TELKOM
NIKA
, Vol.14, No
.1, March 2
0
1
6
, pp. 349~3
6
0
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.3257
349
Re
cei
v
ed
No
vem
ber 1
2
, 2015; Re
vi
sed
Jan
uar
y 7, 20
16; Accepted
Jan
uary 26, 2
016
Determining Process Model Using Time-Based Process
Mining and Control-Flow Pattern
Riy
a
narto Sarno*, Wid
y
asari A
y
u Wibo
w
o
,
Kar
t
ini, Yutika Ameli
a
, Kelly
Rossa
Jurusan T
e
knik Informati
ka, F
a
kultas T
e
knol
ogi Informas
i
, Institut
T
e
knolo
g
i Sep
u
lu
h No
pemb
e
r,
Kampus IT
S Sukoli
lo, Jal
an R
a
ya IT
S,
Suraba
ya, Ja
w
a
T
i
mur 601
11, (03
1
)
5994
25
1
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: ri
y
a
narto
@if.its.ac.id
A
b
st
r
a
ct
Deter
m
in
in
g ri
ght
mod
e
l
of b
u
sin
e
ss proc
es
s from
eve
n
t lo
g is th
e pur
pos
e of pr
ocess
di
scovery.
How
e
ver so
me pro
b
l
e
ms i.
e the
ina
b
i
lity
to disc
over
OR, noise
an
d eve
n
t lo
g i
n
co
mp
leten
e
ss
ar
e
emmerg
ed w
h
i
l
e
deter
mi
nin
g
rig
h
t
mod
e
l
of
bus
iness
pr
oc
ess. F
i
rst, OR rel
a
tion
is
often
discov
e
re
d
as
AND rel
a
tio
n
. Secon
d
, no
ise
prob
le
m is
oc
cured w
h
e
n
th
ere ar
e trunc
ated a
nd l
o
w
freque
ncy traces
i
n
event lo
g.
T
h
u
s
contro
l-flow
pattern is use
d
to
s
o
lv
e iss
ues
of sa
me
n
o
ise
rel
a
tio
n
fr
equ
ency
he
nc
e it
discov
e
rs re
lati
on
bas
ed
on
transacti
on
funct
i
on
of
acti
vity.
Cons
equ
ently,
it can
refi
ne
no
n n
o
ise
re
latio
n
i
n
busi
ness
proc
ess
mo
del. T
h
ird, i
n
co
mp
lete
ness
lea
d
s to
incorrect
disc
o
v
ery of
par
all
e
l pr
ocess
mod
e
l;
therefore w
e
used T
i
me
d-b
a
sed Pr
oc
ess
Mining w
h
ic
h
utili
z
e
d
non-
li
near d
epe
nd
e
n
ce to solve
the
inco
mplete
nes
s. F
i
nally
this
p
aper
pro
pos
ed
combi
natio
n
of T
i
med-b
a
se
d
Process Mi
ni
n
g
a
nd c
ontro
l-flow
pattern to disc
over OR an
d han
dle s
a
me frequ
ency n
o
is
e and
inco
mpl
e
teness. F
r
o
m
the exper
i
m
e
n
t i
n
section 3, this
prop
osed
meth
od man
a
g
e
s to
get right proc
e
ss mod
e
l fro
m
event lo
g.
Ke
y
w
ords
: Co
nditi
ona
l OR, Control-F
l
ow
Pattern, Incomplet
eness, No
ise, T
i
me
d-b
a
sed P
r
ocess Mini
ng
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Determinin
g model of business
process from event
log is t
he main purpo
se of process
discovery. Proce
s
s discov
ery is a chall
engin
g
task
in pro
c
e
ss mi
ning. It is a set of techniq
ues
whi
c
h a
u
tom
a
tically con
s
truct
a mo
del
of an o
r
ga
n
i
zation
`s current activities and its maj
o
r
activities vari
ations. The
s
e techniq
u
e
s
use event
log of activities within a
n
orga
nization.
The
busi
n
e
ss
pro
c
e
ss
mod
e
l i
s
an
alyze
d
to
sho
w
th
e co
mplexity of issue
s
in
a
c
tivities an
d ho
w to
solve them. T
hese issu
es
exist in any field, e.
g. bu
si
ness [1], envi
r
onm
ent [2, 3
]
smartp
hon
e
[4],
and frau
d [5].
Process discovery come
s up with man
y
algorit
hms,
e.g. alpha, alpha
+, alph
a++ [6]. The
alpha, al
pha
+, and alp
h
a
+
+ cann
ot deal
with n
o
is
e, i
n
com
p
leten
e
ss i
s
sue
s
an
d OR conditi
onal.
