I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
,
p
p
.
4
9
1
9
~
493
2
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
5
.
p
p
4
9
1
9
-
493
2
4919
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
G
enera
tive a
rtifi
c
ia
l int
ellig
ence
pe
rso
na
liza
tion a
nd cons
umer
beha
v
io
r ana
ly
tics
E
nd
a
ng
T
j
a
hja
nin
g
s
ih
1
,
E
le
n P
us
pita
s
a
ri
2
,
F
elix
Andrea
s
Su
t
a
nto
3
,
Dew
i H
a
nd
a
y
a
ni Unt
a
ri
Ning
s
ih
3
,
Dwi
B
u
di Sa
nto
s
o
4
1
D
e
p
a
r
t
me
n
t
o
f
M
a
n
a
g
e
me
n
t
,
F
a
c
u
l
t
y
o
f
Ec
o
n
o
mi
c
s a
n
d
B
u
s
i
n
e
ss,
U
n
i
v
e
r
si
t
a
s St
i
k
u
b
a
n
k
,
S
e
mara
n
g
,
I
n
d
o
n
e
s
i
a
2
D
e
p
a
r
t
me
n
t
o
f
A
c
c
o
u
n
t
i
n
g
,
F
a
c
u
l
t
y
o
f
Ec
o
n
o
mi
c
s
a
n
d
B
u
si
n
e
ss
,
U
n
i
v
e
r
si
t
a
s St
i
k
u
b
a
n
k
,
S
e
m
a
r
a
n
g
,
I
n
d
o
n
e
si
a
3
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
I
n
f
o
r
ma
t
i
o
n
T
e
c
h
n
o
l
o
g
y
a
n
d
I
n
d
u
st
r
y
,
U
n
i
v
e
r
si
t
a
s S
t
i
k
u
b
a
n
k
,
S
e
m
a
r
a
n
g
,
I
n
d
o
n
e
si
a
4
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
o
n
S
y
s
t
e
m,
F
a
c
u
l
t
y
o
f
I
n
f
o
r
ma
t
i
o
n
T
e
c
h
n
o
l
o
g
y
a
n
d
I
n
d
u
st
r
y
,
U
n
i
v
e
r
si
t
a
s S
t
i
k
u
b
a
n
k
,
S
e
maran
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
Sep
3
0
,
2
0
2
5
R
ev
is
ed
J
u
l 2
8
,
2
0
2
6
Acc
ep
ted
Au
g
1
2
,
2
0
2
6
Ba
ti
k
m
icro
,
sm
a
ll
a
n
d
m
e
d
i
u
m
e
n
terp
rise
s
(
M
S
M
Es)
su
sta
in
I
n
d
o
n
e
sia
'
s
tex
ti
le
h
e
rit
a
g
e
a
n
d
l
o
c
a
l
li
v
e
li
h
o
o
d
s
b
u
t
stru
g
g
le
wit
h
c
u
sto
m
e
r
re
ten
ti
o
n
d
u
e
t
o
c
o
n
v
e
n
ti
o
n
a
l
m
a
rk
e
ti
n
g
a
n
d
li
m
it
e
d
a
rti
ficia
l
i
n
telli
g
e
n
c
e
(AI)
a
d
o
p
t
io
n
.
T
h
e
sh
ift
t
o
d
ig
i
tal
c
o
m
m
e
rc
e
re
q
u
ires
a
ffo
rd
a
b
le,
c
u
lt
u
ra
ll
y
re
lev
a
n
t
AI
stra
teg
ies
.
Th
is
stu
d
y
p
r
o
p
o
se
s
a
n
in
teg
ra
te
d
fr
a
m
e
wo
rk
c
o
m
b
in
i
n
g
re
c
e
n
c
y
,
fre
q
u
e
n
c
y
,
m
o
n
e
tary
(RF
M
)
a
n
a
ly
ti
c
s,
K
-
m
e
a
n
s
c
lu
ste
rin
g
(K
=
5
)
,
ra
n
d
o
m
fo
re
st
c
h
u
rn
p
re
d
icti
o
n
,
a
n
d
tran
sf
o
r
m
e
r
-
b
a
se
d
g
e
n
e
ra
ti
v
e
AI
fo
r
p
e
rso
n
a
li
z
e
d
p
r
o
m
o
ti
o
n
s
d
e
li
v
e
re
d
v
ia
W
h
a
tsAp
p
b
u
si
n
e
ss
API.
Ev
a
lu
a
ted
o
v
e
r
1
2
m
o
n
th
s
with
2
0
b
a
ti
k
M
S
M
Es
i
n
Ce
n
tral
Ja
v
a
(3
0
0
c
u
sto
m
e
rs,
8
7
3
tran
sa
c
ti
o
n
s),
t
h
e
fra
m
e
wo
rk
e
m
p
lo
y
e
d
tem
p
o
ra
l
d
a
ta
sp
li
tt
i
n
g
(9
m
o
n
t
h
s
trai
n
in
g
,
3
m
o
n
th
s
tes
ti
n
g
)
a
n
d
A/B
tes
ti
n
g
a
g
a
i
n
st
g
e
n
e
ric
c
a
m
p
a
ig
n
s.
Hu
m
a
n
e
v
a
lu
a
ti
o
n
a
ss
e
ss
e
d
m
e
ss
a
g
e
re
lev
a
n
c
e
a
n
d
c
u
lt
u
ra
l
a
p
p
ro
p
riate
n
e
ss
,
wh
i
le
a
ll
d
a
ta
we
re
a
n
o
n
y
m
ize
d
with
in
fo
rm
e
d
c
o
n
se
n
t.
Re
su
lt
s
sh
o
w
im
p
ro
v
e
m
e
n
ts
o
f
+
1
7
%
i
n
re
ten
ti
o
n
p
re
d
ictio
n
a
c
c
u
ra
c
y
,
+
2
3
%
in
c
li
c
k
-
t
h
ro
u
g
h
ra
te,
a
n
d
+
2
1
%
i
n
p
u
rc
h
a
se
fre
q
u
e
n
c
y
v
e
rsu
s
b
a
se
li
n
e
s
(p
<
0
.
0
5
,
C
o
h
e
n
'
s
d
=
0
.
6
8
,
p
o
we
r
>
0
.
8
)
.
T
h
e
s
y
ste
m
c
o
sts
b
e
lo
w
USD
2
0
p
e
r
M
S
M
E
p
e
r
m
o
n
th
,
o
ffe
ri
n
g
a
sc
a
lab
le,
h
u
m
a
n
-
c
e
n
tere
d
p
a
th
wa
y
t
o
stre
n
g
th
e
n
l
o
y
a
lt
y
in
e
m
e
rg
in
g
m
a
rk
e
ts
.
K
ey
w
o
r
d
s
:
C
u
s
to
m
er
r
eten
tio
n
E
m
er
g
in
g
m
ar
k
ets
Gen
AI
p
er
s
o
n
aliza
tio
n
K
-
m
ea
n
s
clu
s
ter
in
g
R
an
d
o
m
f
o
r
est
R
FM
an
aly
tics
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
:
Dwi
B
u
d
i San
to
s
o
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
Sy
s
tem
,
Facu
lty
o
f
I
n
f
o
r
m
atio
n
T
ec
h
n
o
lo
g
y
an
d
I
n
d
u
s
tr
y
,
U
n
iv
er
s
itas
Sti
k
u
b
an
k
Ken
d
en
g
V,
B
en
d
an
Ng
is
o
r
,
Sem
ar
an
g
,
C
en
tr
al
J
av
a,
I
n
d
o
n
e
s
ia
E
m
ail:
d
b
s
@
ed
u
.
u
n
is
b
an
k
.
ac
.
i
d
1.
I
NT
RO
D
UCT
I
O
N
B
ey
o
n
d
th
eir
ec
o
n
o
m
ic
c
o
n
tr
i
b
u
tio
n
,
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
p
r
eser
v
e
cu
ltu
r
al
h
e
r
itag
e
an
d
s
u
p
p
o
r
t
lo
ca
l
liv
elih
o
o
d
s
[
1
]
.
Ho
wev
er
,
t
h
ese
en
ter
p
r
is
es
f
ac
e
ch
allen
g
es
in
r
etain
in
g
cu
s
to
m
er
s
d
u
e
t
o
r
elian
ce
o
n
co
n
v
en
tio
n
al
m
a
r
k
etin
g
,
wh
ile
th
e
s
h
if
t
to
d
ig
ital
co
m
m
er
ce
h
ig
h
lig
h
ts
th
e
u
r
g
en
t
n
ee
d
f
o
r
ar
tific
ial
in
tellig
en
ce
(
AI
)
-
d
r
iv
en
s
tr
ateg
ies
to
e
n
h
an
ce
co
m
p
etitiv
en
ess
.
W
ith
co
n
s
u
m
er
b
e
h
av
io
r
s
in
cr
ea
s
in
g
ly
s
h
ap
e
d
b
y
e
-
co
m
m
er
ce
p
latf
o
r
m
s
,
ad
o
p
tin
g
AI
-
b
ased
m
ar
k
etin
g
s
tr
ateg
ies
h
as
b
ec
o
m
e
ess
en
tial
f
o
r
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es to
r
em
ain
s
u
s
tain
ab
le
in
a
r
ap
i
d
ly
ev
o
l
v
in
g
m
a
r
k
et
[
2
]
,
[
3
]
.
T
h
e
co
r
e
p
r
o
b
lem
ad
d
r
ess
ed
in
t
h
is
s
tu
d
y
is
t
h
e
a
b
s
en
ce
o
f
AI
-
d
r
iv
en
p
er
s
o
n
aliza
tio
n
in
th
e
m
ar
k
etin
g
s
tr
ateg
ies
o
f
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es,
wh
ich
lead
s
to
wea
k
cu
s
to
m
er
en
g
ag
em
e
n
t
an
d
lo
w
lo
y
alty
[
4
]
.
