Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
3
,
Ma
rch
201
9
, p
p.
1124
~
11
29
IS
S
N:
25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
3
.pp
1124
-
11
29
1124
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Jo
bs
eeke
r
-
i
nd
ustry
m
atc
hin
g
s
yst
em using
a
ut
omated
k
eyw
or
d
s
electi
on
and
v
i
sualiz
ation
a
pp
roac
h
No
r
ha
sli
nd
a Kam
aruddi
n
,
Ab
d
ul W
ahab
A
b
dul
Rahm
an
,
Ramiz
ah
Amira
h
Mohd
Lawi
Advanc
ed
An
alytic
s
Engi
ne
eri
ng
Cent
er
,
Fa
cul
t
y
of
Com
pute
r and
Mathe
m
atic
al
S
ci
en
ce
s,
Univer
siti
Te
kno
logi
MA
RA,
Ma
lay
s
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
8
, 2
018
Re
vised Dec
6
,
2018
Acc
epte
d Dec
15
, 201
8
Le
arn
ing
opport
unit
ie
s
are
ava
i
l
abl
e
wi
th
the
acce
ss
ibi
lit
y
of
ne
w
le
arn
in
g
te
chno
logi
es,
d
iscove
r
y
of
untra
d
it
ional
learni
ng
pat
hwa
y
s
and
a
ware
ness
of
the
importan
ce
of
connect
ing
cur
ren
t
knowl
edge
with
new
le
arn
ing.
Such
situa
ti
on
allows
the
expa
nsion
i
n
the
num
ber
of
cour
ses,
progra
m
s
and
profe
ss
iona
l
ce
r
t
ifi
c
at
ions
offe
r
e
d
to
the
students
result
ing
to
the
inc
rement
of
the
num
ber
of
gra
duat
es
ann
ual
l
y
.
The
gra
d
uat
es
the
n
empl
o
y
ed
b
y
th
e
industr
y
for exe
c
uti
ng
the
job.
Ho
weve
r,
th
ere
is a
growing
conc
ern
about
the
inc
rement
of
un
emplo
y
ed
gra
du
at
es
in
the
job
m
ark
et.
One
of
the
rea
sons
of
the
m
ism
at
ch
bet
wee
n
gra
du
at
e
s’
skill
s
and
e
m
plo
y
ers’
nee
ds
is
tha
t
the
jobsee
ker
s
te
nd
to
choose
wrong
job
bec
ause
th
e
y
a
re
over
whel
m
ed
by
th
e
choi
c
es
and
t
y
pi
ca
l
l
y
they
just
r
andoml
y
send
t
he
appl
i
cation
b
ec
ause
it
is
ti
m
e
consum
ing
to
fil
t
er
r
el
ev
ant
adve
rt
.
Such
a
c
ti
on
m
a
y
have
r
epe
rcu
ss
ion
to
the
industr
y
bec
ause
the
emplo
y
ers
nee
d
to
sele
c
t
releva
n
t
c
andi
da
te
s
to
fil
l
up
th
e
post
f
rom
the
unf
il
t
er
ed
pi
le
o
f
app
lic
at
i
ons
m
aki
ng
t
he
sel
ec
t
ion
proc
ess l
engt
h
y
and
ti
m
e
consu
m
ing.
In
thi
s pa
p
er
we
proposed
a
n
aut
om
at
ed
appr
oac
h
to
m
atch
the
gr
adua
t
es’
and
emplo
y
ers’
nee
ds using
a
h
y
brid
of text
m
ini
ng
and
visual
i
za
t
ion
appr
oa
ch
to
fa
ci
l
it
a
te
j
obsee
ker
s’
ta
sk
of
rel
ev
an
t
jo
b
applic
at
ion
.
The
import
ant
k
e
y
words
are
autom
at
ic
a
lly
ext
ra
ct
ed
base
d
on
the
fre
qu
ency
of
th
e
word
used
in
th
e
adve
r
ts.
The
n
,
th
e
gra
d
uat
es’
skill
s
are
m
at
ch
ed
f
rom
the
ir
per
sonali
z
ed
profi
le
.
R
el
ev
ant
v
isual
izat
ion
appr
oac
h
es
ar
e
inc
orpora
te
d
to
fac
i
li
t
at
e
the
sel
ec
t
ion.
I
t
is
pra
ct
i
ca
l
and
fea
sibl
e
for
th
e
proposed
appr
oac
h
to
b
e
incorporat
ed
in
jo
b
sea
rch
ing
website
s
th
at
can
opti
m
iz
e
jobs
ee
ker
s
and
empl
o
y
ers
ti
m
e
and
eff
ort
for
a
suita
ble m
at
ch
.
Ke
yw
or
ds:
Adve
rt
f
il
te
rin
g
Au
t
om
at
ed
k
ey
word
e
xtracti
on
Jo
b
s
ea
rch
i
ng
Sk
il
l
m
a
tc
hin
g
Visu
al
iz
at
ion
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Norhasl
ind
a
Ka
m
aru
ddin
,
Adva
nced A
na
ly
ti
cs En
gin
ee
r
ing
Ce
nter
,
Faculty
of Com
pu
te
r
an
d
Ma
them
a
ti
cal
Scie
nces,
Un
i
ver
sit
i Te
knol
og
i M
ARA
,
Sh
a
h Alam
, S
el
angor, Mal
ay
sia
.
Em
a
il
:
no
r
haslind
a
@tm
sk
.u
it
m
.ed
u.
m
y
1.
