g
a
d
ap
tiv
e
r
ec
o
m
m
en
d
atio
n
s
,
s
u
p
p
o
r
tin
g
s
p
atial
b
u
s
in
ess
in
tellig
en
ce
,
an
d
f
ac
ilit
atin
g
e
x
p
er
ien
tial
e
n
tr
ep
r
e
n
eu
r
s
h
ip
le
ar
n
in
g
.
I
ts
ef
f
ec
tiv
en
ess
is
ev
a
lu
ated
th
r
o
u
g
h
ML
p
er
f
o
r
m
an
ce
,
s
y
s
tem
im
p
lem
e
n
tatio
n
,
an
d
f
ield
v
alid
atio
n
w
ith
MSM
E
s
an
d
in
s
titu
tio
n
al
s
tak
eh
o
ld
er
s
.
As
a
n
I
C
T
co
n
tr
ib
u
tio
n
,
t
h
e
s
tu
d
y
tr
an
s
p
ar
en
tly
d
o
c
u
m
en
ts
wh
er
e
a
s
u
r
v
ey
-
d
er
iv
e
d
,
r
u
le
-
b
ased
s
en
s
in
g
lay
er
s
u
cc
ee
d
s
an
d
f
alls
s
h
o
r
t o
f
r
eli
ab
le
au
to
m
ated
i
n
f
er
en
ce
,
p
r
o
v
id
in
g
a
b
aselin
e
f
u
tu
r
e
s
y
s
tem
s
ca
n
b
u
ild
u
p
o
n
.
T
h
e
p
r
im
ar
y
n
o
v
elty
lies
in
a
n
en
d
-
to
-
en
d
AI
-
d
r
iv
en
in
c
u
b
atio
n
f
r
am
ewo
r
k
th
at
s
im
u
ltan
eo
u
s
ly
in
teg
r
ates
au
to
m
ate
d
r
ea
d
in
ess
class
if
icatio
n
,
ad
ap
tiv
e
r
ec
o
m
m
en
d
atio
n
,
g
e
o
s
p
atial
in
tell
ig
en
ce
,
d
ash
b
o
ar
d
an
aly
tics
,
an
d
s
tu
d
en
t
–
MS
ME
co
llab
o
r
atio
n
i
n
o
n
e
en
v
ir
o
n
m
en
t.
Un
lik
e
c
o
n
v
e
n
tio
n
al
ap
p
r
o
ac
h
es
em
p
h
asizin
g
g
en
er
ic
m
en
t
o
r
in
g
,
AI
-
SIM
d
eliv
e
r
s
p
er
s
o
n
alize
d
,
ev
i
d
en
ce
-
b
ased
in
ter
v
en
ti
o
n
p
ath
way
s
wh
ile
ex
p
licitly
d
o
cu
m
e
n
tin
g
th
e
em
p
ir
ical
lim
its
o
f
it
s
s
u
r
v
ey
-
d
er
i
v
ed
p
r
o
x
ies
—
a
tr
an
s
p
ar
en
cy
p
r
ac
tice
th
at
its
elf
co
n
tr
ib
u
tes
to
tr
u
s
two
r
th
y
,
a
u
d
i
tab
le
I
C
T
in
cu
b
atio
n
s
y
s
tem
s
.
T
h
e
f
r
am
ew
o
r
k
also
ali
g
n
s
with
em
er
g
in
g
p
er
s
p
ec
tiv
es
o
n
AI
-
en
a
b
led
cr
ea
tiv
e
ec
o
n
o
m
ies
[
2
5
]
,
d
i
g
ital
en
tr
ep
r
e
n
eu
r
ial
ec
o
s
y
s
tem
s
[
2
6
]
,
an
d
th
e
Qu
in
tu
p
le
Helix
in
n
o
v
atio
n
m
o
d
el
[
2
7
]
,
co
n
tr
ib
u
tin
g
a
s
ca
lab
le,
r
ep
licab
le
I
C
T
f
r
am
ew
o
r
k
f
o
r
MSM
E
d
ig
ital
tr
an
s
f
o
r
m
atio
n
an
d
e
n
tr
ep
r
e
n
e
u
r
s
h
ip
ed
u
c
atio
n
.
T
h
e
r
em
ain
d
e
r
o
f
t
h
is
p
ap
er
p
r
o
ce
ed
s
as f
o
llo
ws.
Sectio
n
2
d
escr
ib
es th
e
m
eth
o
d
o
lo
g
y
,
c
o
v
er
in
g
d
ata
ac
q
u
is
itio
n
,
p
r
ep
r
o
ce
s
s
in
g
,
ML
m
o
d
elin
g
,
s
y
s
tem
im
p
lem
e
n
tatio
n
,
an
d
v
alid
atio
n
.
Sectio
n
3
p
r
esen
ts
r
esu
lts
an
d
d
is
cu
s
s
io
n
,
in
clu
d
in
g
m
o
d
el
p
er
f
o
r
m
a
n
ce
,
f
ea
t
u
r
e
im
p
o
r
tan
ce
,
d
ash
b
o
a
r
d
im
p
lem
en
tatio
n
,
a
n
d
co
m
p
ar
ativ
e
p
o
s
itio
n
in
g
in
th
e
I
C
T
-
in
cu
b
atio
n
liter
atu
r
e.
Sectio
n
4
co
n
cl
u
d
es
with
m
ajo
r
f
in
d
in
g
s
,
co
n
tr
ib
u
tio
n
s
,
lim
itatio
n
s
,
an
d
f
u
tu
r
e
d
ir
e
ctio
n
s
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
is
s
tu
d
y
a
d
o
p
ts
a
d
esig
n
s
cien
ce
r
esear
ch
ap
p
r
o
ac
h
co
m
p
r
is
in
g
f
iv
e
s
tag
es:
(
1
)
d
ata
a
cq
u
is
itio
n
,
(
2
)
p
r
ep
r
o
ce
s
s
in
g
an
d
f
ea
tu
r
e
en
g
in
ee
r
in
g
,
(
3
)
ML
-
b
ased
d
ig
ital
-
r
ea
d
in
ess
class
if
ic
atio
n
,
(
4
)
s
y
s
tem
im
p
lem
en
tatio
n
,
an
d
(
5
)
v
alid
atio
n
th
r
o
u
g
h
f
ield
d
ep
lo
y
m
en
t,
illu
s
tr
ated
i
n
Fig
u
r
e
1
.
T
h
is
s
tr
u
ct
u
r
e
d
wo
r
k
f
lo
w
en
s
u
r
es b
o
th
tec
h
n
i
ca
l r
ig
o
r
a
n
d
p
r
ac
tical
r
elev
an
ce
f
o
r
MSM
E
em
p
o
wer
m
en
t
[
1
0
]
,
[
1
3
]
,
[
2
0
]
.
2
.
1
.
Da
t
a
a
cquis
it
io
n
Prim
ar
y
d
ata
wer
e
c
o
llected
f
r
o
m
4
,
2
7
6
MSM
E
s
th
r
o
u
g
h
s
tr
u
ctu
r
e
d
s
u
r
v
e
y
s
an
d
i
n
s
titu
tio
n
al
d
atab
ases
b
etwe
en
J
an
u
ar
y
a
n
d
J
u
n
e
2
0
2
5
ac
r
o
s
s
L
am
p
u
n
g
Pro
v
in
ce
,
I
n
d
o
n
esia,
u
s
in
g
s
tr
atif
ied
r
an
d
o
m
s
am
p
lin
g
co
o
r
d
in
ated
with
th
e
Dep
ar
tm
en
t
o
f
C
o
o
p
e
r
ativ
e
s
,
Sm
all
an
d
Me
d
iu
m
E
n
ter
p
r
is
es
o
f
L
am
p
u
n
g
Pro
v
in
ce
.