T
h
is
ch
allen
g
e
is
am
p
lifie
d
b
y
ch
an
g
in
g
c
o
n
s
u
m
er
e
x
p
ec
tatio
n
s
:
s
u
r
v
e
y
s
r
ev
ea
l
th
at
9
1
%
o
f
cu
s
to
m
er
s
p
r
ef
er
b
r
a
n
d
s
o
f
f
er
i
n
g
tailo
r
ed
p
r
o
m
o
tio
n
s
,
an
d
8
5
%
ar
e
m
o
r
e
lik
ely
to
p
u
r
ch
as
e
wh
en
ex
p
o
s
ed
to
in
d
iv
id
u
alize
d
r
ec
o
m
m
e
n
d
atio
n
s
[
5
]
.
Ho
we
v
er
,
m
o
s
t
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
lack
th
e
tech
n
o
l
o
g
ical
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
9
1
9
-
493
2
4920
ca
p
ab
ilit
y
an
d
e
x
p
er
tis
e
to
i
m
p
lem
en
t
s
u
c
h
p
e
r
s
o
n
aliza
tio
n
[
6
]
,
[
7
]
,
c
r
ea
tin
g
a
p
r
ess
in
g
d
ig
ital
d
iv
i
d
e
in
em
er
g
in
g
m
ar
k
ets.
C
u
s
to
m
er
s
eg
m
en
tatio
n
was
p
er
f
o
r
m
e
d
u
s
in
g
r
ec
en
c
y
,
f
r
eq
u
en
cy
,
an
d
m
o
n
etar
y
(
R
FM)
an
aly
s
is
[
8
]
an
d
k
-
m
ea
n
s
clu
s
ter
in
g
[
9
]
,
[
1
0
]
,
wh
ile
c
u
s
to
m
er
ch
u
r
n
was
p
r
e
d
icted
u
s
in
g
th
e
r
an
d
o
m
f
o
r
est
alg
o
r
ith
m
[
1
1
]
.
T
h
e
r
esu
ltin
g
i
n
s
ig
h
ts
wer
e
in
teg
r
ated
with
a
g
en
er
ativ
e
a
r
tific
ial
in
tellig
en
ce
(
Gen
AI
)
m
o
d
el
to
p
r
o
d
u
ce
p
e
r
s
o
n
alize
d
p
r
o
m
o
tio
n
al
m
ess
ag
es a
n
d
p
r
o
d
u
ct
r
ec
o
m
m
en
d
atio
n
s
[
1
2
]
.
E
m
p
ir
ically
,
t
h
e
s
tu
d
y
d
r
aws
o
n
p
r
im
ar
y
d
ata
f
r
o
m
2
0
b
ati
k
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
l
o
ca
ted
in
So
lo
,
Pek
alo
n
g
a
n
,
a
n
d
B
ak
ar
an
(
Pati),
in
v
o
lv
in
g
3
0
0
r
esp
o
n
d
en
ts
ac
r
o
s
s
an
eig
h
t
-
m
o
n
th
im
p
lem
en
tatio
n
p
er
io
d
.
T
h
e
s
y
s
tem
’
s
ef
f
ec
tiv
en
ess
is
ev
alu
ated
u
s
in
g
p
r
ed
ictiv
e
r
eten
tio
n
ac
cu
r
ac
y
,
click
-
th
r
o
u
g
h
r
ate
(
C
T
R
)
[
1
3
]
,
an
d
p
u
r
ch
ase
f
r
eq
u
en
c
y
as
p
er
f
o
r
m
an
ce
in
d
icato
r
s
[
1
4
]
.
E
x
is
tin
g
s
tu
d
i
es
h
av
e
ex
p
lo
r
ed
d
ig
italizatio
n
f
o
r
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es,
f
o
cu
s
in
g
p
r
im
ar
ily
o
n
s
k
ill
d
ev
elo
p
m
en
t,
cr
ea
tiv
e
d
esig
n
,
o
r
g
en
er
ic
o
n
lin
e
p
r
o
m
o
tio
n
[
3
]
,
[
1
5
]
.
Oth
er
wo
r
k
s
ap
p
lied
ch
u
r
n
p
r
ed
ictio
n
o
r
cu
s
to
m
er
s
eg
m
en
tatio
n
in
d
ep
en
d
en
tly
.
Yet,
th
ese
ap
p
r
o
ac
h
es
d
id
n
o
t
i
n
teg
r
ate
a
d
v
an
ce
d
p
er
s
o
n
aliza
tio
n
tech
n
iq
u
es
o
r
lev
er
a
g
e
Gen
AI
[
1
6
]
.
Un
lik
e
p
r
io
r
r
ese
ar
ch
th
at
tar
g
ete
d
lar
g
e
-
s
ca
le
en
ter
p
r
is
es
in
d
ev
elo
p
ed
co
n
te
x
ts
[
1
7
]
,
th
is
s
tu
d
y
em
p
h
asizes
cu
ltu
r
ally
g
r
o
u
n
d
ed
an
d
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
[
1
8
]
,
h
ig
h
lig
h
tin
g
a
g
a
p
in
lo
ca
lized
AI
ad
o
p
tio
n
[
1
9
]
.
T
h
is
s
tu
d
y
a
d
d
r
ess
es
th
e
g
a
p
in
th
e
liter
atu
r
e
o
n
th
e
ap
p
l
icatio
n
o
f
AI
-
d
r
iv
en
p
er
s
o
n
aliza
tio
n
in
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
in
em
er
g
in
g
m
ar
k
ets
[
2
0
]
.
W
h
ile
m
an
y
s
tu
d
ies
f
o
cu
s
o
n
s
m
all
b
u
s
in
es
s
en
ter
p
r
is
es
d
ig
italizatio
n
,
m
o
s
t
em
p
h
asize
co
n
v
en
tio
n
al
m
a
r
k
etin
g
o
r
lar
g
e
-
s
ca
le
tech
n
o
l
o
g
y
u
s
e
[
2
1
]
,
[
2
2
]
.
T
h
e
m
ain
g
ap
is
th
e
in
ab
ilit
y
o
f
m
an
y
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
to
ad
o
p
t
ad
v
a
n
ce
d
tec
h
n
o
lo
g
ies
lik
e
Ge
n
AI
af
f
o
r
d
a
b
ly
[
2
3
]
.
T
h
is
s
tu
d
y
p
r
o
p
o
s
es
an
in
teg
r
ated
s
o
lu
tio
n
co
m
b
in
i
n
g
b
e
h
av
io
r
al
an
aly
tics
,
AI
p
er
s
o
n
aliza
tio
n
,
an
d
W
h
atsAp
p
b
u
s
in
ess
ap
p
licatio
n
p
r
o
g
r
am
m
in
g
in
ter
f
ac
e
(
API
)
[
2
4
]
–
[
2
6
]
to
en
h
a
n
ce
cu
s
to
m
er
lo
y
alty
a
n
d
c
o
m
p
etit
iv
en
ess
in
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
[
2
7
]
.
An
in
teg
r
ated
m
ar
k
etin
g
f
r
am
ewo
r
k
is
in
tr
o
d
u
ce
d
b
y
in
c
o
r
p
o
r
atin
g
Ge
n
AI
in
to
m
ar
k
etin
g
s
tr
ateg
ies
f
o
r
lo
ca
l
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
[
2
8
]
,
co
m
b
in
in
g
co
n
s
u
m
er
an
aly
tics
with
a
d
ap
tiv
e
p
er
s
o
n
aliza
tio
n
[
2
9
]
,
an
d
ap
p
ly
in
g
h
u
m
a
n
-
ce
n
te
r
ed
AI
p
r
in
cip
les
to
im
p
r
o
v
e
u
s
a
b
ilit
y
wh
ile
m
ain
tain
in
g
cu
ltu
r
al
r
elev
an
ce
[
3
0
]
.
B
y
o
f
f
er
in
g
an
af
f
o
r
d
a
b
le
an
d
s
ca
lab
le
s
o
lu
tio
n
,
th
is
s
tu
d
y
p
r
o
v
id
es
an
alter
n
ativ
e
a
p
p
r
o
ac
h
th
at
en
r
ich
es
ex
is
tin
g
m
eth
o
d
s
o
f
AI
-
en
a
b
l
ed
m
ar
k
etin
g
in
em
e
r
g
in
g
m
a
r
k
ets
[
3
1
]
.
T
h
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
b
r
o
a
d
en
s
th
e
s
co
p
e
o
f
AI
-
d
r
iv
en
co
n
s
u
m
e
r
b
eh
av
io
r
an
aly
tics
[
3
2
]
b
y
tr
an
s
latin
g
an
aly
tical
in
s
ig
h
ts
in
to
s
u
s
tain
ab
le
cu
s
to
m
er
lo
y
alty
p
r
ac
tices f
o
r
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es.
2.
M
E
T
H
O
D
2
.
1
.
Resea
rc
h desi
g
n
T
h
is
m
eth
o
d
o
lo
g
y
in
v
o
lv
es
a
th
r
ee
-
s
tag
e
p
r
o
ce
s
s
.
First,
in
p
u
t
d
ata
is
co
llected
f
r
o
m
b
a
tik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
th
r
o
u
g
h
tr
an
s
ac
tio
n
lo
g
s
,
co
n
s
u
m
er
s
u
r
v
ey
s
,
a
n
d
s
o
cial
m
ed
ia
in
ter
ac
tio
n
s
[
3
3
]
.
Pre
-
p
r
o
ce
s
s
in
g
in
clu
d
es d
ata
c
lean
in
g
,
n
o
r
m
aliza
tio
n
,
an
d
ex
p
lo
r
ato
r
y
an
aly
s
is
.