INTROD
U
CTION
L
ifel
ong
le
ar
nin
g
c
on
ce
pt
al
lows
le
ar
ners
to
ed
ucate
them
se
lves
reg
ar
dl
ess
on
their
sta
ge
of
ca
ree
r
and
li
fe
. W
it
h
t
he
uncertai
nty
of
the
c
urre
nt
econom
ic
con
di
ti
on
,
the n
ee
d
f
or
c
on
ti
nui
ng
s
tud
ie
s
is
nee
de
d
f
or
up
s
kill
ing
for
bette
r
w
ork
pe
rfor
m
ance,
se
ekin
g
a
highe
r
sal
ary
an
d
be
tt
er
opportu
ni
ti
es
,
upgr
a
ding
the
te
chnolo
gy
under
sta
nd
i
ng
a
nd
im
pr
ovin
g
m
ark
et
abili
ty
.
This
tre
nd
ca
n
be
obse
rv
e
d
with
high
nu
m
ber
of
higher
e
ducat
ion
i
ns
ti
tuti
on
s
(H
E
I)
i
n
Ma
la
ysi
a
that
is
reco
r
de
d
to
be
69
9
in
2018
[
1
]
.
The
higher
e
ducat
ion
insti
tuti
on
s
in
Ma
la
ysi
a
inclu
des
pu
blic
un
i
ver
sit
ie
s
(IPT
A)
,
pr
i
vate
hi
gh
e
r
ed
ucati
on
insti
tuti
on
s
(
I
PTS),
po
ly
te
ch
nics,
c
omm
un
it
y
colleg
es
a
nd
te
ch
ni
cal
and
vocat
ion
al
trai
ning
i
ns
ti
tuti
on
s
.
T
he
detai
le
d
nu
m
ber
of
the
H
EIs
is
pr
esented
in
Fig
ur
e
1(a)
.
Although
t
he
nu
m
be
r
of
IP
T
As
is
m
ai
ntained,
t
he
num
ber
of
I
PTS
is
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Jobseek
er
-
i
ndust
ry matc
hing
s
yst
em usin
g aut
omated
keyw
ord sele
ct
io
n and
...
(
Nor
hasli
nda K
ama
rud
din
)
1125
sk
yr
ock
et
e
d.
T
his
is
beca
us
e
t
he
dem
and
is
hi
gh
des
pite
the
sign
ific
a
ntly
charge
d
higher
f
ees
w
he
n
c
ompare
d
to
pu
blic
insti
tuti
ons.
Garwe
(20
16)
c
onduc
te
d
a
st
ud
y
i
n
Zim
bab
we
a
nd
repor
te
d
t
hat
the
m
ai
n
fact
or
of
influ
e
ncin
g
st
udents
’
c
ho
ic
es
,
are
nam
ely;
a
ccess
an
d
opport
un
it
y,
prom
otion
al
i
nfor
m
at
ion
a
nd
m
ar
keting,
ref
e
ren
ce
or
i
nf
l
uen
ce
by
ot
her
s
,
qu
al
it
y
of
te
achi
ng
an
d
le
arn
i
ng,
fees
an
d
c
os
t
str
uc
ture
a
nd
aca
dem
ic
reputat
ion
a
nd
recog
niti
on
[2
]
. S
uc
h
sit
uatio
n
is al
so
a
ppli
ed
in Mal
ay
sia
. Hence
, it i
s n
ot
su
r
pri
sing
t
o
s
ee the
nu
m
ber
of gra
du
at
es
is also
incr
ease
d
.
The
gr
a
duat
es
will
then
enter
the
j
ob
m
ark
e
t
and
ap
ply
f
or
the
rele
van
t
po
sts
.
T
he
opport
un
it
y
t
o
seek
a
rele
van
t
j
ob
beco
m
e
m
or
e
c
halle
ngin
g
beca
us
e
t
he
gr
a
duat
es
nee
d
to
com
pete
with
each
oth
e
r
to
see
k
the
jo
b.
As
the
num
ber
of
gra
du
at
e
i
ncr
ease
s,
the
dem
and
and
com
petit
ion
f
or
a
j
ob
will
be
m
or
e
strin
gen
t
.
Ba
sed
on
the
grad
uate tracer s
tud
y co
nducte
d by Mi
nistry of
H
igh
e
r
Ed
ucat
ion
[
1
]
, th
e tre
nd
for
grad
ua
te
s f
or
ICT
sect
or
inc
reases
e
xpone
ntial
ly
as
il
lustrate
d
in
Fig
ure
1(b
).
T
he
num
ber
of
gra
duat
es
in
ICT
sect
or
rangin
g
from
22,64
2
to
27,
735
grad
uates
that
co
m
pr
ise
s
of
ap
pro
xim
at
el
y
8.
69
%
f
r
om
aver
age
of
total
nu
m
ber
of
gra
du
at
es
in
Ma
l
ay
sia
(2
75,
465
gr
a
du
at
es
)
from
20
13
to
2017.
In
the
gradu
at
e
trace
r
stud
y,
Ma
la
ysi
a
gr
ad
uate
em
plo
yme
nt
sta
tus
is
al
so
prese
nted
[
1
]
.
G
raduate
e
m
plo
yme
nt
sta
tus
can
be
div
i
ded
i
nto
five
cat
eg
ori
es
,
w
hich
a
re;
e
m
plo
ye
d,
fur
ther
st
ud
y,
up
gr
a
ding
s
kill
s,
wait
ing
f
or
work
place
m
e
nt
an
d
un
em
plo
ye
d.
T
able
1
shows
t
he
total
nu
m
ber
of
gra
du
at
es
and
thei
r
e
m
plo
ym
ent
s
ta
tus
from
20
13
to
2017.
Fr
om
Table
1,
it
is
ob
ser
ve
d
that
25%
t
o
30
%
grad
uat
es
are
unem
plo
ye
d
a
nd
fro
m
the
sa
m
e
rep
ort
,
the
reas
ons
of
un
em
plo
ym
ent
are
disc
us
se
d.