T
h
e
s
u
r
v
ey
ca
p
t
u
r
e
d
in
d
icat
o
r
s
m
ap
p
ed
to
f
iv
e
c
o
n
s
tr
u
cts
d
ig
ital
i
n
f
r
astru
ctu
r
e
,
o
n
lin
e
m
a
r
k
etin
g
ad
o
p
tio
n
,
f
in
an
cial
m
an
a
g
em
en
t
ca
p
ab
ilit
y
,
I
T
s
k
ills
,
an
d
g
eo
g
r
a
p
h
ic
clu
s
ter
d
er
iv
e
d
f
r
o
m
MSM
E
d
ig
ital
-
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
6
8
2
-
1
695
1684
tr
an
s
f
o
r
m
atio
n
f
r
am
ewo
r
k
s
[
2
]
-
[
4
]
,
with
s
p
atial
attr
ib
u
tes
in
co
r
p
o
r
ated
v
ia
Geo
Pan
d
as
f
o
r
g
eo
s
p
atial
clu
s
ter
in
g
[
1
7
]
,
[
1
8
]
.
R
esp
o
n
d
en
ts
s
p
an
n
ed
m
u
ltip
le
r
eg
en
cies
a
n
d
cities
in
L
am
p
u
n
g
Pro
v
in
ce
in
clu
d
i
n
g
B
an
d
ar
L
am
p
u
n
g
,
Me
tr
o
,
Pes
awa
r
an
,
P
r
in
g
s
ewu
,
an
d
L
am
p
u
n
g
Se
latan
r
etain
ed
as
s
elec
t
ab
le
r
e
g
io
n
al
f
ilter
s
in
th
e
AI
-
SIM
d
ash
b
o
ar
d
t
o
ca
p
tu
r
e
h
eter
o
g
en
e
o
u
s
r
eg
io
n
al
r
ea
d
in
ess
p
atter
n
s
r
ath
er
th
an
a
s
in
g
le
-
lo
ca
lity
s
n
ap
s
h
o
t.
T
h
e
au
th
o
r
s
ac
k
n
o
wled
g
e
th
a
t
a
s
ec
to
r
-
lev
el
b
r
ea
k
d
o
wn
(
e
.
g
.
,
cu
lin
a
r
y
,
f
ash
io
n
an
d
cr
af
t,
r
etail
an
d
tr
ad
e
,
ag
r
icu
ltu
re
-
b
ased
,
s
er
v
ices)
was n
o
t sep
ar
ately
en
co
d
ed
as a
m
o
d
elin
g
f
ea
tu
r
e
in
th
is
r
elea
s
e;
th
is
is
ad
d
r
ess
ed
as
a
d
ata
-
en
r
ich
m
en
t
p
r
io
r
it
y
in
Sectio
n
3
.
4
.
Fig
u
r
e
2
s
u
m
m
ar
izes
th
e
d
ataset
co
n
s
tr
u
ctio
n
:
two
-
s
tep
d
ed
u
p
licatio
n
r
ed
u
ce
d
th
e
4
,
2
7
6
r
aw
r
ec
o
r
d
s
to
9
6
1
u
n
iq
u
e
an
aly
tical
r
ec
o
r
d
s
,
a
n
d
th
e
en
g
in
ee
r
ed
f
ea
tu
r
es
o
f
Pip
elin
e
2
ar
e
g
r
o
u
p
ed
in
to
f
ir
m
o
g
r
ap
h
ic,
ca
teg
o
r
ized
,
a
n
d
s
p
atial
ty
p
es.
Fig
u
r
e
1
.
R
esear
ch
f
r
am
ewo
r
k
o
f
th
e
AI
-
SIM
d
ev
elo
p
m
en
t
p
r
o
ce
s
s
Fig
u
r
e
2
.
Data
s
et
MSM
E
s
2
.
2
.
Da
t
a
prepro
ce
s
s
ing
a
nd
f
ea
t
ure
eng
ineering
A
cr
itical
p
r
ep
r
o
ce
s
s
in
g
s
tep
in
v
o
lv
ed
two
-
lev
el
d
e
d
u
p
licat
io
n
to
p
r
ev
en
t
d
ata
leak
ag
e.
E
x
ac
t
r
o
w
d
u
p
licates
wer
e
f
ir
s
t
r
em
o
v
e
d
,
r
e
d
u
cin
g
4
,
2
7
6
r
aw
o
b
s
er
v
atio
n
s
to
2
,
3
3
4
r
ec
o
r
d
s
;
a
s
tr
ict
r
u
le
b
ased
o
n
th
e
co
m
p
o
s
ite
k
ey
o
f
b
r
a
n
d
n
am
e,
p
h
o
n
e
n
u
m
b
er
,
an
d
tim
e
s
tam
p
th
en
y
ield
ed
9
6
1
u
n
iq
u
e
r
ec
o
r
d
s
.
T
h
is
s
tep
is
m
eth
o
d
o
lo
g
ically
ess
en
tial:
r
etain
in
g
d
u
p
licates
ac
r
o
s
s
tr
ain
/tes
t
s
p
lit
s
ar
tific
ially
i
n
f
lated
ac
cu
r
ac
y
to
93
–
9
9
% in
p
r
elim
in
a
r
y
ex
p
er
i
m
en
ts
,
wh
ich
d
o
es n
o
t r
ef
le
ct
l
eg
itima
te
g
en
er
aliza
tio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Dev
elo
p
men
t o
f a
n
A
I
-
b
a
s
ed
d
ig
ita
l sta
r
tu
p
in
cu
b
a
tio
n
mo
d
e
l fo
r
th
e
emp
o
w
ermen
t o
f
… (
Uma
r
B
a
kti
)
1685
Nu
m
er
ical
attr
ib
u
tes we
r
e
n
o
r
m
alize
d
u
s
in
g
th
e
s
tan
d
a
r
d
z
-
sc
o
r
e
tr
an
s
f
o
r
m
atio
n
s
h
o
wn
in
(
1
)
:
′
=
(
−
)
/
(
1
)
C
ateg
o
r
ical
v
ar
iab
les
wer
e
en
co
d
ed
u
s
in
g
o
n
e
-
h
o
t
o
r
o
r
d
in
al
en
co
d
in
g
d
ep
e
n
d
in
g
o
n
v
ar
iab
le
ty
p
e
,
wit
h
f
ea
tu
r
e
r
elev
an
ce
f
o
llo
we
d
MSM
E
d
ig
ital
m
atu
r
ity
s
tu
d
ies
[
1
4
]
,
[
2
1
]
.
T
wo
f
ea
tu
r
e
co
n
f
ig
u
r
atio
n
s
wer
e
ev
alu
ated
(
Sectio
n
2
.
3
)
.
T
h
e
m
an
u
s
cr
ip
t
-
alig
n
e
d
v
ec
to
r
is
g
iv
en
in
(
2
)
:
=
[
,
,
,
,
,
,
]
(
2
)
wh
er
e:
−
DI
=
Dig
ital I
n
f
r
astru
ctu
r
e
(
p
r
o
x
ied
b
y
a
3
-
lev
el
d
ig
ital c
ap
a
city
in
d
icato
r
,
r
escaled
to
1
–
5)
−
OM
=
On
lin
e
Ma
r
k
etin
g
(
p
r
o
x
ied
b
y
b
in
ar
y
e
-
co
m
m
e
r
ce
ac
c
ess
,
r
escaled
to
1
–
5)
−
FM
=
Fin
an
cial
Ma
n
ag
em
en
t
(
p
r
o
x
ie
d
b
y
b
u
s
in
ess
leg
al
s
tatu
s
,
r
escaled
to
1
–
5)
−
I
T
=
I
T
Sk
ills
(
p
r
o
x
ied
b
y
AI
p
r
o
f
icien
c
y
lev
el,
r
escaled
to
2
–
5)
−
L
C
=
L
o
ca
tio
n
C
lu
s
ter
(
lab
el
-
en
co
d
ed
r
eg
en
c
y
/city
)
−
OB
=
Mo
n
th
ly
R
ev
en
u
e
(
o
m
z
et
b
u
lan
an
,
p
ar
s
ed
f
r
o
m
f
r
ee
-
t
ex
t r
esp
o
n
s
es)
−
KY
=
E
m
p
lo
y
ee
C
o
u
n
t (
k
ar
y
a
wan
,
p
ar
s
ed
f
r
o
m
f
r
ee
-
tex
t
r
esp
o
n
s
es)
T
h
e
v
ar
iab
le
b
u
s
in
ess
_
ag
e
was
ex
clu
d
ed
f
r
o
m
all
m
o
d
els:
p
r
elim
in
ar
y
in
v
esti
g
atio
n
id
en
tifie
d
it
as
d
o
m
in
an
t
(
7
1
.