Seco
n
d
,
co
n
s
u
m
er
b
eh
a
v
io
r
is
an
aly
ze
d
u
s
in
g
th
e
R
FM
f
ea
tu
r
e
[
2
9
]
,
s
eg
m
e
n
ted
with
k
-
m
e
an
s
(K
=
5
)
[
3
4
]
,
[
3
5
]
an
d
ch
u
r
n
r
is
k
is
p
r
ed
icted
with
r
an
d
o
m
f
o
r
est
(
1
0
0
tr
ee
s
,
1
0
-
f
o
ld
cu
s
to
m
er
life
tim
e
v
a
lu
e
o
r
C
L
V)
.
T
h
ir
d
,
h
ig
h
-
p
r
io
r
ity
cu
s
to
m
er
s
ar
e
id
en
tifie
d
u
s
in
g
a
c
o
m
b
in
ed
c
h
u
r
n
-
v
alu
e
s
co
r
e,
f
o
llo
wed
b
y
Gen
AI
m
ess
ag
e
g
en
er
atio
n
(
t
r
an
s
f
o
r
m
er
-
b
ased
,
en
h
an
ce
d
with
b
atik
d
o
m
ai
n
p
r
o
m
p
ts
)
,
au
to
m
atic
d
eliv
er
y
v
ia
W
h
atsAp
p
b
u
s
in
ess
A
PI,
an
d
ev
alu
atio
n
f
o
cu
s
es
o
n
m
etr
ics
d
i
r
ec
tly
lin
k
ed
to
th
e
im
p
lem
e
n
ted
m
eth
o
d
s
.
Fo
r
ch
u
r
n
p
r
ed
ictio
n
(
r
an
d
o
m
f
o
r
est
)
,
ac
cu
r
ac
y
,
a
n
d
F1
-
s
co
r
e
ar
e
u
s
ed
as
p
r
im
ar
y
in
d
icato
r
s
.
Fo
r
Gen
AI
p
er
s
o
n
aliza
tio
n
,
ca
m
p
a
ig
n
ef
f
ec
tiv
e
n
ess
is
ass
es
s
ed
th
r
o
u
g
h
C
T
R
an
d
A/B
test
in
g
ag
ain
s
t
b
aselin
e
g
e
n
er
ic
ca
m
p
aig
n
s
,
u
s
in
g
p
u
r
ch
ase
f
r
eq
u
en
c
y
an
d
r
eten
tio
n
u
p
lift
as
k
ey
co
m
p
ar
ativ
e
m
etr
ics.
T
h
e
b
e
h
av
io
r
al
an
aly
s
is
co
n
s
id
er
ed
th
r
ee
p
r
im
ar
y
asp
ec
ts
as
s
h
o
wn
in
Fig
u
r
e
1
:
t
h
e
g
eo
g
r
ap
h
ical
d
is
tr
ib
u
tio
n
o
f
p
ar
ti
cip
atin
g
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
ac
r
o
s
s
B
ak
ar
an
Pati,
So
lo
,
an
d
Pek
alo
n
g
an
as
in
Fig
u
r
e
1
(
a
)
,
cu
s
to
m
er
ag
e
ch
ar
ac
ter
is
tics
as
i
n
Fig
u
r
e
1
(
b
)
,
an
d
tr
an
s
ac
tio
n
v
alu
e
p
atter
n
s
as in
Fig
u
r
e
1
(
c)
.
2
.
2
.
S
y
s
t
e
m
f
l
o
w
G
e
n
A
I
p
e
r
s
o
n
a
l
i
z
a
t
i
o
n
a
n
d
c
o
n
s
u
m
e
r
b
e
h
a
v
i
o
r
a
n
a
l
y
t
i
c
s
e
n
h
a
n
c
e
b
a
t
i
k
c
u
s
t
o
m
e
r
l
o
y
a
l
t
y
T
h
i
s
s
y
s
t
em
f
l
o
w
i
l
l
u
s
t
r
a
t
es
t
h
e
i
n
t
eg
r
a
t
io
n
o
f
G
e
n
AI
,
c
o
n
s
u
m
er
b
e
h
a
v
io
r
a
n
a
ly
t
i
c
s
,
a
n
d
a
u
t
o
m
a
t
e
d
c
a
m
p
a
i
g
n
o
r
c
h
e
s
tr
a
t
i
o
n
to
p
er
s
o
n
a
l
i
z
e
cu
s
t
o
m
er
e
x
p
er
i
e
n
c
e
a
n
d
i
n
c
r
e
a
s
e
b
a
t
ik
c
o
n
s
u
m
er
l
o
y
a
l
ty
(
F
i
g
u
r
e
2
)
.
T
h
e
f
l
o
w
co
n
n
ec
t
s
u
p
s
t
r
e
am
b
e
h
a
v
io
r
a
l
s
e
g
m
e
n
t
a
t
io
n
w
i
t
h
d
o
wn
s
t
r
e
am
a
d
a
p
t
iv
e
c
o
n
t
e
n
t
g
e
n
er
a
t
i
o
n
a
n
d
d
e
l
i
v
er
y
,
en
s
u
r
i
n
g
th
a
t
ea
c
h
c
u
s
to
m
e
r
r
e
c
e
i
v
e
s
c
o
n
t
e
x
tu
a
l
l
y
r
e
le
v
an
t
e
n
g
a
g
em
e
n
t.
B
y
b
r
i
d
g
i
n
g
a
n
a
ly
t
i
c
o
u
tp
u
t
s
w
i
th
g
e
n
er
a
t
i
v
e
c
ap
ab
i
l
i
t
i
e
s
,
th
e
s
y
s
t
e
m
en
a
b
le
s
c
o
n
t
i
n
u
o
u
s
,
cu
l
t
u
r
a
l
l
y
g
r
o
u
n
d
e
d
p
e
r
s
o
n
a
l
i
z
a
t
io
n
a
t
s
c
a
l
e.
2
.
3
.
P
re
-
pro
ce
s
s
ing
Pre
-
p
r
o
ce
s
s
in
g
was
p
er
f
o
r
m
e
d
to
e
n
h
an
ce
d
ata
q
u
ality
an
d
en
s
u
r
e
co
m
p
atib
ilit
y
with
s
u
b
s
eq
u
en
t
an
aly
tical
p
r
o
ce
d
u
r
es
[
3
6
]
.
Fig
u
r
e
3
s
u
m
m
ar
izes
t
h
e
s
eq
u
en
ce
o
f
o
p
e
r
atio
n
s
ap
p
li
ed
b
ef
o
r
e
m
o
d
el
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Gen
era
tive
a
r
tifi
cia
l in
tellig
en
ce
p
ers
o
n
a
liz
a
tio
n
a
n
d
co
n
s
u
mer b
eh
a
vio
r
…
(
E
n
d
a
n
g
Tja
h
ja
n
in
g
s
ih
)
4921
im
p
lem
en
tatio
n
,
n
am
ely
d
ata
clea
n
in
g
[
3
7
]
,
n
u
m
er
ical
n
o
r
m
aliza
tio
n
[
3
8
]
,
an
d
ex
p
l
o
r
ato
r
y
d
ata
an
aly
s
is
(
E
DA)
f
o
r
b
eh
av
i
o
r
p
atter
n
s
.
‒
Data
clea
n
in
g
:
t
o
im
p
r
o
v
e
d
ataset
r
eliab
ilit
y
,
d
u
p
licate
r
ec
o
r
d
s
wer
e
elim
in
ated
,
a
n
d
in
co
m
p
lete
o
b
s
er
v
atio
n
s
wer
e
s
y
s
tem
atica
lly
r
eso
lv
ed
.
Statis
tical
im
p
u
tatio
n
(
m
ea
n
,
m
e
d
ian
,
o
r
m
o
d
e)
was
em
p
lo
y
ed
f
o
r
r
ec
o
v
er
ab
le
m
is
s
in
g
v
alu
es,
wh
er
ea
s
r
ec
o
r
d
s
with
m
o
r
e
th
an
5
0
%
u
n
av
aila
b
le
in
f
o
r
m
atio
n
wer
e
o
m
itted
.
Fo
r
m
at
s
tan
d
ar
d
izatio
n
an
d
lo
g
ical
v
alid
atio
n
wer
e
s
u
b
s
eq
u
en
tly
p
e
r
f
o
r
m
ed
to
en
s
u
r
e
d
ata
co
n
s
is
ten
cy
,
in
clu
d
in
g
v
er
i
f
icatio
n
th
at
p
r
o
d
u
ct
p
r
ices r
em
ain
ed
with
in
v
alid
,
n
o
n
-
n
eg
ativ
e
r
an
g
es.
‒
Nu
m
er
ical
n
o
r
m
aliza
tio
n
:
n
u
m
er
ical
n
o
r
m
aliza
tio
n
is
ca
r
r
ied
o
u
t
s
o
th
at
n
u
m
er
ical
d
ata
is
o
n
a
u
n
if
o
r
m
s
ca
le,
wh
ich
will
f
ac
ilit
ate
th
e
ap
p
licatio
n
o
f
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
[
3
9
]
an
d
s
tatis
tical
an
aly
s
is
[
4
0
]
,
wh
ich
in
clu
d
es:
i
)
m
in
-
m
ax
s
ca
lin
g
[
4
1
]
,
t
h
is
is
u
s
ef
u
l
to
en
s
u
r
e
th
at
ea
ch
n
u
m
er
ica
l
f
ea
tu
r
e
h
as
a
co
m
p
ar
ab
le
in
f
l
u
en
ce
in
th
e
m
o
d
el
;
ii
)
s
tan
d
ar
d
izatio
n
(
z
-
s
co
r
e
n
o
r
m
aliza
tio
n
)
[
4
2
]
;
an
d
iii
)
lo
g
tr
an
s
f
o
r
m
atio
n
[
4
3
]
,
[
4
4
]
.
‒
E
DA
:
E
DA
f
o
r
b
eh
av
io
r
p
atter
n
[
4
5
]
is
o
b
tain
ed
b
as
ed
o
n
co
n
s
u
m
e
r
b
eh
av
io
r
p
atter
n
s
[
4
6
]
,
tr
an
s
ac
tio
n
s
,
an
d
s
o
cial
m
ed
ia
[
4
7
]
.