Heuris
tic
miner algorithms [7, 9] c
o
me up to s
o
lve
the noise pro
b
lem. Ho
wev
e
r, mo
st of the
algorith
m
s a
r
e unable to find OR
con
d
itional mod
e
l. The existing
algorith
m
fre
quently disco
v
ers
the OR
con
d
itional as A
N
D pa
rallel o
r
XOR condi
tio
nal. The thou
ght of parall
e
l
model di
sco
v
ery
will change the result of activities [8].
Whe
n
“wait a
nd se
e” b
eha
vior model
sy
nch
r
oni
zatio
n
is
occured, it n
eed
s O
R
p
a
rallel to m
odel
the pa
ra
ll
el
split a
nd joi
n
. The
“wait an
d see” be
hav
ior
model
synch
r
onization o
c
cured
whe
n
th
e acto
r ca
n c
hoo
se only o
ne activity, all activity, or more
than o
ne a
c
ti
vity in parall
e
l split and
j
o
in. In thi
s
p
aper we p
r
o
posed i
dea
s
to discove
r
OR
con
d
itional wi
thin busi
n
e
ss
pro
c
e
ss m
o
d
e
l.
One of im
po
rtant things from
process
mining i
s
the
idea of
com
p
l
e
tene
ss whi
c
h
is relate
d
to noise. Inco
mpletene
ss lead
s to false
parallel
rel
a
tions di
scovery, e.
g
the discovered pa
ra
llel
relation
is XO
R b
u
t the
righ
t parall
e
l
relat
i
on in
bu
sine
ss process i
s
OR. T
he
new rep
r
e
s
e
n
tatio
n
of OR-split u
s
e
s
combi
nat
ion the exi
s
ting XO
R-sp
lit
and A
N
D-spli
t to make
the
model
ea
sier to
be an
alyze
d
[13]. In other hand
temp
o
r
al a
c
tivity-ba
s
ed
algo
rith
m [8] and
co
ntrol-flo
w
p
a
ttern
can han
dle discovery of
busi
n
e
ss pro
c
e
ss mod
e
l with
in
com
p
letene
ss and
sam
e
fre
q
u
ency
noise issue
s
. Non-li
nea
r depe
nden
ce
in temporal
activity-based
algorithm is used to sol
v
e
incom
p
leten
e
s
s p
r
obl
em
since
it
can
di
scover
mo
re relation
than
linear
de
pen
den
ce. Cont
rol-
flow p
a
ttern i
s
u
s
e
d
to
so
lve sam
e
a
m
ounts of n
o
ise freq
uen
cy i
s
sue
s
b
e
cau
s
e it
discove
r
s
relation b
a
se
d on tran
sa
ction function
of activity,
therefo
r
e it can
choo
se no
n noise relatio
n
in
busi
n
e
ss p
r
o
c
e
ss m
odel.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 349 – 3
5
9
350
2. Rese
arch
Metho
d
In
Figu
re 1 we de
scribe
the
propo
se
d
method
to
discover the
right m
odel
i
n
bu
sin
e
ss
pro
c
e
ss.
On
e of the m
e
thod
s u
s
ed
in this
re
search i
s
mo
dified tempo
r
al a
c
tivity-base
d
algorith
m
. T
e
mpo
r
al
acti
vity-based
al
gorithm
is modified
to
discover con
d
itional OR and
overcome noi
se and in
com
p
letene
ss. The modificatio
n
is done by usin
g non
-lin
ear de
pen
de
nce
to deal with
incom
p
leten
e
ss a
nd modif
i
ying parall
e
l
relation to d
i
stingui
sh pa
rallel AND a
n
d
con
d
itional O
R
. The non
-li
near d
epe
nd
ence utilize
s
doubl
e timest
amped eve
n
t log to disco
ver
seq
uen
ce a
n
d
co
ncurrent
relation i
n
a
ca
se in
eve
n
t
log; whe
r
ea
s line
a
r d
epe
nden
ce utili
zes
singl
e timest
amped
event
log to disco
v
er only seq
uen
ce relatio
n
in a
case i
n
event log [
8
].
Finally
the di
scovere
d
m
o
del can be matche
d with
formal
control-flow patte
rn existe
d in
real
busi
n
e
ss p
r
o
c
e
ss m
odel. In Figure 1 we
expl
ained a
n
a
lytical step
s
of propo
se
d method.