4
%
im
p
o
r
tan
ce
)
in
u
n
co
n
tr
o
lled
ex
p
er
im
en
ts
,
in
d
icatin
g
lik
ely
s
p
u
r
i
o
u
s
co
r
r
elatio
n
o
r
leak
ag
e
r
ath
er
th
an
g
en
u
i
n
e
s
ig
n
al.
T
h
is
au
to
m
ated
f
ea
tu
r
e
-
im
p
o
r
tan
ce
au
d
itin
g
s
er
v
ed
as
a
leak
ag
e
-
d
etec
tio
n
m
ec
h
an
is
m
an
d
is
r
ec
o
m
m
e
n
d
ed
as
a
s
tan
d
a
r
d
p
r
e
-
d
e
p
l
o
y
m
en
t
c
h
ec
k
f
o
r
f
u
tu
r
e
AI
-
en
ab
led
in
c
u
b
atio
n
s
y
s
tem
s
.
2
.
3
.
Dig
it
a
l r
ea
din
ess
cla
s
s
if
i
ca
t
io
n
R
an
d
o
m
f
o
r
est
class
if
ied
MSM
E
s
in
to
T
r
ad
itio
n
al,
Stab
le,
an
d
Dig
ital
R
ea
d
y
ca
teg
o
r
ies
v
ia
en
s
em
b
le
d
ec
is
io
n
tr
ee
s
[
1
6
]
,
[
2
1
]
,
ch
o
s
en
o
v
e
r
d
ee
p
-
lear
n
in
g
alter
n
ativ
es
f
o
r
th
r
ee
r
ea
s
o
n
s
.
First,
s
am
p
le
-
s
ize
ad
eq
u
ac
y
:
with
o
n
ly
9
6
1
leak
ag
e
-
f
r
ee
r
ec
o
r
d
s
,
d
ee
p
er
ar
ch
itectu
r
es
r
is
k
o
v
er
f
itti
n
g
.
Seco
n
d
,
in
ter
p
r
etab
ilit
y
:
R
an
d
o
m
f
o
r
est
'
s
f
ea
tu
r
e
-
im
p
o
r
ta
n
ce
r
an
k
in
g
s
(
Sectio
n
3
.
3
)
a
r
e
d
ir
ec
tly
u
s
ab
le
b
y
n
o
n
-
tech
n
ical
in
cu
b
atio
n
s
taf
f
,
a
p
r
ac
tical
r
eq
u
ir
em
en
t
f
o
r
d
as
h
b
o
ar
d
-
em
b
e
d
d
ed
d
ec
is
io
n
s
u
p
p
o
r
t.
T
h
ir
d
,
p
r
io
r
ev
id
en
ce
s
h
o
ws
it
co
m
p
etitiv
e
with
o
r
s
u
p
e
r
io
r
to
m
o
r
e
c
o
m
p
lex
lear
n
er
s
o
n
s
im
ilar
MSM
E
class
if
icatio
n
task
s
[
1
6
]
,
[
1
9
]
,
[
2
1
]
.
T
wo
p
ip
elin
es we
r
e
ev
alu
ated
in
p
ar
all
el:
−
Pip
elin
e
1
(
Or
ig
in
al)
:
ten
f
ea
tu
r
es
ex
tr
ac
ted
d
ir
ec
tly
f
r
o
m
b
in
ar
y
an
d
o
r
d
in
al
s
u
r
v
e
y
r
esp
o
n
s
es
(
d
ig
ital
ca
p
ac
ity
,
clo
u
d
ac
ce
s
s
/s
k
ill,
AI
ac
ce
s
s
/s
k
ill,
e
-
co
m
m
er
ce
ac
ce
s
s
/s
k
ill,
leg
al
s
tatu
s
,
m
ar
k
etp
lac
e
p
ar
ticip
atio
n
,
an
d
b
u
s
in
ess
ag
e)
.
−
Pip
elin
e
2
(
Ma
n
u
s
cr
ip
t
-
Alig
n
ed
)
:
th
e
s
ev
en
-
f
ea
tu
r
e
v
ec
t
o
r
in
(
2
)
,
d
ir
ec
tly
r
ef
lectin
g
th
e
m
an
u
s
cr
i
p
t'
s
co
n
ce
p
tu
al
c
o
n
s
tr
u
cts.
Fo
r
Pip
elin
e
1
,
Gr
id
Sear
ch
C
V
with
5
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
id
en
tifie
d
class
_
weig
h
t='
b
alan
ce
d
'
,
m
ax
_
d
e
p
th
=N
o
n
e,
m
in
_
s
am
p
les_
s
p
lit=2
,
n
_
esti
m
ato
r
s
=1
0
0
as
o
p
tim
al;
f
o
r
Pip
elin
e
2
,
c
lass
_
weig
h
t=No
n
e,
m
ax
_
d
e
p
th
=2
0
,
m
in
_
s
am
p
le
s
_
s
p
lit=1
0
,
n
_
esti
m
ato
r
s
=2
0
0
.
T
h
e
Pip
elin
e
2
s
ea
r
c
h
ev
alu
ate
d
2
8
8
h
y
p
er
p
ar
am
eter
co
m
b
i
n
atio
n
s
u
n
d
er
5
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
(
Sectio
n
3
.
6
)
,
s
wee
p
in
g
n
_
esti
m
ato
r
s
,
m
ax
_
d
e
p
th
,
m
in
_
s
am
p
les_
s
p
lit,
an
d
class
_
weig
h
t jo
in
tly
.
T
h
e
p
r
e
d
ictio
n
f
u
n
ctio
n
is
g
iv
en
in
(
3
)
:
ŷ
=
(
₁
(
)
,
₂
(
)
,
…
,
ₙ
(
)
)
(
3
)
wh
er
e
T
ₙ
r
ep
r
esen
ts
ea
ch
d
ec
i
s
io
n
tr
ee
.
T
h
e
d
ataset
(
9
6
1
r
ec
o
r
d
s
)
was
s
p
lit
in
to
tr
ain
in
g
(
7
6
8
,
7
9
.
9
%)
an
d
test
in
g
(
1
9
3
,
2
0
.
1
%)
u
s
in
g
a
s
tr
atif
ied
8
0
:2
0
r
atio
(
s
ee
d
=
4
2
)
to
p
r
eser
v
e
class
d
is
tr
ib
u
tio
n
(
p
r
e
-
SMOT
E
tr
ain
in
g
d
is
tr
ib
u
tio
n
:
2
2
9
/2
5
6
/2
8
3
f
o
r
Dig
ital
R
ea
d
y
/St
ab
le/T
r
ad
itio
n
al)
.
SMOT
E
b
alan
cin
g
was
ap
p
lied
to
t
h
e
tr
ain
in
g
s
et
o
n
ly
.
T
o
ass
ess
r
o
b
u
s
tn
ess
,
r
an
d
o
m
f
o
r
est
was
b
en
ch
m
a
r
k
ed
ag
ain
s
t
d
ec
is
io
n
tr
ee
,
lo
g
is
tic
r
eg
r
ess
io
n
,
an
d
SVM
o
n
Pi
p
elin
e
1
u
s
in
g
th
e
s
am
e
8
0
:2
0
s
p
lit,
with
5
-
f
o
l
d
cr
o
s
s
-
v
alid
atio
n
v
er
if
y
in
g
s
tab
ilit
y
ac
r
o
s
s
p
ar
titi
o
n
s
.
Fig
u
r
e
3
tr
ac
es
th
e
en
d
-
to
-
en
d
p
ip
elin
e
f
r
o
m
t
h
e
d
ed
u
p
lic
ated
9
6
1
-
r
ec
o
r
d
s
u
r
v
ey
d
atas
et
th
r
o
u
g
h
f
ea
tu
r
e
en
co
d
in
g
o
f
th
e
s
ev
e
n
m
an
u
s
cr
ip
t
-
alig
n
e
d
in
d
ica
to
r
s
(
{DI
,
OM
,
FM,
I
T
,
L
C
,
O
B
,
KY}
)
,
s
tr
atif
ied
8
0
:2
0
tr
ain
–
test
s
p
litt
in
g
,
SM
OT
E
b
alan
cin
g
o
f
t
h
e
tr
ain
in
g
p
ar
titi
o
n
,
Gr
id
Sear
ch
C
V
h
y
p
er
p
ar
am
eter
tu
n
in
g
,
an
d
r
an
d
o
m
f
o
r
est
m
o
d
el
f
it
tin
g
,
to
th
e
f
in
al
th
r
ee
-
class
d
ig
ital
-
r
ea
d
in
ess
p
r
ed
ictio
n
.