Alm
o
st
70
%
to
75%o
f
the
pa
r
ti
ci
pan
ts
sta
te
d
that
the
reas
on
that
they
are
unem
plo
ye
d
is
beca
us
e
they
are
st
il
l
seeking
f
or
a
j
ob
ra
ngin
g
from
37
,31
6
gr
a
duat
es
in
2017
to
39,86
4
gr
a
duat
es
in
2013.
Ot
her
reas
ons
giv
en
are;
ta
king
a
break
,
wait
ing
for
placem
ent
t
o
f
ur
t
he
r
stu
dy
,
respo
ns
ibil
it
y
t
ow
a
r
ds
fam
ily,
j
obs
off
ere
d
not
su
it
able,
c
hoose
not
to
w
ork
,
not
interest
ed
to
w
ork
,
la
ck
of
sel
f
-
c
onfide
nc
e
to
face
w
orkin
g
e
nv
i
ronm
ent,
healt
h
pro
blem
,
ref
us
e
t
o
m
ov
e
t
o
a
nothe
r
place
a
nd
oth
e
r
reasons
. T
he dist
ribu
ti
on
of th
e grad
uate trac
er r
e
su
lt
of
t
he un
em
plo
ym
ent
f
act
ors is
de
picte
d
in
Fig
ur
e
2
.
Figure
1
(a
)
.
N
um
ber
of m
al
a
ysi
a h
ig
her
e
du
cat
ion
insti
tute (he
i)
by
yea
r
[
1
]
Figure
1(b
).
Pe
rcen
ta
ge
of
m
al
ay
sia
n
gr
a
duat
e in ict
sect
or
from
2
013 an
d 2
017 [
1]
Figure
2
.
Fact
ors
of
grad
uate
un
em
plo
ym
ent
in
m
al
ay
sia
f
ro
m
2
013
a
nd
2018
[
1]
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
3
,
Ma
rc
h 201
9
:
1124
–
11
29
1126
Table
1.
Mal
ay
sia
n Gr
a
du
at
es
Em
plo
ya
bili
ty Stat
us
by
Year
[
1]
Year
E
m
p
lo
y
ed
Fu
rther Stud
y
Up
g
radin
g
Sk
ills
W
aitin
g
f
o
r
W
o
rk
Place
m
en
t
Un
e
m
p
lo
y
m
en
t
Total
Gradu
ates
2013
1
0
1
,286
4
2
,44
3
2
,91
1
1
2
,90
8
5
3
,28
2
2
1
2
,830
2014
1
0
1
,619
4
3
,83
6
3
,22
3
8
,94
1
5
2
,21
9
2
0
9
,838
2015
1
2
1
,740
4
0
,06
7
3
,77
6
9
,13
3
5
4
,85
2
2
2
9
,568
2016
1
3
4
,561
3
4
,51
0
5
,39
4
9
,61
9
5
4
,10
3
2
3
8
,187
2017
1
4
1
,257
4
3
,49
5
5
,06
8
1
1
,90
6
5
3
,37
3
2
5
5
,099
Fr
om
Fig
ur
e
2,
it
is
ob
vious
t
hat
gra
du
at
es
a
re
sti
ll
unem
pl
oyed
beca
us
e
t
hey
are
sti
ll
try
ing
t
o
fin
d
a
su
it
able
j
ob.
T
he
j
ob
searc
hi
ng
pr
ocess
is
te
dious
a
nd
ti
m
e
co
nsum
ing
be
cause
jobse
ek
er
needs
to
ev
al
uate
the
releva
ncy
of
the
ad
ve
rt
befor
e
a
pp
ly
in
g
an
d
for
the
po
te
ntial
e
m
pl
oyer
to
ga
ug
e
the
su
it
abili
ty
of
the
cand
i
date.
T
he
re
is
a
nee
d
to
m
at
ch
betwee
n
the
gr
a
duat
es’
sk
il
l
set
s
with
the
nee
d
of
the
industry
f
or
t
he
job
po
ste
d.
T
o
da
te
,
there
is
onl
y
a
basic
filt
er
prov
i
ded
by
the
jo
b
ad
ver
ti
s
e
m
ent
web
sit
e
s
for
jo
b
sel
ect
ion
by
sp
eci
fyi
ng
le
ve
l
of
ed
ucati
on
,
locat
io
n,
s
pec
ific
at
ion
a
nd
m
ini
m
u
m
sal
ary.
I
n
this
pa
pe
r,
we
a
re
pro
posin
g
a
pr
act
ic
al
ap
pro
ach
to
m
at
ch
t
he
gr
a
duat
es’
sk
il
l
set
s
an
d
the
nee
d
of
th
e
pote
ntial
em
plo
ye
r
by
us
in
g
a
n
autom
at
ed
ke
yword
e
xtracti
on
an
d
in
c
orp
or
at
es
visu
al
iz
at
ion
to
facil
it
at
e
sel
ect
ion
.
For
sim
plificati
on
,
we foc
us
on
ICT w
e
b
-
relat
ed j
ob
s
posit
io
n.
This
pa
per
is
orga
nized
in
the
f
ollow
i
ng
m
ann
er
.
Sect
ion
2
desc
ribes
on
avail
able
job
sea
rch
i
ng
web
sit
es
an
d
di
ff
ere
nt
ap
proa
ches
us
e
d
to
e
xtr
act
releva
nt
keyw
ords
f
ro
m
do
c
um
ents.
Sect
ion
3
prese
nts
the
ov
e
rall
fr
am
ework
of
the
pr
opos
e
d
ap
proa
ch.