Figure 1. Flow ch
art of pro
posed metho
d
2.1. Modifica
tion
Timed-ba
s
e
d
Process M
i
ning Algorithm
Figure 2. Modification Tim
ed-b
a
sed Pro
c
e
ss Mini
ng
Algorithm
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Determ
inin
g Process Mod
e
l Usi
ng Tim
e
-Based Pro
c
e
ss Mini
ng an
d… (Ri
y
a
n
a
r
to Sarno
)
351
The Time
d-b
a
se
d Pro
c
e
s
s Minin
g
alg
o
rithm i
s
introdu
ced by
Rizka et. al. [8]. The
algorith
m
is created
ba
se
d
on
several d
e
finitions to
n
o
te the
relati
on bet
wee
n
a
c
tivities in
event
log e.g b
e
fore
and me
ets, o
v
erlap
s
, co
ntains, i
s
-f
ini
s
h
ed-by, eq
ual
s, and sta
r
ts [8
]. Howeve
r, the
algorith
m
can
not distin
gui
sh O
R
o
r
A
N
D relation
s. T
hus th
e alg
o
rithm is mo
dified by ad
ding
a
step to
distin
guish O
R
a
n
d
AND relati
ons (sho
wn
i
n
ste
p
4
-
8).
The m
odifica
tion algo
rith
m is
written in Fig
u
re 2.
2.2. Con
t
rol-
flo
w
Patter
n
s
In busin
ess
pro
c
e
ss m
o
d
e
ls the
r
e a
r
e
often use
d
pro
c
e
ss fo
rm
s whi
c
h
are
called
Control-flo
w
pattern
s. The
r
e a
r
e 2
0
p
a
ttern
s availa
bl
e for u
s
e
d
in
multi model
d
i
agra
m
ba
se
d
on
Aalst pape
r [10]. Those
pattern
s are
too abstra
c
t for practical use in b
u
sin
e
ss process
modellin
g, th
us
we
tried
to
formali
z
e
the
patterns u
s
a
ge in
two
ste
p
s, T
op
Level
Abstractio
n
and
Low Level Execution.
T
op
Level
A
b
stra
ction
d
e
fines
bu
sin
e
ss process model in
multi
orga
nization
s platform wh
ile Low
Lev
el Abstra
ctio
n is u
s
ed to
discover th
e real a
c
tivities
perfo
rmed in
side an organi
zation.
2.2.1.
Top Lev
e
l A
b
stra
ction
In real life ap
plicatio
n even
t-logs of a co
mp
lete bu
sin
e
ss pro
c
e
s
s hardly can be
found in
one data
b
a
s
e. Each emb
r
oiled o
r
g
ani
zation
kee
p
s their own d
a
ta in sep
a
rate databa
se
s to
ensure
data
se
curity. Thu
s
to obtain a
compl
e
te
vie
w
poi
nt in bu
siness p
r
o
c
e
s
s mod
e
l we
n
eed
to define the
form of part
nership
between o
r
gani
za
tions pa
rtici
p
ated in bu
sin
e
ss process.
The
definition of partne
r
ship f
o
rm bet
wee
n
orga
nizati
o
n
is calle
d To
p Level Abst
ractio
n. It is only
explained th
e frame
w
ork of a busine
ss p
r
o
c
e
ss b
u
t not nece
s
sarily explai
n
ed the activities
whi
c
h a
r
e ex
ecute
d
by a certain o
r
ga
nization. Ps
eu
d
o
co
de for T
o
p Level Abst
raction i
s
liste
d in
Figure 3.
There a
r
e fo
ur type
s of
partne
r
ship
whi
c
h i
s
reg
u
larly a
pplie
d in multi
-
organi
zatio
n
b
u
s
in
es
s
pr
oc
es
s
:
a. Con
s
e
c
utive
Partnershi
p
b. Substitutive
Partne
rship
c. Compl
e
me
ntory
Partne
rsh
i
p
d. Subprocess
Partnershi
p
Pseud
ocode
Rule
Defining
Top Level Ab
stra
ction i
s
explained in
Figure 3.
Figure 3. Rul
e
of Top Leve
l
Abstra
ction
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 349 – 3
5
9
352
2.2.2. Lo
w
Lev
e
l
Execu
tion
Definin
g
form of patnership b
e
twe
e
n
embroiled
orga
nization
s only explained oute
r
frame
w
ork
of busi
n
e
s
s pro
c
e
s
s mod
e
l. Thus we n
e
e
d
to loo
k
u
p
i
n
sid
e
ea
ch
organi
zation
s a
n
d
model exe
c
ut
ed activities
of real bu
sin
e
ss pro
c
e
s
s, this step call
ed Low L
e
vel
Execution. The
model i
s
in
correctly di
scovered
whil
e
it contain
s
the noi
se rel
a
tion. The
r
ef
ore, control-fl
ow
pattern of bu
siness process is employed
to confir
m the
discovere
d
relation in bu
si
ness process.
We a
c
knowl
e
dge Aalst con
t
rol-flo
w
patterns [10]
a
s
theoreti
c
al gui
d
ance. We defi
ne 8 out of
20 availa
ble
pattern
s [11
]
are a
b
le to
help in
de
ciding p
a
rall
el
model
whi
c
h is limite
d
f
o
r
orga
nizationa
l collabo
ratio
n
. Those the
8 patterns a
r
e expresse
d
in formal an
d grap
hical way.