E
ac
h
s
tag
e
in
th
e
d
iag
r
am
co
r
r
esp
o
n
d
s
d
ir
ec
tly
to
a
s
tep
in
th
e
Alg
o
r
ith
m
1
p
s
eu
d
o
co
d
e
b
elo
w,
allo
win
g
t
h
e
p
r
o
ce
d
u
r
e
to
b
e
r
ep
r
o
d
u
ce
d
eith
e
r
v
is
u
ally
o
r
p
r
o
g
r
am
m
atica
lly
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
6
8
2
-
1
695
1686
Fig
u
r
e
3
.
MSM
E
d
ig
ital r
ea
d
i
n
ess
class
if
icatio
n
p
ip
elin
e
(
A
lg
o
r
ith
m
1
)
Alg
o
r
ith
m
1
: M
SME
d
ig
ital r
e
ad
in
ess
class
if
icat
io
n
Input: Deduplicated analytical dataset D (961 records, features {DI, OM, FM, IT, LC, OB,
KY})
Output: Trained classifier ŷ = mode(T
₁
(x), …, T
ₙ
(x)) and evaluation metrics
1: D ← deduplicate(D_raw, keys = {b
rand, phone, timestamp}) // Sec. 2.2
2: X ← normalize(D.numeric); X ← encode(D.categorical) // Eq. (1)
3: exclude business_age from X // leakage control
4: (X_train, y_train), (X_test, y_te
st) ← stratified_split(X, y, ratio = 0.8, seed = 42)
5: X_train, y_train ← SMOTE(X_train, y_train) // train set only
6: params ← GridSearchCV(RandomForest, grid, cv = 5, scoring = accuracy)
7: model ← RandomForest(**params).fit(X_t
rain, y_train)
8: ŷ_test ← model.predict(X_test) // Eq. (3)
9: metrics ← evaluate(y_test, ŷ_test) // accuracy, precision, recall, F1
—
Eqs. (4
–
7)
10: importances ← model.feature_importances_ //
Sec. 3.3
11: return model, metrics, importances
2
.
4
.
P
er
f
o
r
m
a
nce
m
e
t
rics
Mo
d
el
ev
alu
atio
n
u
s
ed
th
e
f
o
ll
o
win
g
s
tan
d
ar
d
class
if
icatio
n
m
etr
ics,
g
iv
en
in
(
4
)
–
(
7
)
:
Acc
u
r
ac
y
=
(
T
P +
T
N)
/ (
T
P +
T
N
+
FP
+
FN)
(
4
)
Pre
cisi
o
n
=
T
P /
(
T
P +
FP
)
(
5
)
R
ec
all
=
T
P /
(
T
P +
FN)
(
6
)
F1
=
2
×
(
Pre
cisi
o
n
×
R
ec
all)
/
(
Pre
cisi
o
n
+
R
ec
all)
(
7
)
wh
er
e
T
P,
T
N,
FP
,
an
d
FN
d
en
o
te
tr
u
e
p
o
s
itiv
es,
tr
u
e
n
eg
ativ
es,
f
alse
p
o
s
itiv
es,
an
d
f
alse
n
eg
ativ
es,
r
esp
ec
tiv
ely
.
2
.
5
.
Sy
s
t
e
m
a
rc
hite
ct
ure
a
n
d im
plem
ent
a
t
io
n
Fas
tAPI
h
an
d
led
b
ac
k
e
n
d
in
f
er
en
ce
an
d
Stre
am
lit
p
r
o
v
id
e
d
d
ash
b
o
ar
d
v
is
u
aliza
tio
n
[
2
4
]
;
ad
ap
tiv
e
r
ec
o
m
m
en
d
atio
n
u
s
ed
AI
-
b
as
ed
in
ter
v
en
tio
n
lo
g
ic
[
2
2
]
,
[
2
3
]
,
an
d
s
tu
d
en
ts
wer
e
m
ap
p
ed
to
MSM
E
s
b
y
r
ea
d
in
ess
lev
el
f
o
r
ex
p
er
ien
t
ial
lear
n
in
g
[
1
9
]
.
T
h
is
s
tateless
-
API
/d
ec
o
u
p
led
-
vi
s
u
aliza
tio
n
d
esig
n
f
o
llo
ws
s
tan
d
ar
d
I
C
T
p
r
ac
tice,
allo
win
g
th
e
class
if
icatio
n
b
ac
k
en
d
to
b
e
u
p
g
r
a
d
ed
(
e.
g
.
,
with
r
ich
er
f
ea
tu
r
es
p
e
r
Sectio
n
3
.
4
)
with
o
u
t c
h
a
n
g
in
g
th
e
d
ash
b
o
a
r
d
f
r
o
n
t e
n
d
.
2
.
6
.
Va
lid
a
t
i
o
n a
nd
ev
a
lua
t
i
o
n
Valid
atio
n
u
s
ed
a
tr
ian
g
u
late
d
ap
p
r
o
ac
h
—
q
u
an
titativ
e
ML
m
etr
ics,
f
ield
d
ep
lo
y
m
en
t,
an
d
f
o
cu
s
g
r
o
u
p
d
is
cu
s
s
io
n
s
—
f
o
llo
wi
n
g
in
c
u
b
atio
n
ass
ess
m
en
t
f
r
am
ewo
r
k
s
[
6
]
,
[
7
]
,
[
9
]
,
d
is
cu
s
s
ed
in
r
elatio
n
to
d
ig
ital e
n
tr
ep
r
e
n
eu
r
ial
ec
o
s
y
s
t
em
s
[
2
5
]
-
[
2
7
]
in
s
ec
tio
n
3.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Dev
elo
p
men
t o
f a
n
A
I
-
b
a
s
ed
d
ig
ita
l sta
r
tu
p
in
cu
b
a
tio
n
mo
d
e
l fo
r
th
e
emp
o
w
ermen
t o
f
… (
Uma
r
B
a
kti
)
1687
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
ec
tio
n
p
r
esen
ts
ex
p
er
im
en
tal
r
esu
lts
f
o
r
AI
-
SIM
an
d
d
is
cu
s
s
es
MS
ME
d
ig
ital
r
ea
d
in
ess
,
b
u
s
in
ess
ch
ar
ac
ter
is
tics
,
an
d
in
cu
b
atio
n
o
u
tco
m
es,
v
is
u
a
lized
th
r
o
u
g
h
class
if
icatio
n
m
etr
ics,
d
ash
b
o
ar
d
an
aly
tics
,
an
d
d
escr
ip
tiv
e
d
is
tr
ib
u
tio
n
s
[
3
]
,
[
2
4
]
.
3
.
1
.
Dig
it
a
l
re
a
din
ess
cla
s
s
if
i
ca
t
io
n per
f
o
rma
nce
Usi
n
g
th
e
f
in
al
9
6
1
-
r
ec
o
r
d
d
ataset,
r
an
d
o
m
f
o
r
est
ac
h
iev
ed
3
3
.
6
8
%
ac
cu
r
a
cy
(
P
ip
elin
e
1
,
ten
f
ea
tu
r
es)
an
d
3
0
.
5
7
%
(
Pip
elin
e
2
,
s
ev
en
f
ea
tu
r
es),
b
o
th
n
ea
r
o
r
b
elo
w
th
e
3
3
.
3
%
r
an
d
o
m
-
ch
an
ce
b
aselin
e
f
o
r
a
b
alan
ce
d
th
r
ee
-
class
p
r
o
b
lem
,
in
d
icatin
g
lim
ited
p
r
e
d
ictiv
e
ca
p
ab
ilit
y
f
r
o
m
th
e
c
u
r
r
en
t
s
u
r
v
ey
-
d
er
iv
e
d
in
d
icato
r
s
.
T
h
is
co
n
tr
asts
wit
h
an
ea
r
lier
u
n
co
n
tr
o
lled
r
u
n
r
ep
o
r
tin
g
9
3
.