As
this
w
ork
is
sti
ll
wo
r
k
in
pro
gr
ess
,
a
prototype
of
t
he
Job
Ma
tc
hin
g
Syst
e
m
is
il
lustrated
as
well
.
To
co
nclu
de
th
e
pap
e
r,
Sect
ion
4
pro
vid
es
su
m
m
ar
y
and
fu
tu
r
e
directi
on
of
t
he
r
esea
rch
.
2.
LIT
ERATUR
E REVIE
W
Gr
a
duat
es
ty
pical
ly
will
find
ing
job
s
by
re
gi
ste
ring
thei
r
pro
file
s
to
job
s
earchi
ng
we
bs
it
es
su
ch
a
s
Jo
bSt
reet.c
om
[
3
]
,
Li
nk
e
dIn
[
4
]
,
Mo
ns
te
r
[
5
]
an
d
ot
hers.
T
he
se
we
bs
it
es
pr
ov
i
de
a
platfo
r
m
fo
r
the
j
obse
eke
r
to
fin
d
a
s
uitab
le
j
ob
a
nd
pote
ntial
e
m
plo
ye
r
s
to
ad
ve
rtise
job
or
posit
io
n
that
they
nee
d
to
hire
.
S
om
e
of
t
he
web
sit
e
pr
ov
i
de
a
per
s
on
al
a
ssist
ance
to
th
e
j
obsee
ker
as
par
t
of
their
serv
ic
e
to
facil
it
at
e
the
proc
ess
of
sel
ect
ing
the
r
el
evan
t
job
f
or
the
app
li
cants
.
The
job
see
ke
r
prof
il
e
is
avail
able
fo
r
the
com
pan
y
to
fu
rt
her
analy
sed.
Mo
r
eov
e
r,
the
j
ob
seeker
s
can
al
so
vie
w
the
c
om
pan
y
prof
il
e
to
help
the
m
to
cho
os
e
t
he
m
os
t
su
it
able
caree
r
p
at
h
f
or
them
.
Jo
bSt
reet.c
om
and
Li
nk
e
dIn
need
t
he
us
e
rs
to
sign
i
n
be
fore
they
can
use
the
serv
ic
es
.
On
the
c
ontrary,
I
ndeed
al
lo
ws
use
r
to
di
rectl
y
search
the
relev
ant
j
ob
by
s
pe
ci
fyi
ng
t
he
job
ti
tl
e,
keyw
ords
a
nd
com
pan
y
as
w
el
l
as
the
locat
i
on
of
the
jo
b
offe
red.
U
sers
c
an
the
n
c
reate
or
uploa
d
resum
e
and
sign
i
n
to
e
nsu
re
the
sec
ur
it
y
of
t
he
create
d
resu
m
e.
Link
e
dIn
offer
s
c
on
necti
on
to
var
i
ou
s
us
e
r
s
ocial
m
edia
accounts
f
or
be
tt
er
reacha
bili
ty
and
visibil
it
y.
More
over,
JobStreet
.c
om
se
gr
e
gated
job
i
n
m
any
fiel
ds
suc
h
as
il
lustrate
d
in
F
igure
3.
In
thi
s
pap
e
r
we
ar
e
fo
c
us
in
g
on
web
relat
ed
jo
b
s
that
co
ntain
s
the
process
of
web
dev
el
op
m
ent.
The
five
jo
b
s
t
hat
can
be
c
onsidere
d
as
the
web
relat
ed
jo
b
s
,
are
nam
el
y;
so
ft
war
e
de
ve
lop
e
r,
web
de
velo
pe
r, so
ftwa
r
e e
ng
i
neer,
Net d
e
vel
op
e
r
a
nd P
HP
dev
el
op
e
r
.
On
ce
sc
ope
ha
s
been
set
,
we
colle
ct
ed
100
job
a
dverti
sem
e
nts
that
are
rel
at
ed
to
the
web
relat
ed
jo
b.
The
releva
nt key
word
s ar
e m
anu
al
ly
ex
tract
ed
from
the sk
i
ll
s r
equ
irem
ent. Th
is i
s to
find co
rr
el
at
io
n
bet
wee
n
sk
il
ls
an
d
job
adv
e
rtise
d
an
d
to
s
ee
wh
et
he
r
the
re
a
re
sim
il
ar
re
qu
ir
em
e
nts
need
e
d
by
m
ul
ti
ply
j
ob
a
dv
e
rts.
The
sk
il
ls
are
m
app
ed
on
t
o
a
Vern
diag
ram
in
Figure
4
to
sh
ow
inter
de
pe
nd
e
ncy
of
the
sk
il
ls
and
the
j
ob.
It
is
noti
ced
th
at
there
a
re
ge
ner
al
s
kil
ls
that
are
need
e
d
by
al
l
web
relat
ed
job
re
qu
irem
ents
suc
h
a
s
MS
SQ
L
,
.N
et
a
nd
A
SP
with
m
ini
m
u
m
qual
ific
at
ion
of
dip
l
om
a
and
/o
r
de
gree
in
the
relat
ed
fiel
d
s
.
Furthe
rm
or
e,
it
is
al
so
obser
ve
d
that
there
are
sp
eci
fic
nee
ds
for
certai
n
s
kill
fo
r
diff
e
re
nt
jo
b.
For
in
sta
nce,
S
hell
Script
pro
gr
am
m
ing
is
ver
y
m
uch
ne
eded
f
or
we
b
dev
el
op
e
r
as
com
par
ed
to
.Net
dev
el
oper
.