(
a)
(
b
)
(
c)
Fig
u
r
e
1
.
Data
an
al
y
s
is
o
f
b
ati
k
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es c
u
s
to
m
er
b
eh
a
v
io
r
: (
a
)
d
is
tr
ib
u
t
io
n
b
y
l
o
ca
tio
n
,
(
b
)
cu
s
to
m
e
r
ag
e
d
is
tr
ib
u
tio
n
,
an
d
(
c)
tr
an
s
ac
tio
n
v
al
u
e
d
is
tr
ib
u
tio
n
F
i
g
u
r
e
2
.
S
y
s
t
e
m
f
l
o
ws
G
e
n
A
I
p
e
r
s
o
n
a
l
i
z
at
i
o
n
a
n
d
c
o
n
s
u
m
e
r
b
e
h
a
v
i
o
r
a
n
a
l
y
t
i
cs
e
n
h
a
n
c
e
b
a
t
ik
c
u
s
t
o
m
e
r
l
o
y
a
l
t
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
9
1
9
-
493
2
4922
Fig
u
r
e
3
.
Pre
-
p
r
o
ce
s
s
in
g
s
tag
e
s
2
.
4
.
B
eha
v
io
r
a
na
ly
s
is
B
eh
av
io
r
an
aly
s
is
p
r
o
v
id
es
th
e
an
aly
tic
f
o
u
n
d
atio
n
f
o
r
th
e
p
er
s
o
n
aliza
tio
n
f
r
am
ew
o
r
k
.
I
t
is
co
n
d
u
cte
d
th
r
o
u
g
h
R
FM
f
ea
tu
r
e
en
g
in
ee
r
in
g
[
4
8
]
t
o
ex
tr
a
ct
R
FM
v
alu
es,
k
-
m
ea
n
s
clu
s
ter
in
g
[
4
9
]
to
f
o
r
m
m
u
ltid
im
en
s
io
n
al
cu
s
to
m
e
r
s
eg
m
en
tatio
n
,
an
d
r
an
d
o
m
f
o
r
es
t
class
if
icatio
n
[
5
0
]
to
p
r
e
d
ict
p
o
ten
tial
cu
s
to
m
er
ch
u
r
n
.
T
o
g
eth
er
,
th
ese
tech
n
i
q
u
es
s
u
p
p
o
r
t
d
ata
-
d
r
iv
e
n
r
ete
n
tio
n
an
d
p
er
s
o
n
aliza
tio
n
s
tr
ateg
ies
b
y
lin
k
in
g
b
eh
av
io
r
al
h
is
to
r
y
to
f
o
r
war
d
-
lo
o
k
in
g
r
is
k
esti
m
atio
n
[
5
1
]
.
2
.
4
.
1
.
RF
M
f
ea
t
ure
eng
ineering
a
nd
no
rm
a
liza
t
io
n
R
FM
f
ea
tu
r
e
en
g
in
ee
r
in
g
tr
an
s
f
o
r
m
s
r
aw
tr
an
s
ac
tio
n
al
h
is
to
r
ies
in
to
in
ter
p
r
etab
le
b
eh
av
io
r
a
l
d
im
en
s
io
n
s
.
T
h
is
s
u
b
s
ec
tio
n
d
escr
ib
es
h
o
w
r
aw
R
FM
v
alu
e
s
ar
e
n
o
r
m
alize
d
an
d
co
m
b
in
e
d
in
to
a
co
m
p
o
s
ite
s
co
r
e
to
en
a
b
le
f
air
c
o
m
p
a
r
is
o
n
ac
r
o
s
s
cu
s
to
m
er
s
.
E
ac
h
n
o
r
m
aliza
tio
n
s
tep
is
in
tr
o
d
u
ce
d
s
ep
ar
ately
b
elo
w
as
d
ef
in
ed
in
(
1
)
to
(
3
)
,
f
o
llo
win
g
estab
lis
h
ed
R
FM
p
r
ac
tices
[
5
2
]
.
i)
Min
-
m
ax
n
o
r
m
aliza
tio
n
(
f
o
r
F a
n
d
M)
:
=
−
(
)
(
)
−
(
)
,
{
,
]
(
1
)
w
h
er
e
r
ep
r
esen
ts
th
e
o
r
ig
in
a
l
v
alu
e,
′
th
e
n
o
r
m
alize
d
v
alu
e,
an
d
(
)
/
(
)
th
e
o
b
s
er
v
e
d
b
o
u
n
d
s
,
with
m
in
–
m
a
x
n
o
r
m
al
izatio
n
ap
p
lied
to
th
e
f
r
e
q
u
en
c
y
(
)
an
d
m
o
n
etar
y
(
)
v
ar
iab
les.
ii)
Fo
r
r
ec
en
cy
(
th
e
s
m
aller
th
e
b
etter
)
,
u
s
e
in
v
er
te
d
s
ca
lin
g
:
=
1
−
−
(
)
(
)
−
(
)
(
2
)
w
h
er
e
d
en
o
tes
th
e
n
o
r
m
alize
d
r
ec
en
cy
s
co
r
e;
is
th
e
tim
e
s
in
ce
th
e
last
p
u
r
ch
ase;
(
)
an
d
(
)
in
d
icate
th
e
d
ataset
b
o
u
n
d
s
iii)
W
eig
h
ted
R
FM
co
m
p
o
s
ite
s
co
r
e
:
RF
M
=
ꭣ
̃
+
ꭣ
̂
+
ꭣ
̂
,
ꭣ
+
ꭣ
+
ꭣ
=
1
(
3
)
w
h
er
e
RF
M
in
d
icate
s
th
e
weig
h
te
d
R
FM
s
co
r
e,
wh
er
e
n
o
r
m
ali
ze
d
̃
,
̂
,
an
d
̂
ar
e
co
m
b
in
ed
u
s
in
g
weig
h
ts
ꭣ
,
ꭣ
,
an
d
ꭣ
.
2
.
4
.
2
.
K
-
m
ea
ns
clus
t
er
ing
(
mu
lt
idi
m
ens
io
na
l seg
m
ent
a
t
io
n)
M
u
l
t
i
d
i
m
e
n
s
i
o
n
a
l
s
eg
m
en
t
a
ti
o
n
g
r
o
u
p
s
cu
s
t
o
m
er
s
w
i
th
s
i
m
i
l
ar
b
e
h
av
i
o
r
a
l
p
r
o
f
il
e
s
.
T
h
i
s
s
u
b
s
e
c
t
i
o
n
f
o
r
m
u
l
a
t
e
s
t
h
e
k
-
m
e
a
n
s
o
p
t
i
m
i
za
t
i
o
n
o
b
j
e
c
t
i
v
e
a
s
d
e
f
i
n
e
d
i
n
(
4
)
a
n
d
e
x
p
l
a
in
s
t
h
e
s
e
l
e
c
t
io
n
o
f
t
h
e
o
p
t
i
m
a
l
n
u
m
b
e
r
o
f
c
lu
s
t
e
r
s
u
s
i
n
g
th
e
e
lb
o
w
m
e
t
h
o
d
a
n
d
s
i
l
h
o
u
e
t
t
e
an
a
l
y
s
i
s
,
as
a
p
p
l
i
ed
i
n
p
r
i
o
r
c
l
u
s
t
e
r
in
g
s
tu
d
i
e
s
[
5
3
]
.
i)
K
-
m
ea
n
s
o
p
tim
izatio
n
g
o
al
:
∑
‖
−
(
)
‖
=
1
{
}
2
2
(
4
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Gen
era
tive
a
r
tifi
cia
l in
tellig
en
ce
p
ers
o
n
a
liz
a
tio
n
a
n
d
co
n
s
u
mer b
eh
a
vio
r
…
(
E
n
d
a
n
g
Tja
h
ja
n
in
g
s
ih
)
4923
w
h
er
e
is
th
e
to
tal
n
u
m
b
er
o
f
d
ata
p
o
in
ts
,
d
en
o
tes th
e
i
-
th
d
at
a
v
ec
to
r
,
is
th
e
ce
n
tr
o
id
o
f
clu
s
ter
,
(
)
r
ep
r
esen
ts
th
e
clu
s
ter
ass
ig
n
e
d
to
,
an
d
‖
−
(
)
‖
2
2
is
th
e
s
q
u
ar
e
d
E
u
cl
id
ea
n
d
is
tan
ce
b
etwe
en
an
d
its
ass
ig
n
ed
ce
n
tr
o
id
.
T
h
e
o
p
tim
izatio
n
g
o
al
is
to
m
in
i
m
ize
th
e
t
o
tal
s
q
u
ar
e
d
d
is
tan
ce
o
f
all
d
at
a
p
o
in
ts
f
r
o
m
th
eir
ass
ig
n
ed
clu
s
ter
ce
n
tr
o
id
.
ii)
Selectio
n
o
f
th
e
o
p
tim
al
n
u
m
b
er
o
f
clu
s
ter
s
:
t
h
e
n
u
m
b
e
r
o
f
clu
s
ter
s
is
d
eter
m
i
n
ed
u
s
in
g
th
e
el
b
o
w
m
eth
o
d
an
d
s
ilh
o
u
ette
a
n
aly
s
is
.
B
ased
o
n
t
h
ese
ev
alu
atio
n
s
,
(
K
=
5
)
is
s
elec
ted
as
th
e
o
p
ti
m
al
n
u
m
b
er
o
f
clu
s
ter
s
.
T
h
e
s
ilh
o
u
ette
s
co
r
e
i
s
ca
lcu
lated
as (
5
)
.
=
(
)
=
(
)
−
(
)
m
ax
{
(
)
,
(
)
}
(
5
)
W
h
er
e
=
(
)
in
d
icate
s
th
e
s
ilh
o
u
ette
s
co
r
e
f
o
r
o
b
s
er
v
atio
n
;
(
)
an
d
(
)
co
r
r
esp
o
n
d
to
i
n
tr
a
-
a
n
d
n
ea
r
est
-
clu
s
ter
d
is
tan
ce
s
,
r
esp
ec
tiv
ely
,
with
ma
x
{
(
)
,
(
)
}
u
s
ed
f
o
r
n
o
r
m
al
izatio
n
.