5
7
%
ac
cu
r
ac
y
,
l
ater
f
o
u
n
d
to
b
e
an
ar
tifa
ct
o
f
d
u
p
licate
r
ec
o
r
d
s
s
p
an
n
in
g
tr
ain
i
n
g
an
d
test
in
g
p
ar
titi
o
n
s
(
Sectio
n
2
.
2
)
;
af
ter
s
tr
ict
d
ed
u
p
licatio
n
,
th
e
r
esu
lts
r
ep
o
r
ted
h
er
e
r
e
f
lec
t g
en
u
in
e
g
en
er
aliza
tio
n
to
u
n
s
ee
n
d
ata.
T
ab
le
1
s
h
o
ws
th
at
th
e
9
6
1
-
M
SME
s
am
p
le
s
k
ews
to
war
d
th
e
lo
wer
en
d
o
f
d
ig
ital
m
atu
r
ity
:
3
6
.
8
4
%
o
f
r
esp
o
n
d
en
ts
f
all
in
t
h
e
T
r
a
d
itio
n
al
ca
teg
o
r
y
,
3
3
.
4
0
%
in
Stab
le,
an
d
o
n
ly
2
9
.
7
6
%
ar
e
class
if
ied
as
Dig
ita
l
R
ea
d
y
.
T
h
is
n
ea
r
-
e
v
en
th
r
ee
-
way
s
p
lit,
with
a
s
lig
h
t
co
n
ce
n
tr
atio
n
in
th
e
least
d
ig
itally
m
atu
r
e
g
r
o
u
p
,
m
o
tiv
ates
th
e
class
if
icat
io
n
task
ad
d
r
ess
ed
in
th
is
s
tu
d
y
an
d
i
s
co
n
s
is
ten
t
with
th
e
m
o
d
est
b
aselin
e
ac
cu
r
ac
ies
r
ep
o
r
ted
s
u
b
s
eq
u
en
tly
.
T
ab
le
2
s
u
m
m
ar
izes
th
e
r
a
n
d
o
m
f
o
r
est
r
esu
lts
f
o
r
b
o
th
p
ip
elin
es.
Pip
elin
e
1
r
ea
ch
es
3
3
.
6
8
%
test
ac
cu
r
ac
y
an
d
a
3
6
.
7
5
%
m
ea
n
f
iv
e
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
ac
cu
r
ac
y
,
wh
er
ea
s
Pip
elin
e
2
r
ea
ch
es
3
0
.
5
7
%
test
ac
cu
r
ac
y
a
n
d
a
3
7
.
2
3
%
cr
o
s
s
-
v
alid
atio
n
m
ea
n
;
in
b
o
th
ca
s
es
th
e
test
ac
cu
r
ac
y
s
tay
s
clo
s
e
to
th
e
3
3
.
3
% r
an
d
o
m
-
ch
an
ce
b
aselin
e.
T
ab
le
3
b
en
c
h
m
ar
k
s
r
an
d
o
m
f
o
r
est
ag
ain
s
t
th
r
ee
alter
n
ativ
e
class
if
ier
s
o
n
Pip
elin
e
1
.
S
VM
attain
s
th
e
h
ig
h
est
r
aw
a
cc
u
r
ac
y
(
3
6
.
7
9
%)
b
u
t
with
s
u
b
s
tan
tially
lo
wer
p
r
ec
is
io
n
(
1
3
.
5
3
%)
an
d
F1
-
s
co
r
e
(
1
9
.
7
9
%),
in
d
icatin
g
a
b
ias
to
war
d
m
ajo
r
ity
-
class
p
r
ed
ictio
n
s
r
ath
er
th
an
b
alan
ce
d
class
if
icati
o
n
.
R
an
d
o
m
f
o
r
est
(
3
3
.
6
8
%
ac
cu
r
ac
y
,
2
7
.
9
4
%
p
r
ec
is
io
n
)
an
d
d
ec
is
io
n
tr
ee
(
3
4
.
2
0
%
ac
c
u
r
ac
y
,
2
8
.
3
0
%
p
r
ec
i
s
io
n
)
o
f
f
er
a
m
o
r
e
b
alan
ce
d
p
r
ec
is
io
n
–
r
ec
all
tr
ad
e
-
o
f
f
,
wh
ile
lo
g
is
tic
r
eg
r
ess
io
n
tr
ails
o
n
all
f
o
u
r
m
etr
ics.
T
h
ese
r
esu
lts
s
u
p
p
o
r
t
th
e
ch
o
ice
o
f
r
an
d
o
m
f
o
r
est
as
th
e
p
r
im
ar
y
class
if
ier
,
as
it
av
o
id
s
th
e
m
ajo
r
ity
-
class
b
ias
ev
id
en
t
in
SV
M'
s
in
f
lated
ac
cu
r
ac
y
f
ig
u
r
e.
T
ab
le
1
.
MSM
E
d
ig
ital r
ea
d
in
ess
d
is
tr
ib
u
tio
n
D
i
g
i
t
a
l
r
e
a
d
i
n
e
ss
c
a
t
e
g
o
r
y
N
u
mb
e
r
o
f
M
S
M
Es
(
n
)
P
e
r
c
e
n
t
a
g
e
(
%)
Tr
a
d
i
t
i
o
n
a
l
3
5
4
3
6
.
8
4
%
S
t
a
b
l
e
3
2
1
3
3
.
4
0
%
D
i
g
i
t
a
l
R
e
a
d
y
2
8
6
2
9
.
7
6
%
To
t
a
l
9
6
1
1
0
0
.
0
0
%
T
ab
le
2
.
C
lass
if
icatio
n
p
er
f
o
r
m
an
ce
r
esu
lts
(
r
an
d
o
m
f
o
r
est
)
P
i
p
e
l
i
n
e
A
c
c
u
r
a
c
y
5
-
F
o
l
d
C
V
M
e
a
n
P
i
p
e
l
i
n
e
1
(
O
r
i
g
i
n
a
l
,
1
0
f
e
a
t
u
r
e
s)
33
.
68
%
3
6
.
7
5
%
P
i
p
e
l
i
n
e
2
(
M
a
n
u
s
c
r
i
p
t
-
A
l
i
g
n
e
d
,
7
f
e
a
t
u
r
e
s)
3
0
.
5
7
%
3
7
.
2
3
%
T
ab
le
3
.
C
o
m
p
a
r
ativ
e
p
er
f
o
r
m
an
ce
o
f
class
if
icatio
n
m
o
d
els
(
Pip
elin
e
1
)
M
o
d
e
l
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
D
e
c
i
s
i
o
n
t
r
e
e
3
4
.
2
0
%
2
8
.
3
0
%
3
4
.
2
0
%
2
9
.
3
0
%
R
a
n
d
o
m
f
o
r
e
s
t
3
3
.
6
8
%
2
7
.
9
4
%
3
3
.
6
8
%
2
8
.
8
9
%
Lo
g
i
s
t
i
c
r
e
g
r
e
ssi
o
n
3
3
.
1
6
%
2
3
.
4
7
%
3
3
.
1
6
%
2
7
.
2
6
%
S
V
M
3
6
.
7
9
%
1
3
.
5
3
%
3
6
.
7
9
%
1
9
.
7
9
%
Fig
u
r
e
4
v
is
u
alize
s
th
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
c
o
r
e
v
alu
es r
ep
o
r
te
d
in
T
ab
le
3
ac
r
o
s
s
th
e
f
o
u
r
class
if
ier
s
.
T
h
e
c
h
ar
t
h
i
g
h
lig
h
ts
th
e
p
r
ec
is
io
n
–
ac
cu
r
a
cy
g
ap
m
o
s
t
clea
r
ly
f
o
r
SV
M,
wh
o
s
e
b
ar
f
o
r
ac
cu
r
ac
y
is
v
is
ib
ly
taller
th
an
its
p
r
ec
is
io
n
an
d
F1
-
s
co
r
e
b
ar
s
,
wh
ile
r
an
d
o
m
f
o
r
est
an
d
d
ec
is
io
n
tr
ee
s
h
o
w
m
o
r
e
ev
en
ly
m
atc
h
ed
b
ar
s
a
cr
o
s
s
all
f
o
u
r
m
etr
ics
—
r
ein
f
o
r
cin
g
th
e
tex
t'
s
co
n
clu
s
io
n
th
at
r
an
d
o
m
f
o
r
est
p
r
o
v
id
es
th
e
m
o
r
e
b
alan
ce
d
a
n
d
r
eliab
le
class
if
ier
.