It
al
so
sh
ows
th
at
it
is
i
m
po
rtant
t
o
pr
epar
e
onesel
f
f
or
t
he
s
kill
s
ne
eded
by
the
in
du
st
ry
f
or
t
hat
par
ti
cula
r
jo
b
t
o
inc
rease
c
ha
nces
of
secur
i
ng
t
he
job
a
dverti
sed
.
Howe
ver
,
m
anu
al
s
kill
m
app
in
g
is
pro
ne
to
error
a
nd
ve
ry
tim
e
con
su
m
ing
.
Data
redu
nd
a
ncy
and
m
iss
ing
data
probl
e
m
will
al
ways
com
pr
om
i
se
the
accura
cy
of
the
m
a
pp
i
ng.
Hen
ce
, a
n
a
utom
at
ed
keyw
ord
e
xtracti
on
ne
eds
t
o be
us
ed
to
sim
plify t
he
process
.
Using
the
c
orrect
keyw
ord
s
for
j
ob
se
arch
i
ng
i
ncr
ea
se
s
the
c
hances
to
fi
nd
r
el
evan
t
jo
b
adv
e
rtise
m
ents
and
getti
ng
shortli
ste
d
f
or
an
intervie
w.
Howev
e
r,
job
see
ke
rs
s
om
eti
m
es
are
not
sure
of
th
e
rig
ht
key
words
of
the
job
that
they
wa
nt
to
s
ecur
e
.
T
hey
m
ay
us
e
brute
-
force
a
ppro
ac
h
t
o
te
st
any
key
words
that
m
ay
be
relevan
t
an
d
s
uc
h
ap
proac
h
is
tim
e
con
su
m
ing
an
d
ene
r
gy
intensi
ve.
He
nc
e,
m
any
research
ers
hav
e
pro
po
se
d
m
any
te
chn
iq
ues
to
aut
om
atical
ly
extract
relevan
t
keywords
from
te
xt
do
c
um
ents.
This
process
i
s
to
sel
ect
w
ords
a
nd
phrases
from
the
te
xt
do
c
um
ent
s
tha
t
can
giv
e
the
gist
of
t
he
i
ntend
e
d
inf
or
m
at
ion
wi
thout
the
hu
m
an
in
volvem
ent
[
6
]
.
Once
rele
van
t
keyw
ords
are
e
xtracted
,
the
inf
orm
ation
ca
n
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Jobseek
er
-
i
ndust
ry matc
hing
s
yst
em usin
g aut
omated
keyw
ord sele
ct
io
n and
...
(
Nor
hasli
nda K
ama
rud
din
)
112
7
be
us
e
d
to
fin
d
a
bette
r
m
at
ch
fo
r
j
ob
searc
h
qu
e
ries.
Bhart
i
et
a
l
.
[
7
]
cat
egorized
e
xtract
ion
syst
em
int
o
fo
ur
cl
asses;
nam
ely,
si
m
ple
sta
t
ist
ic
al
,
li
ng
uis
t
s
,
m
achine
lear
ni
ng
a
nd
hy
br
i
d
ap
proac
he
s.
Sim
ple
statist
ic
appr
oach
syst
em
loo
ks
from
the
perspecti
ve
of
the
ra
w
docum
ent
suc
h
as
f
reque
nc
y
of
the
w
ord
us
e
d,
locat
ion
of
th
e
w
ord
i
n
the
docum
ent
an
d
is
t
he
s
im
plest
appr
oach
to
im
ple
m
ent
com
par
ed
t
o
othe
r
appr
oach
es
m
akin
g
the
pro
cessi
ng
ti
m
e
is
kep
t
to
m
i
nim
a
l.
The
li
nguists
ap
pro
ach
inc
orp
or
at
es
the
unde
rstan
ding
of
la
ngua
ge
a
naly
sis
su
ch
a
s
le
xical
,
synta
ct
ic
,
discours
e
and
sem
antic
of
the
wor
d
us
ed.
The
m
achine
l
earn
i
ng
ap
proa
ch
us
es
t
he
power
of
cl
assifi
e
rs
s
uc
h
as sup
port v
ect
or
m
achine,
naï
ve
bay
es
an
d
deep
le
a
rn
i
ng
to
unde
rstan
d
the
co
ntent
s
of
the
a
dv
e
rt
s
a
nd
us
es
weig
ht
to
m
a
tc
h
between
t
he
j
obs
eeke
r
app
li
cat
io
n
an
d
the
job
ad
ver
ti
sem
ent.
I
n
a
dd
it
io
n
,
hy
br
id
a
ppr
oac
h
c
om
bin
es
two
or
m
or
e
pr
e
vious
appr
oach
es
an
d uti
li
ze the str
eng
t
h of t
he
sel
ect
ed
ap
proac
he
s.
Figure
3
.
Cl
assifi
cat
ion
of jo
b base
d o
n jobst
reet.com
p
ers
pe
ct
ive [3]
Figure
4
.
Job
r
equ
i
rem
ents interd
e
pende
ncy
base
d on extra
ct
ed
jo
b descri
ption
Am
at
o
et
a
l.
[8]
at
tem
p
te
d
to
autom
at
e
the
resu
m
e
m
anag
e
m
ent
fo
r
m
at
ching
ca
nd
i
date
pro
file
s
with
j
ob
desc
riptio
n
by
com
par
ing
se
ver
al
te
chn
i
qu
e
s
su
c
h
as
li
near
SV
C
,
ru
le
-
base
d
and
Lat
ent
Di
r
ic
hlet
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
3
,
Ma
rc
h 201
9
:
1124
–
11
29
1128
Allocat
ion
(L
DA)
to
cl
assif
y
j
ob
ad
ve
rtise
m
ent
against
a
n
occ
upat
ion
c
la
ssifie
r
us
e
d
by
the
Ital
ia
n
N
at
ion
al
In
sti
tute
of
St
at
ist
ic
(I
STA
T
).