I
n
ter
p
r
etatio
n
:
=
(
)
n
ea
r
1
in
d
ic
ates
well
-
s
ep
ar
ated
clu
s
ter
s
,
v
alu
es
ar
o
u
n
d
0
s
u
g
g
est
b
o
u
n
d
ar
y
p
o
in
ts
,
n
eg
ativ
e
v
alu
es
im
p
l
y
m
is
class
if
icatio
n
,
an
d
an
av
er
ag
e
s
co
r
e
a
b
o
v
e
0
.
6
5
r
ef
lects
g
o
o
d
clu
s
ter
in
g
q
u
ality
.
2
.
4
.
3
.
Ra
nd
o
m
f
o
re
s
t
f
o
r
churn pre
dict
io
n
C
h
u
r
n
p
r
ed
ictio
n
esti
m
ates
th
e
lik
elih
o
o
d
th
at
an
ac
tiv
e
cu
s
to
m
er
b
ec
o
m
es
in
ac
tiv
e
in
th
e
u
p
co
m
in
g
p
er
io
d
.
T
h
is
s
u
b
s
ec
tio
n
d
ef
in
e
s
th
e
r
an
d
o
m
f
o
r
est ch
u
r
n
p
r
o
b
ab
ilit
y
as
d
ef
in
ed
in
(
6
)
an
d
s
u
m
m
ar
izes th
e
co
r
e
h
y
p
er
p
ar
am
eter
s
u
s
ed
f
o
r
tr
a
in
in
g
an
d
v
alid
atio
n
u
n
d
e
r
te
m
p
o
r
al
s
p
litt
in
g
.
T
h
e
ap
p
r
o
a
ch
b
u
ild
s
o
n
p
r
io
r
ch
u
r
n
m
o
d
elin
g
wo
r
k
[
1
1
]
,
ad
ap
ted
to
th
e
b
atik
m
icr
o
,
s
m
all
an
d
m
ed
iu
m
en
ter
p
r
is
es
(
MS
ME
)
co
n
tex
t
wh
er
e
class
im
b
alan
ce
an
d
tem
p
o
r
al
leak
ag
e
ar
e
cr
itical
co
n
ce
r
n
s
.
i)
C
h
u
r
n
p
r
o
b
ab
ilit
y
[
1
1
]
:
̂
(
=
ℎ
|
)
=
1
∑
1
{
ℎ
=
1
(
)
=
ℎ
}
(
6
)
w
h
er
e
r
ep
r
esen
ts
cu
s
to
m
er
f
ea
tu
r
es,
th
e
ac
tu
al
class
lab
el,
̂
(
=
ℎ
|
)
th
e
esti
m
ated
c
h
u
r
n
p
r
o
b
a
b
ilit
y
,
th
e
n
u
m
b
e
r
o
f
d
e
cisi
o
n
tr
ee
s
,
ℎ
(
)
th
e
p
r
e
d
ictio
n
f
r
o
m
tr
ee
,
an
d
1
{
⋅
}
in
d
icato
r
f
u
n
ctio
n
.
ii)
C
o
r
e
h
y
p
er
p
ar
am
eter
s
:
T
=
100
tr
e
e
s
,
ma
x
_
de
p
th
=
10
,
10
time
s
C
L
V
2
.
5
.
Ca
m
pa
ig
n
o
rc
hes
t
ra
t
i
o
n
C
am
p
aig
n
o
r
c
h
estra
tio
n
is
d
o
n
e
with
tar
g
et
p
r
io
r
ity
s
co
r
es
to
d
eter
m
in
e
th
e
m
ain
au
d
ien
ce
,
Gen
AI
co
n
tin
u
es
to
p
r
o
d
u
ce
a
d
ap
tiv
e
m
ess
ag
es
[
5
4
]
,
th
en
au
to
m
ati
ca
lly
d
is
tr
ib
u
ted
v
ia
W
h
atsAp
p
b
u
s
in
ess
API
to
in
cr
ea
s
e
th
e
ef
f
ec
tiv
e
n
ess
o
f
cu
s
to
m
er
in
ter
ac
tio
n
s
[
2
6
]
.
T
ar
g
et
p
r
io
r
ity
a
n
d
ca
m
p
ai
g
n
o
r
ch
estra
tio
n
as
d
ef
in
ed
in
(
7
)
.
i)
Prio
r
ity
s
co
r
e
(
co
m
b
in
in
g
cu
s
t
o
m
er
v
alu
e
an
d
c
h
u
r
n
r
is
k
)
.
=
.
̂
(
ℎ
⌊
)
+
.
̂
(
)
,
,
>
0
(
7
)
w
h
er
e
̂
(
)
ca
n
b
e
a
v
alu
e
esti
m
at
io
n
.
ii)
Gen
AI
f
o
r
m
u
la:
a
d
ap
tiv
e
m
ess
ag
es a
r
e
g
en
er
ated
with
c
o
n
d
i
tio
n
al
p
r
o
b
ab
ilit
ies
as
d
ef
in
ed
in
(
8
)
.
∗
=
a
r
g
ma
x
(
|
,
,
)
(
8
)
w
h
e
r
e
∗
r
ep
r
e
s
en
t
s
t
h
e
ad
ap
t
i
v
e
m
e
s
s
a
g
e
;
,
,
a
n
d
c
o
r
r
e
s
p
o
n
d
t
o
cu
s
t
o
m
e
r
em
b
ed
d
in
g
,
r
e
tr
i
e
v
e
d
p
r
o
d
u
c
t
i
n
f
o
r
m
a
t
io
n
,
an
d
c
o
n
te
x
t
p
r
o
f
i
l
e
,
w
h
er
e
a
s
d
ef
i
n
e
s
th
e
G
en
A
I
c
o
n
d
i
t
io
n
a
l
p
r
o
b
a
b
il
i
t
y
.
iii)
Au
to
m
atic
d
is
tr
ib
u
tio
n
v
ia
W
h
atsAp
p
b
u
s
in
ess
API
:
m
ess
ag
e
d
is
tr
ib
u
tio
n
is
d
o
n
e
au
t
o
m
atic
ally
af
ter
t
h
e
tar
g
et
p
r
io
r
ity
s
co
r
e
is
ca
lcu
lated
as (
9
)
.
De
l
ive
r
y
(
)
=
{
1
ℎ
(
)
≥
⋀
(
∗
,
)
=
0
ℎ
(
9
)
w
h
e
r
e
r
ep
r
e
s
en
t
s
th
e
c
u
s
to
m
e
r
,
(
)
t
h
e
p
r
i
o
r
i
t
y
s
c
o
r
e
,
t
h
e
p
r
e
d
e
f
in
e
d
th
r
e
s
h
o
ld
,
∗
t
h
e
a
d
a
p
t
i
v
e
AI
-
g
en
er
a
t
e
d
m
e
s
s
a
g
e
,
an
d
(
∗
,
)
t
h
e
m
e
s
s
a
g
e
d
e
l
i
v
e
r
y
s
t
a
t
u
s
v
ia
W
h
a
t
s
A
p
p
b
u
s
i
n
e
s
s
A
P
I
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
9
1
9
-
493
2
4924
2
.
6
.
E
v
a
lua
t
i
o
n
m
et
rics
T
h
e
ev
al
u
atio
n
is
lim
ited
to
m
etr
ics
d
ir
ec
tly
alig
n
e
d
wit
h
th
e
im
p
lem
e
n
ted
m
eth
o
d
s
,
en
s
u
r
in
g
co
n
s
is
ten
cy
an
d
r
e
p
licab
ilit
y
:
i)
R
an
d
o
m
f
o
r
est ch
u
r
n
p
r
ed
ictio
n
:
‒
Acc
u
r
ac
y
: m
ea
s
u
r
es o
v
er
all
c
o
r
r
ec
tn
ess
o
f
p
r
ed
ictio
n
s
as
d
e
f
in
ed
in
(
1
0
)
.
=
+
+
+
+
(
10
)
T
h
e
n
o
tatio
n
,
,
,
an
d
id
en
tifi
es
co
r
r
ec
t
p
o
s
itiv
e
o
u
tco
m
es,
co
r
r
ec
t
n
e
g
ativ
e
o
u
tco
m
es,
f
alse a
lar
m
s
,
an
d
m
is
s
ed
d
etec
tio
n
s
,
r
esp
ec
tiv
ely
‒
F1
-
s
co
r
e
ev
alu
ates
th
e
b
alan
ce
b
etwe
en
p
r
ec
is
io
n
a
n
d
r
e
ca
ll,
m
ak
in
g
it
p
a
r
ticu
lar
ly
s
u
itab
le
f
o
r
im
b
alan
ce
d
ch
u
r
n
d
atasets
,
as
d
ef
in
ed
in
(
1
1
)
.
=
+
,
=
+
,
1
−
=
2
×
×
+
(
1
1
)
W
h
er
e
T
P,
FP
,
an
d
FN
d
en
o
te
p
o
s
itiv
e
p
r
ed
ictio
n
o
u
tco
m
es.
C
las
s
if
icatio
n
p
er
f
o
r
m
an
ce
is
m
ea
s
u
r
ed
u
s
in
g
p
r
ec
is
io
n
,
r
ec
all
,
an
d
F1
-
s
co
r
e
,
wh
er
e
F1
-
s
co
r
e
r
ef
lect
s
th
e
b
alan
ce
b
etwe
en
b
o
th
m
e
tr
ics.
ii)
Gen
AI
p
er
s
o
n
aliza
tio
n
(
ca
m
p
a
ig
n
ef
f
ec
tiv
e
n
ess
)
‒
C
T
R
: e
v
alu
ates u
s
er
en
g
ag
em
en
t w
ith
p
er
s
o
n
alize
d
m
ess
ag
es
,
as d
ef
in
ed
in
(
1
2
)
.