Fro
m
an
I
C
T
d
esig
n
s
tan
d
p
o
in
t,
a
n
e
ar
-
ch
an
ce
class
if
ier
s
till
f
u
n
ctio
n
s
as
a
d
iag
n
o
s
tic
in
s
tr
u
m
en
t
r
ev
ea
lin
g
wh
ic
h
p
ar
ts
o
f
th
e
s
u
r
v
ey
in
s
tr
u
m
en
t
ca
r
r
y
u
s
ab
le
s
ig
n
al,
g
u
id
in
g
th
e
n
e
x
t iter
atio
n
(
Sec
tio
n
3
.
4
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
6
8
2
-
1
695
1688
Fig
u
r
e
4
.
C
o
m
p
a
r
ativ
e
m
o
d
el
p
er
f
o
r
m
ac
e
m
etr
ics
3
.
2
.
M
SM
E
ca
pita
l
s
t
ruct
ur
e
a
nd
dig
it
a
l a
cc
ess
T
h
i
s
s
u
b
s
ec
tio
n
r
ep
o
r
ts
d
e
s
cr
ip
tiv
e,
n
o
n
-
ML
an
a
ly
s
e
s
o
f
M
SM
E
ca
p
it
al
s
tr
u
ct
u
r
e,
le
g
a
l
f
o
r
m
a
li
za
tio
n
,
an
d
m
ar
k
e
tp
la
ce
a
cc
e
s
s
f
r
o
m
th
e
s
am
e
9
6
1
-
r
ec
o
r
d
d
a
ta
s
e
t,
s
u
r
f
ac
ed
th
r
o
u
g
h
th
e
AI
-
SI
M
Fig
u
r
e
5
.
Di
s
tr
i
b
u
t
io
n
o
f
i
n
it
ia
l
MS
ME
ca
p
i
ta
l
Fig
u
r
e
4
.
C
o
m
p
ar
a
t
iv
e
m
o
d
e
l
p
e
r
f
o
r
m
an
ce
m
e
tr
i
c
s
d
ash
b
o
ar
d
.
Fig
u
r
e
5
s
h
o
w
s
in
it
ia
l
MS
ME
ca
p
i
ta
l:
ab
o
u
t
7
0
%
o
f
en
ter
p
r
i
s
e
s
o
p
er
at
e
b
elo
w
I
D
R
5
0
m
i
l
lio
n
an
d
a
b
o
u
t
9
0
%
b
elo
w
I
D
R
1
0
0
m
i
l
lio
n
,
wi
th
o
n
ly
a
s
m
a
ll
f
r
a
ct
io
n
ex
c
ee
d
in
g
I
D
R
1
0
0
m
il
l
io
n
,
co
n
f
ir
m
in
g
th
a
t
m
o
s
t
p
a
r
t
ic
ip
an
t
s
a
r
e
m
icr
o
-
en
ter
p
r
i
s
e
s
w
ith
l
im
it
ed
a
cc
e
s
s
to
f
in
an
c
ia
l
r
eso
u
r
ce
s
an
d
d
ig
i
ta
l
in
f
r
a
s
tr
u
c
tu
r
e
—
co
n
d
i
tio
n
s
th
at
ju
s
t
if
y
ad
ap
t
iv
e
in
cu
b
at
io
n
s
u
p
p
o
r
t [
2
]
,
[
5
]
.
Fig
u
r
e
5
.
Dis
tr
ib
u
tio
n
o
f
in
itia
l
MSM
E
ca
p
ital
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Dev
elo
p
men
t o
f a
n
A
I
-
b
a
s
ed
d
ig
ita
l sta
r
tu
p
in
cu
b
a
tio
n
mo
d
e
l fo
r
th
e
emp
o
w
ermen
t o
f
… (
Uma
r
B
a
kti
)
1689
Fig
u
r
e
6
s
h
o
ws
leg
al
s
t
atu
s
:
th
e
m
ajo
r
ity
o
f
MSM
E
s
(
ap
p
r
o
x
im
ately
6
5
%)
r
em
ain
in
f
o
r
m
al/u
n
r
eg
is
ter
ed
,
wh
ile
s
m
aller
s
h
ar
es
a
r
e
leg
ally
r
eg
is
ter
ed
(
NI
B
/SIUP;
ap
p
r
o
x
im
ately
2
5
%)
o
r
in
co
r
p
o
r
ated
(
PT/C
V;
ap
p
r
o
x
im
ately
1
0
%),
r
ein
f
o
r
cin
g
p
r
io
r
f
in
d
in
g
s
o
n
s
tr
u
ct
u
r
al
b
a
r
r
ier
s
to
f
in
a
n
cin
g
,
g
o
v
er
n
m
en
t p
r
o
g
r
am
s
,
an
d
d
ig
ital p
latf
o
r
m
s
[
4
]
,
[
6
]
.
Fig
u
r
e
7
s
h
o
ws
m
a
r
k
etp
lace
ac
ce
s
s
ib
ilit
y
:
a
m
in
o
r
ity
o
f
MSM
E
s
(
ap
p
r
o
x
im
ately
2
0
%)
ar
e
f
u
lly
in
teg
r
ated
with
o
n
lin
e
m
a
r
k
etp
lace
s
,
ap
p
r
o
x
im
ately
3
5
%
r
ep
o
r
t
p
ar
tial
ac
ce
s
s
,
an
d
t
h
e
lar
g
est
s
h
ar
e
(
ap
p
r
o
x
im
ately
4
5
%)
r
ep
o
r
t n
o
n
e,
in
d
icatin
g
u
n
ev
en
d
ig
ital
ad
o
p
tio
n
co
n
s
is
ten
t w
ith
p
r
io
r
f
in
d
in
g
s
[
6
]
,
[
9
]
.
Fig
u
r
e
6
.
MSM
E
leg
al
s
tatu
s
.
Per
ce
n
tag
es a
r
e
r
o
u
n
d
e
d
to
th
e
n
ea
r
est 5
%
Fig
u
r
e
7
.
Ma
r
k
etp
lace
ac
ce
s
s
ib
ilit
y
.
Per
ce
n
tag
es a
r
e
r
o
u
n
d
e
d
to
th
e
n
ea
r
est 5
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
15
,
No
.
4
,
Dec
em
b
er
20
26
:
1
6
8
2
-
1
695
1690
3
.
3
.
F
e
a
t
ure
co
ntr
ibu
t
io
n a
nd
AI
-
SI
M
da
s
hb
o
a
rd
ins
ig
hts
Featu
r
e
im
p
o
r
tan
ce
f
o
r
th
e
m
an
u
s
cr
ip
t
-
alig
n
ed
p
i
p
elin
e
is
d
is
tr
ib
u
ted
f
air
ly
ev
e
n
ly
ac
r
o
s
s
th
e
s
ev
e
n
f
ea
tu
r
es
r
ath
er
th
an
c
o
n
ce
n
t
r
a
ted
in
a
f
ew
p
r
e
d
icto
r
s
(
Fig
u
r
e
8
)
.
Dig
ital
in
f
r
astru
ctu
r
e
r
a
n
k
s
h
ig
h
est
(
0
.
2
2
)
,
f
o
llo
wed
b
y
m
o
n
th
ly
r
e
v
en
u
e
(
0
.
1
8
)
,
o
n
lin
e
m
a
r
k
etin
g
(
0
.
1
5
)
,
f
in
an
cial
m
an
ag
e
m
en
t
(
0
.
1
4
)
,
em
p
lo
y
ee
c
o
u
n
t
(
0
.
1
2
)
,
I
T
s
k
ills
(
0
.
1
1
)
,
an
d
l
o
ca
tio
n
clu
s
ter
(
0
.
0
8
)
[
1
7
]
,
[
1
9
]
.