T
hey
re
port
ed
that
L
DA
perform
ed
consi
ste
ntly
well
for
ave
ra
g
e
pr
eci
sion
,
recall
and
FSc
or
e
as
c
om
pared
to
m
achine
le
arn
in
g
a
nd
r
ule
-
based
a
ppr
oach.
Linea
r
S
VC
nee
ds
s
uff
ic
ie
nts
trai
ning
data
t
o
yi
el
d
good
pe
rfor
m
ance
an
d
r
ule
-
base
d
a
ppr
oach
nee
d
extra
at
te
ntio
n
by
the
ex
pe
rt
for
the
ru
le
de
velo
pme
nt.
Hauff
an
d
Gous
io
s
[
9]
propose
d
a
pip
el
i
ne
that
a
uto
m
at
iz
e
j
ob
m
at
ching
ad
ve
rtisem
e
nts
to
dev
el
op
e
rs
by
lookin
g
at
thre
e
m
ai
n
com
po
nen
ts
,
nam
el
y;
con
ce
pt
ext
ra
ct
ion
f
ro
m
j
ob
adv
e
rtise
m
ent
s
an
d
so
ci
al
co
ding
us
er
data,
co
nc
ept
wei
gh
ti
ng
and
co
nce
pt
m
at
ching.
Ba
sic
con
ce
pt
of
Ter
m
Fr
equ
e
ncy
I
nv
e
rs
e
Do
c
um
ent
Fr
e
qu
e
ncy
(T
F
-
IDF)
is
em
plo
ye
d
for
co
nce
pt
w
ei
gh
in
g.
TF
-
I
D
F
giv
es
l
ow
we
igh
t
f
or
c
oncep
t
that
is com
m
on
and
apprea
r
in
m
a
ny doc
um
ents an
d hi
gh w
ei
ght t
o
c
oncepts t
hat o
cc
ur m
any tim
es in a
docum
ent
bu
t
ra
rely
acro
ss
enti
re
c
orpu
s
.
The
resu
l
t
sh
ows
that
there
is
a
subs
ta
ntial
ov
erlap
between
t
he
entit
ie
s
extracte
d
from
j
ob
a
dv
e
rtise
m
ents
an
d
the
e
nt
it
ie
s
extracte
d
from
dev
el
op
e
r
pro
file
s
and
t
he
li
near
c
orrelat
ion
betwee
n
the
nu
m
ber
of
tim
es
a
con
ce
pt
ap
pe
ars
in
de
velo
pe
r
pro
file
s
vs
jo
b
ad
ver
ts
is
r=
0.49.
I
n
m
or
e
recent
work,
M
uth
ya
la
et
al
.
[
10
]
discuss
e
d
a
m
eth
od
ology
that
i
m
pr
oves
us
er
j
ob
sea
rc
hing
exp
e
rience
by
add
i
ng
sk
il
l
set
an
d
c
om
pan
y
at
trib
ute
filt
ers.
T
he
TF
-
I
DF
wei
gh
ti
ng
is
us
e
d
to
cal
c
ulate
the
fr
e
qu
e
ncies
f
o
r
al
l
un
i
qu
e
sk
il
ls
of
the
do
c
um
ent
it
sel
f.
The
n,
s
ever
al
sim
il
ari
t
y
m
easur
em
ents
wer
e
us
e
d
to
m
easur
e
sim
il
arit
y
of
the
t
wo
doc
um
ents
based
on
t
heir
feat
ure
vecto
r.
T
he
j
ob
sea
rch
res
ult
sh
ows
th
at
the
filt
ered
j
obs
are
ranke
d
us
in
g
a
relevan
ce
sc
ore
der
i
ved
f
r
om
a
weigh
te
d
com
bin
at
ion
of
sk
il
l
set
s
and
com
pan
ie
s
extern
al
factors.
He
nce,
in
this
work
w
e are
us
in
g TF
-
ID
F
to e
xtract
releva
nt k
ey
word
s
.
3.
RESEA
R
CH MET
HO
DOL
OGY
Our
m
et
ho
d
c
on
sist
s
of
se
ve
ral
com
pone
nts
that
re
qu
ir
es
pa
rsing,
int
erpreti
ng
a
nd
norm
al
is
ing
sem
i
/
un
struct
ured
data
gathe
red
f
r
om
j
ob
se
eker
re
su
m
e
a
nd
j
ob
ad
ve
rtise
m
ent
to
create
a
reco
m
m
end
at
ion
eng
i
ne
that
will
be
pr
ese
nted
in
a
j
ob
searc
hi
ng
we
bs
it
e
.
C
raw
le
r
will
captur
e
the
poste
d
j
ob
ad
ver
ti
se
m
ents.
In
our
pr
el
im
inary
wor
k,
100
jo
b
a
dv
e
rtisem
ents
are
colle
ct
ed
f
or
five
web
-
rel
at
ed
j
obs,
na
m
el
y;
web
de
vel
op
e
r
,
s
of
twa
re
de
ve
lop
e
r,
s
oft
wa
re
e
ng
i
neer,
P
HP
de
velo
per
and
.
Net
dev
el
op
e
r.
T
he
s
kill
set
s
require
d
by
th
e
com
pan
ie
s
will
be
extract
ed
us
i
ng
T
F
-
I
DF
.
T
he
j
obse
eker
s
res
um
e
is
reco
r
de
d,
a
nd
t
he
j
obsee
ker
ac
quired
sk
il
l
set
are
al
so
extracte
d.