=
(
)
×
100%
(
1
2
)
‒
A/B
test
in
g
with
r
eten
tio
n
m
etr
ic:
co
m
p
ar
es
p
er
f
o
r
m
a
n
ce
o
f
p
e
r
s
o
n
alize
d
ca
m
p
aig
n
s
(
g
r
o
u
p
A)
v
s
g
en
er
ic
ca
m
p
ai
g
n
s
(
g
r
o
u
p
B)
,
as sh
o
wn
in
(
1
3
)
.
△
=
−
(1
3
)
Statis
t
ical
s
ig
n
if
ican
ce
is
test
e
d
u
s
in
g
a
p
air
ed
t
-
test
o
r
ANO
VA
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
o
p
r
o
ce
s
s
an
d
an
aly
ze
d
ata,
p
r
e
-
p
r
o
ce
s
s
in
g
is
cr
u
cial
to
en
s
u
r
e
it
is
r
ea
d
y
f
o
r
f
u
r
th
er
an
aly
s
is
.
B
ased
o
n
s
y
s
tem
f
lo
ws
Gen
AI
p
er
s
o
n
aliza
tio
n
an
d
c
o
n
s
u
m
er
b
eh
av
i
o
r
an
aly
tics
en
h
a
n
ce
b
atik
cu
s
to
m
er
lo
y
alty
,
th
e
ac
tiv
ity
s
tag
e
b
e
g
in
s
with
in
p
u
ttin
g
tr
an
s
ac
tio
n
d
ata
f
o
r
1
2
m
o
n
th
s
,
co
v
er
i
n
g
all
co
m
m
er
cial
ac
tiv
ities
o
f
2
0
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
r
eg
is
ter
ed
in
t
h
e
s
tu
d
y
with
3
0
0
cu
s
to
m
er
s
.
Ov
er
all,
8
7
3
tr
a
n
s
ac
tio
n
r
ec
o
r
d
s
wer
e
p
r
o
ce
s
s
ed
as th
e
p
r
im
ar
y
in
p
u
t
f
o
r
cu
s
to
m
e
r
b
eh
a
v
io
r
a
n
aly
s
is
.
3
.
1
.
P
re
-
pro
ce
s
s
ing
,
lo
a
d,
a
nd
v
a
lid
a
t
e
da
t
a
T
h
e
d
ata
u
s
ed
f
o
r
th
e
an
al
y
s
is
ar
e
s
ev
en
f
iles
in
.
csv
f
o
r
m
at,
n
am
ely
u
m
k
m
_
d
ata.
csv
(
2
0
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es
m
ar
k
etin
g
)
,
cu
s
to
m
e
r
_
d
ata.
csv
(
3
0
0
cu
s
to
m
er
s
)
,
tr
an
s
ac
tio
n
_
d
ata.
csv
(
8
7
3
tr
an
s
ac
tio
n
s
)
,
p
r
o
d
u
c
t
_
d
a
t
a
.
c
s
v
(
4
5
p
r
o
d
u
c
ts
)
,
r
f
m
_
d
a
t
a
.
c
s
v
(
3
0
2
R
F
M
a
n
a
ly
s
i
s
)
,
s
o
ci
a
l
_
m
e
d
i
a
_
d
at
a
.
c
s
v
(
1
,
2
4
8
t
r
a
n
s
a
c
ti
o
n
s
)
,
an
d
ab
_
test
in
g
_
d
ata.
csv
(
3
0
0
A/B
test
in
g
)
.
I
n
th
e
d
ata
lo
ad
in
g
a
n
d
v
alid
atio
n
p
h
ase,
all
s
o
u
r
ce
f
iles
(
s
m
all
an
d
m
ed
iu
m
en
ter
p
r
is
e
s
o
r
SME
s
,
p
r
o
d
u
cts,
cu
s
to
m
er
s
,
tr
an
s
ac
tio
n
s
,
s
o
cial
m
ed
ia
in
ter
ac
tio
n
s
,
R
FM,
an
d
A/B
tes
tin
g
)
ar
e
lo
ad
ed
u
s
in
g
a
Pan
d
as
-
b
ased
I
/O
i
n
ter
f
ac
e
with
ex
p
licit
d
ata
t
y
p
e
ass
ig
n
m
en
t
to
m
in
im
ize
in
f
er
e
n
ce
er
r
o
r
s
.
T
h
e
d
ataset
s
u
m
m
ar
y
is
p
r
ese
n
ted
in
T
ab
le
1
,
wh
ile
th
e
d
is
tr
ib
u
tio
n
o
f
d
ata
b
etwe
en
co
n
s
u
m
er
s
,
R
FM,
an
d
A/B
test
in
g
,
with
lar
g
e
v
o
lu
m
es
o
n
s
o
cial
m
ed
ia
an
d
tr
an
s
ac
tio
n
s
,
is
s
h
o
wn
in
F
ig
u
r
e
4
.
T
h
e
s
u
m
m
ar
y
o
f
d
at
asets
u
s
ed
in
th
e
1
2
-
m
o
n
th
f
ield
s
tu
d
y
is
p
r
esen
ted
i
n
T
ab
le
2
.
T
ab
le
1
.
Su
m
m
a
r
y
o
f
d
atasets
u
s
ed
f
o
r
AI
-
b
ased
b
e
h
av
io
r
al
an
aly
s
is
D
a
t
a
s
e
t
R
o
w
s
C
o
l
s
N
u
l
l
s?
D
u
p
r
o
w
s
D
t
y
p
e
s O
K
D
a
t
e
s
O
K
u
mk
m_
d
a
t
a
.
c
s
v
20
8
No
0
Y
e
s
–
c
u
s
t
o
m
e
r
_
d
a
t
a
.
c
s
v
3
0
0
10
No
0
Y
e
s
–
t
r
a
n
s
a
c
t
i
o
n
_
d
a
t
a
.
c
s
v
8
7
3
9
No
0
Y
e
s
Y
e
s
p
r
o
d
u
c
t
_
d
a
t
a
.
c
sv
45
9
No
0
Y
e
s
–
r
f
m_
d
a
t
a
.
c
s
v
3
0
0
5
No
0
Y
e
s
–
so
c
i
a
l
_
m
e
d
i
a
_
d
a
t
a
.
c
sv
1
,
2
4
8
7
Y
e
s
0
Y
e
s
Y
e
s
a
b
_
t
e
st
i
n
g
_
d
a
t
a
.
c
s
v
3
0
0
8
No
0
Y
e
s
Y
e
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Gen
era
tive
a
r
tifi
cia
l in
tellig
en
ce
p
ers
o
n
a
liz
a
tio
n
a
n
d
co
n
s
u
mer b
eh
a
vio
r
…
(
E
n
d
a
n
g
Tja
h
ja
n
in
g
s
ih
)
4925
Fig
u
r
e
4
.
Data
d
is
tr
ib
u
tio
n
f
o
r
AI
-
b
ased
b
eh
a
v
io
r
al
a
n
aly
s
is
: p
r
o
p
o
r
tio
n
al
v
o
lu
m
e
ac
r
o
s
s
cu
s
to
m
er
,
R
FM,
A/B
test
in
g
,
tr
an
s
ac
tio
n
,
an
d
s
o
cial
m
ed
ia
r
ec
o
r
d
s
T
ab
le
2
.
Su
m
m
a
r
y
o
f
d
atasets
u
s
ed
in
th
e
1
2
-
m
o
n
th
f
ield
s
tu
d
y
D
a
t
a
s
e
t
R
e
c
o
r
d
s
D
e
scri
p
t
i
o
n
P
e
r
i
o
d
u
mk
m_
d
a
t
a
.
c
s
v
20
B
a
t
i
k
M
S
M
Es
(
S
o
l
o
,
P
e
k
a
l
o
n
g
a
n
,
B
a
k
a
r
a
n
)
1
2
mo
n
t
h
c
u
s
t
o
m
e
r
_
d
a
t
a
.
c
s
v
3
0
0
C
u
s
t
o
mer
p
r
o
f
i
l
e
s
1
2
mo
n
t
h
t
r
a
n
s
a
c
t
i
o
n
_
d
a
t
a
.
c
s
v
8
7
3
P
u
r
c
h
a
se
t
r
a
n
sa
c
t
i
o
n
s
1
2
mo
n
t
h
p
r
o
d
u
c
t
_
d
a
t
a
.
c
sv
45
P
r
o
d
u
c
t
c
a
t
a
l
o
g
1
2
mo
n
t
h
r
f
m_
d
a
t
a
.
c
s
v
3
0
2
R
F
M
sc
o
r
e
s
D
e
r
i
v
e
d
so
c
i
a
l
_
m
e
d
i
a
_
d
a
t
a
.
c
sv
1
,
2
4
8
S
o
c
i
a
l
i
n
t
e
r
a
c
t
i
o
n
s
1
2
mo
n
t
h
a
b
_
t
e
st
i
n
g
_
d
a
t
a
.
c
s
v
3
0
0
A
/
B
a
ssi
g
n
me
n
t
s
(
1
5
0
/
1
5
0
)
4
w
e
a
k
3
.
1
.
1
.
Da
t
a
clea
nin
g
a
nd
v
a
li
da
t
io
n
T
h
e
d
ata
clea
n
i
n
g
p
r
o
ce
s
s
r
e
s
u
lted
in
4
7
m
is
s
in
g
v
alu
es
(
5
.
4
%)
u
s
in
g
th
e
ap
p
r
o
p
r
iate
i
m
p
u
tatio
n
tech
n
iq
u
e
;
2
3
an
o
m
alo
u
s
tr
an
s
ac
tio
n
s
(
2
.
6
%)
with
s
tatis
tical
m
eth
o
d
s
an
d
d
o
m
ai
n
k
n
o
wled
g
e
v
alid
atio
n
.