T
h
is
d
is
tr
ib
u
tio
n
s
u
g
g
ests
th
at
n
o
s
in
g
le
in
d
icato
r
d
o
m
in
ates
th
e
cla
s
s
if
icatio
n
:
th
e
f
o
u
r
d
ig
ital
-
ca
p
ab
ilit
y
co
n
s
tr
u
cts
(
d
ig
ital
i
n
f
r
astru
ctu
r
e,
o
n
lin
e
m
ar
k
etin
g
,
f
in
a
n
cial
m
an
a
g
em
en
t,
a
n
d
I
T
s
k
ills
)
to
g
eth
e
r
ac
co
u
n
t
f
o
r
0
.
6
2
o
f
th
e
to
tal
i
m
p
o
r
tan
ce
,
wh
ile
th
e
o
b
jectiv
e
f
ir
m
ch
a
r
ac
ter
is
tics
(
m
o
n
th
l
y
r
ev
e
n
u
e,
e
m
p
lo
y
ee
co
u
n
t,
an
d
lo
ca
tio
n
clu
s
ter
)
a
cc
o
u
n
t
f
o
r
0
.
3
8
.
T
o
g
eth
e
r
wit
h
th
e
n
ea
r
-
c
h
an
ce
ac
cu
r
ac
y
r
ep
o
r
ted
ab
o
v
e,
th
is
in
d
icate
s
th
at
th
e
cu
r
r
en
t
s
u
r
v
ey
in
s
tr
u
m
e
n
t,
wh
ich
r
elies
lar
g
ely
o
n
c
o
ar
s
e
L
ik
e
r
t
-
s
ca
le
item
s
p
r
o
n
e
to
s
o
cial
-
d
esira
b
ilit
y
b
ias,
d
o
es
n
o
t
ca
r
r
y
en
o
u
g
h
d
is
cr
im
in
ato
r
y
s
ig
n
al.
A
f
u
tu
r
e
s
en
s
in
g
lay
er
s
h
o
u
ld
t
h
er
ef
o
r
e
ad
d
o
b
jectiv
e,
s
y
s
tem
-
lo
g
g
e
d
in
d
icato
r
s
(
tr
an
s
ac
tio
n
v
o
lu
m
e
,
p
latf
o
r
m
u
s
ag
e,
d
ig
ital
-
p
ay
m
en
t
r
ec
o
r
d
s
)
,
as
elab
o
r
ated
in
Sectio
n
3
.
4
.
T
h
e
AI
-
SIM
d
ash
b
o
ar
d
(
Fig
u
r
e
9
)
in
te
g
r
ates
th
ese
o
u
t
p
u
ts
i
n
to
an
in
ter
ac
tiv
e
e
n
v
ir
o
n
m
en
t
s
u
r
f
ac
in
g
r
eg
io
n
al,
leg
al
-
s
tatu
s
,
an
d
m
ar
k
etp
lace
-
a
cc
ess
in
d
icato
r
s
in
r
ea
l
tim
e
en
ab
lin
g
p
o
licy
m
ak
e
r
s
,
in
cu
b
ato
r
s
,
an
d
ac
ad
em
ic
p
ar
t
n
er
s
to
m
o
n
ito
r
r
ea
d
i
n
ess
,
ca
p
ital
s
tr
u
ct
u
r
e,
leg
ality
,
an
d
m
ar
k
et
ac
ce
s
s
,
a
n
d
to
v
iew
a
d
ap
tiv
e
tr
ain
in
g
r
ec
o
m
m
en
d
atio
n
s
alig
n
ed
with
p
e
r
s
o
n
alize
d
m
en
t
o
r
in
g
ap
p
r
o
ac
h
es
[
1
4
]
,
[
2
2
]
.
A
s
an
I
C
T
ar
tifa
ct,
it
in
teg
r
ates
class
if
icatio
n
-
b
ased
tr
iag
e,
g
eo
s
p
atial
s
itu
ati
o
n
al
awa
r
en
ess
,
an
d
r
ec
o
m
m
en
d
atio
n
-
d
r
iv
e
n
in
ter
v
en
tio
n
r
o
u
tin
g
in
o
n
e
in
t
er
f
ac
e,
p
o
s
itio
n
in
g
AI
-
SIM
as
a
s
m
ar
t in
cu
b
atio
n
p
latf
o
r
m
[
2
5
]
,
[
2
6
]
.
Fig
u
r
e
8
.
Featu
r
e
im
p
o
r
ta
n
ce
o
f
th
e
r
a
n
d
o
m
f
o
r
est
m
o
d
el
(
P
ip
elin
e
2
)
3
.
3
.
1.
Student
–
M
S
M
E
co
lla
b
o
ra
t
io
n o
utc
o
m
es
Stu
d
en
t
–
MSM
E
co
llab
o
r
atio
n
in
v
o
l
v
ed
7
1
u
n
i
v
er
s
ity
s
tu
d
en
ts
an
d
1
3
MSM
E
s
,
c
o
o
r
d
in
ate
d
b
y
eig
h
t
f
ield
f
ac
ilit
ato
r
s
(
PLUT
MSM
E
Facilitato
r
–
L
am
p
u
n
g
Pro
v
in
ce
)
,
with
s
tu
d
e
n
ts
ass
ig
n
ed
b
y
r
ea
d
i
n
ess
class
if
icatio
n
.
T
r
ain
in
g
n
ee
d
s
co
n
ce
n
tr
ated
in
m
ar
k
etin
g
(
n
=4
4
,
5
2
.
4
%)
an
d
g
en
er
al
b
u
s
in
ess
d
ev
elo
p
m
en
t
(
n
=2
8
,
3
3
.
3
%),
with
s
m
aller
s
h
ar
es
in
co
m
b
in
ed
m
ar
k
etin
g
–
b
u
s
in
ess
(
n
=7
,
8
.
3
%)
an
d
AI
-
s
p
ec
if
ic
tr
ain
in
g
(
n
=5
,
6
.
0
%).
FGDs
in
d
icate
d
im
p
r
o
v
e
d
MSM
E
awa
r
en
ess
o
f
d
ig
ital
to
o
ls
an
d
s
tr
o
n
g
er
s
tu
d
en
t
en
g
ag
em
e
n
t,
s
u
p
p
o
r
tin
g
ac
ad
em
ia
–
i
n
d
u
s
tr
y
in
teg
r
atio
n
a
n
d
en
t
r
ep
r
en
eu
r
s
h
ip
-
ed
u
ca
tio
n
m
o
d
els [
1
9
]
,
[
2
0
]
.
3
.
3
.
2
C
o
m
pa
r
a
t
iv
e
po
s
it
io
nin
g
wit
hin
I
C
T
-
ba
s
ed
i
ncub
a
t
io
n lit
er
a
t
ure
T
ab
le
4
co
m
p
ar
es
AI
-
SIM
’
s
class
if
icatio
n
p
er
f
o
r
m
an
ce
wi
th
two
in
d
ep
en
d
en
tly
p
u
b
lis
h
ed
,
m
o
r
e
r
ec
en
t
s
tu
d
ies
(
2
0
2
6
)
th
at
also
ap
p
ly
r
a
n
d
o
m
f
o
r
est
to
s
u
r
v
ey
-
d
e
r
iv
ed
MSM
E
in
d
icato
r
s
f
o
r
a
s
tr
u
ctu
r
ally
s
im
ilar
th
r
ee
-
way
r
ea
d
in
ess
/d
i
g
italizatio
n
class
if
icatio
n
task
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Dev
elo
p
men
t o
f a
n
A
I
-
b
a
s
ed
d
ig
ita
l sta
r
tu
p
in
cu
b
a
tio
n
mo
d
e
l fo
r
th
e
emp
o
w
ermen
t o
f
… (
Uma
r
B
a
kti
)
1691
F
ig
u
r
e
9
.
T
h
e
AI
-
SIM
d
ash
b
o
ar
d
(
p
e
r
ce
n
tag
es a
s
in
Fig
u
r
es
6
an
d
7
)
T
ab
le
4
.
Qu
a
n
titativ
e
B
en
ch
m
ar
k
in
g
o
f
AI
-
SIM
a
g
ain
s
t r
ec
e
n
t (
2
0
2
6
)
I
C
T
-
b
ased
MSM
E
d
ig
ital
-
r
ea
d
in
ess
/d
ig
italizatio
n
class
if
icatio
n
s
tu
d
ies
S
t
u
d
y
C
l
a
s
si
f
i
c
a
t
i
o
n
t
a
sk
S
a
mp
l
e
(
Tr
a
i
n
/
Te
st
)
R
e
p
o
r
t
e
d
a
c
c
u
r
a
c
y
AI
-
S
I
M
–
P
i
p
e
l
i
n
e
1
(
t
h
i
s
st
u
d
y
)
M
S
M
E
d
i
g
i
t
a
l
r
e
a
d
i
n
e
ss,
3
c
l
a
sses
(
r
a
n
d
o
m f
o
r
e
s
t
)
7
6
8
/
1
9
3
3
3
.