The
ac
qu
i
re
d
job
see
ke
r
an
d
the
nee
ded
c
om
pan
y
sk
il
l
s
et
s
are
ranke
d
usi
ng
featur
e
ra
nkin
g
m
et
ho
d.
Thi
s
is
to
inc
reas
e
the
pe
rfo
rm
ance
of
t
he
re
com
m
end
er
syst
e
m
.
Fo
r
t
his
prel
i
m
inary
w
ork,
a
s
i
m
ple
fr
eq
uen
c
y
cal
culat
ion
is
us
ed
.
T
hen,
a
con
ce
pt
m
at
c
hing
is
i
m
ple
m
ente
d
to
determ
ine
the
sim
il
arity
of
the
a
dv
e
rtise
d
job
an
d
jo
bse
eker
pr
of
il
e.
We
pro
pose
to
us
e
c
os
ine
sim
il
arity
m
et
ho
d. T
he pr
opos
e
d j
ob r
ec
omm
end
er
syst
e
m
w
ork
flo
w
i
s prese
nted
i
n Fi
gure
5.
The
res
ult
will
then
be
in
t
he
range
of
pe
rce
ntage
wh
e
re
th
e
higher
sc
or
e
ind
ic
at
es
highe
r
si
m
il
arity.
Fo
r
t
he
ease
of
sel
ect
io
n,
a
visu
al
iz
at
ion
appr
oach
is
use
d
f
or
the
jo
b
sel
ect
ion
w
ebsite
as
dep
i
ct
ed
in
Figure
6.
I
nfo
r
m
at
ion
su
ch
a
s
j
ob
ti
tl
e,
lo
c
at
ion
,
sk
il
ls
ne
eded
a
nd
e
xp
e
ct
ed
sal
ary
are
rev
eal
ed
to
th
e
j
ob
seeker
for t
he
m
to m
ake a
w
el
l
-
inform
ed
de
ci
sion
a
nd
pr
e
sented
in
t
he fo
rm
o
f
da
shb
oard.
4.
CONCL
US
I
O
N
The
e
xtracti
on
of
rele
van
t
i
nfo
rm
ation
for
j
ob
s
el
ect
ion
i
s
not
a
tri
vial
ta
sk
.
E
na
bling
autom
at
ed
keyw
ord
e
xtra
ct
ion
fr
om
j
ob
s
eeker res
um
e a
nd jo
b
ad
ve
rtisem
ent
m
ay
f
aci
li
ta
te
j
ob
see
ke
r
to
fin
d
r
el
eva
nt jo
b
in
a
m
ini
m
al
t
i
m
e
and
help
com
pan
ie
s
to
get
bett
er
ca
ndidate
s
to
be
c
onside
red
f
or
th
e
j
ob.
Alth
ough
th
e
work
prese
nted
is
only
pr
el
i
m
anar
y
w
ork,
it
sh
ows
pote
nt
ia
l
to
be
em
b
edd
e
d
in
t
he
c
urren
t
j
ob
sea
r
chin
g
web
sit
es.
F
ur
t
her
w
orks
nee
d
to
be
i
ncor
porate
d
to
e
nsu
re
the
su
cces
s
of
the
job
re
com
m
end
er
sy
stem
.
Su
c
h
syst
em
can
be
em
power
e
d
with
li
felo
ng
le
ar
ning
to
to
f
os
te
r
the
c
on
ti
nuous
de
velo
pm
e
nt
an
d
i
m
pr
ovem
ent o
f
the
kn
ow
le
dge an
d
s
kill
s n
ee
ded f
or em
plo
ym
ent an
d pe
rs
onal
fulfil
m
ent [
11
]
.
Figure
5
.
The
pro
po
se
d j
ob re
com
m
end
er
s
yst
e
m
w
orkf
l
o
w
V
i
s
u
a
l
i
z
a
ti
o
n
M
a
tc
h
i
n
g
J
o
b
see
ke
r
R
esum
e
C
o
m
p
an
ies
J
o
b
adve
r
tise
m
en
t
Grad
u
ate
Skil
l
A
cqu
ired
Co
m
p
an
y
Skil
l
N
ee
d
ed
Au
to
mati
c
ke
y
w
o
r
d
ex
t
r
a
c
ti
o
n
Ran
king
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Jobseek
er
-
i
ndust
ry matc
hing
s
yst
em usin
g aut
omated
keyw
ord sele
ct
io
n and
...
(
Nor
hasli
nda K
ama
rud
din
)
1129
Figure
6
.
Exa
m
ple o
f
jo
b
r
ec
omm
end
er
w
e
bs
it
e b
ase
d o
n visuali
zat
ion a
ppr
oach
ACKN
OWLE
DGE
MENT
The
a
uthors
would
li
ke
to
than
k
Un
i
versi
ti
Teknolo
gi
MARA
(U
iT
M),
I
nter
natio
nal
Islam
ic
Un
i
ver
sit
y
Ma
la
ysi
a
(I
I
UM)
and
Mi
nistry
of
Hi
gh
e
r
E
du
cat
ion
Ma
la
ysi
a
(
MO
HE
)
for
pro
vid
i
ng
fin
ancial
su
pp
or
t
th
rou
gh
the
Tra
nsdis
ipli
nar
y
Re
search
Gr
a
nt
Schem
e
T
RGS
(
T
RGS16
-
04
-
02
-
002
)
to
co
nduc
t
the
work pu
blishe
d
in
this
pa
per
.
REFERE
NCE
S
[1]
B.
Ahm
ad,
“
La
pora
n
Kaj
ian
Pengesa
nan
Gradua
n,
”
2
016.
Avail
ab
le:
htt
ps://
ww
w.
m
ohe.
gov.
m
y
/m
uat
-
turun/
awa
m
/stati
stik/
190
-
la
por
an
-
kaj
i
an
-
penge
s
a
nan
-
gra
duan
.