Data
v
alid
atio
n
e
n
s
u
r
ed
1
0
0
%
co
n
s
is
ten
cy
,
r
esu
ltin
g
in
a
r
eliab
le
d
ataset
f
o
r
f
u
r
t
h
er
a
n
aly
s
is
.
Fig
u
r
e
5
p
r
esen
ts
an
o
v
er
v
iew
o
f
th
e
d
ata
clea
n
in
g
an
d
v
alid
atio
n
p
r
o
ce
s
s
.
Sp
ec
if
ically
,
Fig
u
r
e
5
(
a)
illu
s
tr
ates
th
e
m
is
s
in
g
v
alu
e
h
an
d
lin
g
p
r
o
ce
s
s
,
Fig
u
r
e
5
(
b
)
p
r
esen
ts
an
o
m
alo
u
s
tr
an
s
ac
tio
n
d
etec
tio
n
an
d
v
alid
atio
n
,
an
d
Fig
u
r
e
5
(
c
)
d
em
o
n
s
tr
ates th
e
p
o
s
t
-
v
alid
ati
o
n
co
n
s
is
ten
cy
o
f
th
e
d
ataset.
(
a)
(
b
)
(
c)
Fig
u
r
e
5
.
Data
clea
n
in
g
an
d
v
a
lid
atio
n
o
v
er
v
iew:
(
a)
m
is
s
in
g
v
alu
e
h
an
d
lin
g
(
4
7
r
ec
o
r
d
s
,
5
.
4
%),
(
b
)
an
o
m
alo
u
s
tr
a
n
s
ac
tio
n
d
et
ec
tio
n
(
2
3
r
ec
o
r
d
s
,
2
.
6
%),
an
d
(
c)
p
o
s
t
-
v
ali
d
atio
n
co
n
s
is
ten
cy
(
1
0
0
%)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
9
1
9
-
493
2
4926
3
.
1
.
2
.
Num
er
ica
l no
rma
liza
t
i
o
n a
nd
ex
plo
ra
t
o
ry
da
t
a
a
na
ly
s
is
Nu
m
er
ical
n
o
r
m
aliza
tio
n
s
u
c
ce
s
s
f
u
lly
s
tan
d
ar
d
ized
1
5
v
a
r
iab
les
with
d
if
f
er
e
n
t
s
ca
les
[
3
8
]
‒
[
4
0
]
.
C
u
s
to
m
er
ag
e
d
is
tr
ib
u
tio
n
is
co
n
ce
n
tr
ated
in
2
5
-
4
5
y
ea
r
s
,
tr
an
s
ac
tio
n
v
alu
es
ar
e
r
ig
h
t
-
s
k
e
wed
,
an
d
tr
ad
itio
n
al
b
atik
p
atter
n
s
a
r
e
p
r
ef
er
r
e
d
.
C
o
r
r
elatio
n
a
n
aly
s
is
in
d
icate
s
o
l
d
er
cu
s
to
m
er
s
te
n
d
to
g
e
n
er
ate
h
ig
h
er
tr
an
s
ac
tio
n
v
alu
es
(
r
=
0
.
7
8
)
[
4
6
]
.
T
h
e
ef
f
ec
t
s
ize
o
f
p
er
s
o
n
aliza
tio
n
o
n
tr
an
s
ac
tio
n
v
alu
e
wa
s
m
ed
iu
m
-
to
-
la
r
g
e
(
C
o
h
en
'
s
d
=
0
.
6
8
)
,
c
o
m
p
lem
en
tin
g
s
ig
n
if
ica
n
ce
test
in
g
.
Fig
u
r
e
6
p
r
o
v
i
d
es
an
o
v
er
all
v
is
u
aliza
tio
n
o
f
t
h
e
ef
f
ec
t
o
f
Z
-
s
co
r
e
s
tan
d
ar
d
izat
io
n
o
n
t
h
e
s
ca
le
d
is
tr
ib
u
tio
n
o
f
th
e
1
5
n
u
m
er
ical
v
ar
iab
les
.
Fig
u
r
e
6
(
a
)
s
h
o
ws
th
e
p
r
e
-
n
o
r
m
aliza
tio
n
s
ca
le
h
eter
o
g
en
eity
,
wh
ile
Fig
u
r
e
6
(
b
)
d
e
m
o
n
s
tr
ates
p
o
s
t
-
s
tan
d
ar
d
izatio
n
alig
n
m
en
t
,
en
ab
lin
g
u
n
b
iased
k
-
m
ea
n
s
an
d
r
an
d
o
m
f
o
r
est tr
ain
in
g
.
(
a)
(
b
)
Fig
u
r
e
6
.
Z
-
s
co
r
e
s
tan
d
ar
d
izatio
n
f
o
r
1
5
n
u
m
e
r
ical
v
ar
iab
les:
(
a)
b
ef
o
r
e
n
o
r
m
aliza
tio
n
,
s
h
o
win
g
h
eter
o
g
en
e
o
u
s
s
ca
les
an
d
(
b
)
a
f
ter
s
tan
d
ar
d
izatio
n
,
ce
n
ter
e
d
ar
o
u
n
d
ze
r
o
3
.
2
.
B
eha
v
io
r
a
na
ly
s
is
a
nd
p
re
dict
iv
e
m
o
delin
g
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
em
p
ir
ical
o
u
tc
o
m
es
o
f
b
e
h
av
io
r
al
m
o
d
elin
g
a
n
d
p
r
ed
ictiv
e
e
v
alu
atio
n
.
R
esu
lts
ar
e
o
r
g
an
ized
b
y
R
FM
en
g
in
ee
r
in
g
,
clu
s
ter
in
g
,
an
d
ch
u
r
n
p
r
ed
ictio
n
,
ea
ch
lin
k
ed
to
s
u
b
s
eq
u
en
t
Gen
AI
p
er
s
o
n
aliza
tio
n
an
d
c
am
p
aig
n
r
esu
lts
.
All
f
in
d
in
g
s
ar
e
r
e
p
o
r
ted
with
r
esp
ec
t
t
o
th
e
1
2
-
m
o
n
th
f
iel
d
s
tu
d
y
an
d
tem
p
o
r
al
h
o
ld
-
o
u
t p
r
o
to
co
l d
escr
ib
e
d
in
th
e
m
eth
o
d
s
.
i)
R
FM
f
ea
tu
r
e
en
g
in
ee
r
in
g
:
R
F
M
an
aly
s
is
s
u
cc
ess
f
u
lly
g
r
o
u
p
ed
3
0
0
cu
s
to
m
e
r
s
,
with
d
is
tr
ib
u
tio
n
s
co
r
e
s
s
h
o
win
g
4
0
%
in
th
e
h
ig
h
r
ec
e
n
cy
ca
teg
o
r
y
,
f
r
eq
u
en
cy
s
co
r
e
s
id
en
tify
in
g
2
5
%
o
f
cu
s
to
m
e
r
s
as
f
r
eq
u
en
t
b
u
y
er
s
(
≥
5
tr
a
n
s
ac
tio
n
s
)
,
an
d
m
o
n
etar
y
s
co
r
es
r
ev
ea
lin
g
th
at
3
0
%
o
f
c
u
s
to
m
er
s
co
n
tr
i
b
u
te
7
0
%
o
f
to
tal
r
ev
en
u
e
.
Fig
u
r
e
7
p
r
esen
ts
th
e
R
FM
f
ea
tu
r
e
en
g
in
ee
r
i
n
g
r
esu
lts
.
Fig
u
r
e
7
(
a)
s
h
o
ws
t
h
e
d
is
tr
ib
u
tio
n
o
f
R
FM
s
co
r
es,
wh
ile
Fig
u
r
e
7
(
b
)
s
h
o
ws
th
e
c
o
m
p
o
s
ite
R
FM
s
co
r
e
d
en
s
ity
(
3
–
1
5
)
.
C
o
m
p
o
s
ite
R
FM
s
co
r
es
(
3
-
1
5
)
s
h
o
w
co
n
ce
n
tr
at
io
n
in
t
h
e
m
id
d
le
r
an
g
e
(
7
-
1
2
)
,
wh
ich
r
e
f
lects
b
eh
av
i
o
r
al
h
et
er
o
g
en
eity
in
th
e
cu
s
to
m
er
b
ase.
ii)
K
-
m
ea
n
s
clu
s
ter
in
g
:
k
-
m
ea
n
s
clu
s
ter
in
g
with
K
=
5
s
u
cc
ess
f
u
lly
id
en
tifie
d
d
is
tin
ct
cu
s
to
m
er
s
eg
m
en
ts
b
y
s
h
o
win
g
c
h
am
p
io
n
s
4
2
c
u
s
to
m
er
s
(
1
4
.
0
%),
l
o
y
al
cu
s
t
o
m
er
s
7
3
(
2
4
.
3
%),
at
r
is
k
3
3
cu
s
to
m
er
s
(
1
1
.
0
%),
ca
n
'
t lo
s
e
5
2
cu
s
to
m
er
s
(
1
7
.
3
%),
an
d
l
o
s
t 1
0
0
c
u
s
to
m
er
s
(
3
3
.
3
%)
as sh
o
wn
in
Fig
u
r
e
8
.
iii)
R
an
d
o
m
f
o
r
est
f
o
r
ch
u
r
n
p
r
e
d
ictio
n
:
th
e
r
an
d
o
m
f
o
r
est
class
if
ier
f
o
r
ch
u
r
n
p
r
ed
ictio
n
ac
h
iev
ed
8
7
.
3
%
ac
cu
r
ac
y
with
a
p
r
ec
is
io
n
o
f
0
.
8
5
an
d
a
r
ec
all
o
f
0
.
8
9
as
s
h
o
wn
in
Fig
u
r
e
9
.