6
8
%
AI
-
S
I
M
–
P
i
p
e
l
i
n
e
2
(
t
h
i
s
st
u
d
y
)
M
S
M
E
d
i
g
i
t
a
l
r
e
a
d
i
n
e
ss,
3
c
l
a
sses
(
r
a
n
d
o
m f
o
r
e
s
t
)
7
6
8
/
1
9
3
3
0
.
5
7
%
R
o
y
a
n
i
e
t
a
l
.
[
2
8
]
M
S
M
E
d
i
g
i
t
a
l
i
z
a
t
i
o
n
l
e
v
e
l
,
r
a
n
d
o
m f
o
r
e
st
v
s.
C
4
.
5
1
0
0
r
e
sp
o
n
d
e
n
t
s (T
e
g
a
l
R
e
g
e
n
c
y
,
I
n
d
o
n
e
si
a
)
9
1
.
4
3
%
R
i
a
n
s
y
a
h
e
t
a
l
.
[
2
9
]
M
S
M
E
d
i
g
i
t
a
l
-
t
r
a
n
sf
o
r
m
a
t
i
o
n
r
e
a
d
i
n
e
ss,
3
l
e
v
e
l
s
(
r
a
n
d
o
m f
o
r
e
st
)
H
o
l
d
-
o
u
t
t
e
st
s
e
t
,
9
3
r
e
c
o
r
d
s
9
1
.
4
0
%
Po
s
itio
n
ed
ag
ain
s
t
th
ese
s
tu
d
ies,
AI
-
SIM
’
s
n
ea
r
-
ch
a
n
ce
ac
cu
r
ac
y
is
m
a
r
k
ed
ly
lo
wer
.
T
h
is
g
ap
is
co
n
s
is
ten
t
with
Sectio
n
3
.
3
’
s
f
in
d
in
g
t
h
at
AI
-
SIM
’
s
p
r
e
d
ictiv
e
weig
h
t
is
s
p
r
ea
d
th
i
n
ly
ac
r
o
s
s
co
ar
s
e,
s
elf
-
r
ep
o
r
ted
s
u
r
v
ey
c
o
n
s
tr
u
c
ts
,
with
n
o
d
o
m
in
a
n
t
f
ea
tu
r
e
(
th
e
lar
g
est
im
p
o
r
ta
n
ce
is
o
n
ly
0
.
2
2
)
,
wh
e
r
ea
s
R
o
y
an
i
et
a
l
.
[
2
8
]
an
d
R
ian
s
y
ah
et
a
l
.
[
2
9
]
b
o
t
h
r
e
p
o
r
t
r
an
d
o
m
f
o
r
est
ex
ce
e
d
in
g
9
0
%
ac
c
u
r
ac
y
w
h
en
f
ea
tu
r
es
m
o
r
e
d
ir
ec
tly
ca
p
tu
r
e
o
r
g
an
iz
atio
n
al
o
r
tr
an
s
ac
tio
n
al
ch
ar
ac
ter
is
tics
.
T
h
i
s
co
m
p
ar
is
o
n
r
ef
r
am
es
th
e
lim
itatio
n
id
en
tifie
d
in
Sectio
n
3
.
4
as
p
r
im
ar
ily
a
f
ea
t
u
r
e
-
en
g
in
ee
r
in
g
g
ap
in
th
e
cu
r
r
en
t
s
u
r
v
ey
i
n
s
tr
u
m
en
t
r
ath
er
th
a
n
a
s
h
o
r
tco
m
in
g
o
f
th
e
r
a
n
d
o
m
f
o
r
est
alg
o
r
ith
m
its
elf
,
an
d
r
ein
f
o
r
ce
s
th
e
m
iti
g
atio
n
alr
e
ad
y
p
r
o
p
o
s
ed
th
er
e
r
ep
lacin
g
c
o
ar
s
e
s
elf
-
r
ep
o
r
ted
p
r
o
x
ies
with
o
b
jectiv
e,
s
y
s
tem
-
lo
g
g
e
d
in
d
icato
r
s
.
T
h
e
d
ataset
was
co
llected
as
a
cr
o
s
s
-
s
ec
tio
n
al
s
u
r
v
ey
r
at
h
er
th
an
a
lo
n
g
itu
d
in
al
ev
al
u
atio
n
,
s
o
p
r
e
-
p
o
s
t
r
ea
d
i
n
ess
co
m
p
ar
is
o
n
s
an
d
ca
teg
o
r
y
-
tr
a
n
s
itio
n
r
ates
wer
e
n
o
t
ca
p
t
u
r
ed
.
T
h
is
lim
itatio
n
is
ad
d
r
ess
ed
in
Sectio
n
3
.
4
,
with
lo
n
g
itu
d
in
al
tr
ac
k
in
g
r
ec
o
m
m
en
d
ed
f
o
r
f
u
t
u
r
e
AI
-
SIM
iter
atio
n
s
.
3
.
4
.
Da
t
a
a
nd
m
o
del
lim
it
a
t
io
ns
Fiv
e
in
ter
r
elate
d
lim
itatio
n
s
co
n
s
tr
ain
th
e
p
r
ed
ictiv
e
v
alid
it
y
an
d
p
r
ac
tical
-
im
p
ac
t
ass
ess
m
en
t
o
f
th
is
s
tu
d
y
;
ea
ch
is
p
air
ed
b
elo
w
with
a
co
n
cr
ete
I
C
T
-
o
r
ie
n
ted
m
itig
atio
n
.
Sam
p
le
s
ize
af
ter
d
ed
u
p
licatio
n
:
th
e
two
-
lev
el
d
ed
u
p
licatio
n
n
e
ed
ed
to
p
r
ev
en
t
d
ata
leak
ag
e
r
ed
u
ce
d
th
e
s
am
p
le
f
r
o
m
4
,
2
7
6
to
9
6
1
r
ec
o
r
d
s
,
co
n
s
tr
ain
in
g
s
tatis
tical
p
o
wer
.
Mitig
atio
n
:
ex
p
an
d
d
ata
co
lle
ctio
n
ac
r
o
s
s
co
h
o
r
ts
an
d
p
r
o
v
i
n
ce
s
,
an
d
co
n
s
id
er
p
r
iv
ac
y
-
p
r
eser
v
in
g
f
ed
er
ated
lear
n
in
g
ac
r
o
s
s
p
ar
tn
er
in
s
titu
tio
n
s
to
p
o
o
l
s
am
p
l
e
s
izes
with
o
u
t
ce
n
tr
alizin
g
s
en
s
itiv
e
r
ec
o
r
d
s
.
Pro
x
y
co
n
s
tr
u
ct
v
alid
atio
n
a
n
d
s
ec
to
r
d
iv
er
s
ity
:
th
e
f
iv
e
co
n
s
tr
u
cts
wer
e
o
p
er
atio
n
alize
d
f
r
o
m
av
ailab
le
s
u
r
v
ey
item
s
r
ath
er
th
an
v
alid
ated
L
ik
er
t
-
s
ca
le
in
s
tr
u
m
en
ts
(
e.
g
.
,
I
T
s
k
ills
ap
p
r
o
x
im
ated
v
ia
AI
p
r
o
f
ic
ien
c
y
)
,
a
n
d
i
n
d
iv
id
u
ally
ca
r
r
y
o
n
l
y
m
o
d
est
p
r
e
d
ictiv
e
weig
h
t
(
0
.
0
8
–
0
.
2
2
)
(
Sectio
n
3
.
3
)
;
s
ec
to
r
-
le
v
el
d
iv
er
s
ity
was
also
n
o
t
s
ep
ar
at
ely
en
c
o
d
ed
.
Mitig
atio
n
:
r
ed
e
s
ig
n
th
e
s
u
r
v
ey
ar
o
u
n
d
v
alid
ated
L
ik
e
r
t
item
s
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.