[2]
Garwe
E
.
C.
,
“
Inc
rea
se
in
the
Dem
and
for
Pri
vat
e
Higher
Ed
uca
t
ion:
Unm
asking
the
“
Para
dox”
,”
Int
ernati
o
nal
Journal
of
Edu
c
ati
onal Manage
ment
,
vo
l/
issue:
30(2)
,
pp
.
232
-
51
,
2016
.
[3]
“
Jobs
tre
et
.
com,
”
htt
ps://
ww
w.j
ob
stree
t
.
com.m
y
/en/job
-
se
ar
ch/
job
-
v
ac
an
c
y
.
php
?
ke
y
=
ic
t
&l
o
ca
t
ion=&
spec
ializati
on
=
1
91%
2C192%2C
193&a
re
a=
&sa
l
ar
y
=&
o
js=3&
src
=12.
[4]
htt
ps://
ww
w.l
ink
edi
n.
com
[5]
htt
ps://
ww
w.m
onster.
com.m
y
/
[6]
Zha
ng
C.
,
“
Au
t
om
at
ic
Ke
y
wor
d
Ext
ra
ct
ion
fr
om
Docum
ent
s
using
Condit
io
nal
Random
Fi
el
ds
,”
Journal
of
Computati
onal I
nformation
Syst
e
ms
,
vol/
issue:
4(
3)
,
pp
.
1169
-
80
,
2008.
[7]
Bhart
i
S
.
K
.
and
Babu
K
.
S.
,
“
Autom
at
ic
Ke
y
wo
rd
Ext
ra
ct
ion
for
Te
xt
Sum
m
ari
z
at
ion:
A
Surve
y
,”
ArXiv
preprint
arXiv
:1704.
0324
2
,
2017
.
[8]
Am
at
o
F
.
,
et
al
.
,
“
Cla
ss
ifica
t
io
n
of
W
eb
Job
Advert
isements:
A
Case
Stud
y
,”
Proceedi
ngs
of
the
23
rd
Itali
an
Symposium on
A
dvanc
ed
Databa
se
Syste
ms
(
SEBD
2015)
.
Gae
t
a,
Ita
l
y
,
pp
.
144
-
15
1
,
2015
.
[9]
Hauff
C
.
and
G
ousios
G.
,
“
Mat
chi
ng
GitHub
Deve
lop
er
Profil
e
s
to
Job
Advert
i
sem
ent
s
,”
Pro
ceedi
ngs
of
the
12
th
Working
Conf
ere
nce on
M
ini
ng
Soft
ware
Re
posit
ories.
Floren
ce, Ita
l
y
,
pp
.
362
-
36
6
,
2015
.
[10]
Muth
y
a
la
R
.
,
e
t
al.
,
“
Data
-
driv
en
Job
Sear
ch
Eng
ine
Us
ing
Skil
ls
and
Com
pan
y
A
tt
ri
but
e
Fil
te
rs
,”
Proce
ed
ings
of
IEE
E
Inte
rnat
io
nal
Conf
ere
nce
on
Data
Min
ing Wor
kshops (
ICD
MW)
,
Lousia
na
,
US
,
pp.
199
-
206
,
2017
.
[11]
Ham
iz
M
.
,
et
al.
,
“
As
sessment
Anal
y
tic
Theoret
i
ca
l
Fram
e
work
Based
on
Le
arn
e
rs’
Conti
nuous
Le
arn
in
g
Im
prove
m
ent
,”
Indone
sian
J
ournal
of
E
lectric
al
Eng
ine
e
ring
and
Computer
Sc
ie
nc
e
,
vol/
issue:
11(
2)
,
pp.
682
-
687
,
20
18.
[12]
Muth
y
a
la
R,
W
ood
S,
Jin
Y,
Q
in
Y,
Gao
H,
Rai
A.
Data
-
dr
iven
Job
Sear
ch
Engi
ne
Us
ing
Skill
s
and
Com
pany
Attri
bute
Filt
ers.
Proce
edi
ngs
of
IEE
E
In
te
rn
at
io
nal
Confer
ence
on
Data
Mining
W
orkshops
(IC
DM
W
)
,
Lousia
na,
US
.
2017
;
199
-
2
06
.
[13]
La
wrenc
e
S,
Le
e
Gile
s
C,
Boll
a
cke
r
K.
Digi
ta
l
li
br
aries
an
d
aut
onom
ous
ci
ta
ti
on
ind
exi
ng
.
IEE
E
Journals
on
Com
pute
r
.
vo
l.
32
,
no
.
6
,
pp
.
6
7
-
71,
June
1999.
[14]
Yih
W
T
,
Good
m
an
J
,
Carva
lh
o
V
R
.
Finding
adve
rti
sing
ke
ywords
on
web
p
age
s.
In
Proc
ee
d
ings
of
the
15th
int
ern
at
ion
al
con
fe
renc
e
on
Worl
d
Wide We
b
(
WWW
'06)
.
ACM,
New Y
ork
,
N
Y
,
USA
,
2006
,
213
-
222.
[15]
Ham
iz
M,
B
akr
i
M,
Ka
m
aru
ddin
N,
Moham
ed
A.
As
sessment
Anal
y
ti
c
Th
eor
etic
al
Fram
ework
Based
on
Le
a
rne
rs
’
Conti
nuous
Lea
rning
Im
prove
m
ent
.
Indone
si
an
Journal
of
El
e
ct
ri
cal
En
gine
ering
and
Computer
Sci
e
n
ce
.
2018;11(2):
682
-
687.
Evaluation Warning : The document was created with Spire.PDF for Python.