T
h
e
m
o
d
el
id
en
tifie
d
4
5
cu
s
to
m
er
s
(
1
5
%)
with
h
ig
h
ch
u
r
n
r
is
k
wh
o
r
e
q
u
ir
ed
i
m
m
ed
iate
in
ter
v
en
tio
n
s
tr
ateg
ies
,
as
s
h
o
wn
in
Fig
u
r
e
10
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Gen
era
tive
a
r
tifi
cia
l in
tellig
en
ce
p
ers
o
n
a
liz
a
tio
n
a
n
d
co
n
s
u
mer b
eh
a
vio
r
…
(
E
n
d
a
n
g
Tja
h
ja
n
in
g
s
ih
)
4927
C
am
p
aig
n
ef
f
ec
tiv
e
n
ess
(
A/B
)
:
o
v
er
th
e
4
-
wee
k
r
an
d
o
m
ized
A/B
test
(
n
=
1
5
0
p
er
g
r
o
u
p
)
,
G
e
n
A
I
-
p
e
r
s
o
n
a
li
z
e
d
m
es
s
a
g
es
(
g
r
o
u
p
A
)
a
c
h
i
e
v
e
d
C
T
R
1
2
.
4
%
v
e
r
s
u
s
1
0
.
1
%
f
o
r
g
e
n
e
r
i
c
b
r
o
a
d
c
a
s
ts
(
g
r
o
u
p
B
)
,
a
r
elativ
e
u
p
lift
o
f
+2
3
%
(
d
e
lta
=
2
.
3
p
p
,
two
-
p
r
o
p
o
r
tio
n
z
-
test
p
=
0
.
0
1
8
,
C
o
h
e
n
'
s
h
=
0
.
2
2
)
.
Pu
r
ch
ase
f
r
eq
u
e
n
cy
p
e
r
s
eg
m
en
t
in
cr
ea
s
ed
b
y
+2
1
%
an
d
r
ete
n
tio
n
(
1
-
ch
u
r
n
_
4
w)
im
p
r
o
v
e
d
f
r
o
m
5
5
%
(
B
)
to
6
7
%
(
A)
,
d
eltaReten
tio
n
=
+1
2
p
p
(
+2
1
%
r
elativ
e)
.
Statis
tical
test
s
co
n
f
ir
m
ed
im
p
r
o
v
e
m
en
t
s
wer
e
s
ig
n
if
ican
t
(p
<
0
.
0
5
)
with
p
o
wer
>
0
.
8
f
o
r
m
e
d
iu
m
e
f
f
ec
ts
.
An
ab
l
atio
n
an
aly
s
is
co
n
f
ir
m
e
d
th
at
r
em
o
v
in
g
Gen
AI
p
er
s
o
n
aliza
tio
n
r
e
d
u
ce
s
C
T
R
b
y
1
8
%
an
d
r
ete
n
tio
n
u
p
lift
b
y
1
4
p
p
,
v
er
if
y
i
n
g
its
ca
u
s
al
c
o
n
tr
ib
u
tio
n
b
ey
o
n
d
R
FM/k
-
m
ea
n
s
alo
n
e.
Hu
m
an
ev
al
u
atio
n
a
n
d
s
af
ety
r
ev
iew:
th
r
ee
b
atik
MS
ME
p
r
ac
titi
o
n
er
s
a
n
d
two
m
ar
k
etin
g
ex
p
er
ts
b
lin
d
ly
r
ated
6
0
g
e
n
er
ated
m
ess
ag
es
o
n
5
-
p
o
in
t
L
ik
er
t
s
ca
les
f
o
r
r
elev
a
n
ce
(
M
=
4
.
3
)
,
cu
ltu
r
al
ap
p
r
o
p
r
iaten
ess
(
M
=
4
.
5
)
,
a
n
d
clar
ity
(
M
=
4
.
4
)
[
3
0
]
.
All
m
ess
ag
es
wer
e
s
cr
ee
n
ed
f
o
r
eth
ical
co
m
p
lian
ce
,
en
s
u
r
in
g
n
o
m
is
lead
in
g
,
h
ar
m
f
u
l,
o
r
c
u
ltu
r
ally
i
n
ap
p
r
o
p
r
iate
co
n
ten
t,
an
d
wer
e
co
n
s
tr
ain
e
d
v
ia
p
r
o
m
p
t
f
ilter
s
(
m
ax
3
0
wo
r
d
s
,
n
o
s
u
p
er
lativ
es
with
o
u
t
ev
id
en
ce
)
.
T
h
is
h
u
m
an
-
in
-
th
e
-
lo
o
p
ch
ec
k
co
m
p
lem
en
ts
au
to
m
atic
m
etr
ics
an
d
alig
n
s
with
h
u
m
a
n
-
ce
n
ter
ed
AI
p
r
i
n
cip
les
[
2
9
]
.
Fro
m
a
d
e
p
lo
y
m
e
n
t
p
er
s
p
ec
ti
v
e,
th
e
s
y
s
tem
was
r
u
n
o
n
lig
h
tweig
h
t
clo
u
d
s
er
v
ices
an
d
th
e
W
h
atsAp
p
b
u
s
in
e
s
s
API
with
an
esti
m
ated
o
p
er
atio
n
al
co
s
t
b
elo
w
USD
2
0
p
er
MSM
E
p
er
m
o
n
t
h
,
m
ak
in
g
it a
f
f
o
r
d
ab
le
f
o
r
s
m
all
en
ter
p
r
is
es in
em
er
g
in
g
m
a
r
k
ets.
(
a)
(
b
)
F
i
g
u
r
e
7
.
R
F
M
f
e
at
u
r
e
e
n
g
i
n
e
er
i
n
g
:
(
a
)
d
i
s
t
r
i
b
u
t
i
o
n
o
f
R
F
M
s
c
o
r
e
s
,
a
n
d
(
b
)
c
o
m
p
o
s
i
t
e
R
FM
s
c
o
r
e
d
e
n
s
it
y
(
3
-
15)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
9
1
9
-
493
2
4928
Fig
u
r
e
8
.
K
-
m
ea
n
s
clu
s
ter
in
g
(
K=
5
)
Fig
u
r
e
9
.
Mo
d
el
m
etr
ics an
d
b
aselin
e
co
m
p
ar
is
o
n
(
R
F v
s
lo
g
is
tic
r
eg
r
ess
io
n
an
d
d
ec
is
io
n
tr
ee
)
Fig
u
r
e
1
0
.
Hig
h
ch
u
r
n
r
is
k
d
is
tr
ib
u
tio
n
4.
CO
NCLU
SI
O
N
T
h
e
co
m
b
in
atio
n
o
f
R
FM
an
aly
s
is
,
k
-
m
ea
n
s
clu
s
ter
in
g
,
r
an
d
o
m
f
o
r
est
ch
u
r
n
p
r
e
d
ictio
n
,
an
d
Gen
AI
p
er
s
o
n
aliza
tio
n
ef
f
ec
tiv
ely
s
t
r
e
n
g
th
en
e
d
d
i
g
ital
m
ar
k
etin
g
s
tr
ateg
ies
f
o
r
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es,
im
p
r
o
v
in
g
r
eten
tio
n
p
r
ed
ictio
n
ac
cu
r
ac
y
b
y
1
7
%,
ca
m
p
aig
n
C
T
R
b
y
2
3
%,
a
n
d
p
u
r
c
h
ase
f
r
eq
u
en
cy
b
y
2
1
%
co
m
p
ar
ed
with
g
en
e
r
ic
p
r
o
m
o
tio
n
s
(
all
p
<
0
.
0
5
,
d
=
0
.
6
8
)
.
T
h
eo
r
etica
lly
,
th
e
s
tu
d
y
e
x
ten
d
s
R
FM
-
b
ased
s
eg
m
en
tatio
n
with
tr
an
s
f
o
r
m
er
-
d
r
iv
e
n
p
er
s
o
n
aliza
tio
n
in
em
er
g
in
g
m
ar
k
ets.
Pra
ctica
lly
,
it
o
f
f
er
s
b
atik
MSM
E
s
a
s
ca
lab
le,
af
f
o
r
d
ab
le
(
<
USD
2
0
/m
o
n
t
h
)
,
an
d
cu
lt
u
r
ally
r
ele
v
an
t
s
o
lu
tio
n
th
at
b
r
id
g
es
ad
v
an
ce
d
AI
an
d
s
m
all
-
b
u
s
in
ess
n
ee
d
s
.
All
cu
s
to
m
er
d
ata
wer
e
a
n
o
n
y
m
iz
ed
b
ef
o
r
e
an
aly
s
is
,
ex
p
licit
co
n
s
en
t
was
o
b
tain
ed
f
r
o
m
p
ar
ticip
atin
g
MSM
E
o
wn
er
s
,
an
d
n
o
p
er
s
o
n
ally
id
en
tifia
b
le
in
f
o
r
m
atio
n
was
s
to
r
ed
,
en
s
u
r
in
g
co
m
p
lian
ce
with
d
ata
p
r
o
tecti
o
n
an
d
eth
ical
r
esear
ch
p
r
in
cip
les.
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
th
o
r
s
th
an
k
th
e
p
ar
ticip
atin
g
b
atik
s
m
all
b
u
s
in
ess
en
ter
p
r
is
es,
th
e
r
esear
ch
team
,
an
d
s
tu
d
en
ts
o
f
Un
iv
er
s
itas
Sti
k
u
b
an
k
,
Sem
ar
an
g
f
o
r
th
eir
co
o
p
er
atio
n
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
was
s
u
p
p
o
r
ted
b
y
th
e
Min
is
tr
y
o
f
Hig
h
er
E
d
u
ca
tio
n
,
Scien
ce
,
an
d
T
ec
h
n
o
l
o
g
y
o
f
th
e
R
ep
u
b
lic
o
f
I
n
d
o
n
esia
th
r
o
u
g
h
th
e
Ap
p
lied
R
esear
ch
Gr
a
n
t
Sch
em
e,
f
is
ca
l
y
ea
r
2
0
2
5
,
u
n
d
er
co
n
tr
ac
t
n
o
.
Evaluation Warning : The document was created with Spire.PDF for Python.