B
a
sic P
a
t
t
e
rn
s
This g
r
ou
p of
pattern
s cont
ains el
eme
n
tary as
pe
cts o
f
workflow
proce
s
s. Tho
s
e
pattern
s are
listed bel
ow:
1) Sequen
ce
The p
a
ttern
as
se
en i
n
Fi
gure
4, d
e
fin
e
s
simpl
e
st f
o
rm
of a
c
tivity executio
n i
n
control
-
flow. Sequ
en
ce
de
scribe
s an
activity in workfl
ow is ena
bled
after the
co
mplet
i
on of it
s inp
u
t
ac
tivity in the
s
a
me proc
ess
.
1
0
2
a
b
Figure 4. Sequen
ce pattern
2) Parallel
Split
The
pattern
as
se
en i
n
F
i
gure
5, d
e
fin
e
s
splitting
p
o
int in
a
workflow where
a si
ngle
thread
of pro
c
e
ss
divised i
n
to two o
r
m
o
re b
r
a
n
ch
es
of
pro
c
e
ss control whi
c
h can be
exe
c
u
t
ed
silmultan
eou
sly in any orde
r.
Figure 5. Parallel split patt
e
rn
3) Synchroni
zati
on
The patte
rn a
s
seen in Fi
g
u
re 6, defin
e
s
a poi
nt in a
workflo
w
wh
ere multipl
e
pro
c
e
ss
conve
r
ge
int
o
on
e
singl
e
process, th
us
syn
c
h
r
oni
zation
of m
u
ltiple thread
of p
r
o
c
e
s
s is
happ
ened.
Figure 6. Synchroni
zation
pattern
4) Exclusive
Ch
oice
The pattern as se
en in Fig
u
re 7, define
s
a point in a
workflo
w
wh
e
r
e ba
sed on
deci
s
io
n
control only o
ne out of som
e
n multiple b
r
an
che
s
is ex
ecute
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Determ
inin
g Process Mod
e
l Usi
ng Tim
e
-Based Pro
c
e
ss Mini
ng an
d… (Ri
y
a
n
a
r
to Sarno
)
353
Figure 7. Exclusive choi
ce
pattern
5) Simple
Merge
The pattern
as seen i
n
F
i
gure
8, describe
s
a workf
l
ow when two or mo
re b
r
anche
s
conve
r
ge into
a single process thre
a
d
wi
thout synchro
n
izatio
n. The
assumptio
n
o
f
this pattern is
none of multi
p
le bra
n
che
s
is ever exe
c
u
t
ed simultan
e
ously.
Figure 8. Simple merge pat
tern
a) Advan
c
e Condition
al Bra
n
chi
ng Patterns
This
gro
up of
pattern
s i
s
sl
ightly more
complex than
bran
chi
ng p
a
tterns in b
a
si
c pattern,
yet the p
a
tterns li
st i
n
thi
s
grou
p
are
u
s
ually o
ften
u
s
ed in
real life
pro
c
e
s
ses.
T
hose p
a
ttern
s are
listed bel
ow:
1) Multi
Choi
ce
The pattern as se
en in Fig
u
re 9, is an i
m
prov
em
ent of Exclusive choi
ce patte
rn. In Multi
choi
ce p
a
ttern bran
ch
es a
r
e able to be
executed in
parall
e
l or se
quentially de
pendi
ng on t
he
deci
s
io
n of executio
n time.
Figure 9. Multi choi
ce patte
rn
2) Synchroni
zin
g
Merg
e
The pattern
as se
en
i
n
Figure
1
0
,
combine
s
bet
wee
n
Synchronization an
d
Simple
Merg
e pattern, where the deci
s
io
n of sychroni
za
tion
or merge p
r
o
c
e
ss d
epe
nd
s on exe
c
utio
n.
Figure 10. Synch
r
oni
zin
g
merg
e pattern
b) Stru
ctural
Pattern
The pattern a
s
se
en in Fig
u
re 11, de
scri
bes different restri
ction in workflo
w
mo
de
ls su
ch
as
c
y
c
l
es
pattern. In this
sec
t
ion, we pres
ent
a patte
rn
which rep
r
e
s
ent typical
workflo
w
manag
eme
n
t system
s structural re
stri
ctio
ns.
1) Arbitra
r
y
cycl
es
Arbitra
r
y cycl
es define
s
a
point in a
work
flo
w
wh
ere on
e or more a
c
tivities can b
e
execute
d
rep
eatedly.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 349 – 3
5
9
354
Figure 11. Arbitrary cy
cle
s
pattern
The rul
e
use
d
to define the model i
s
sim
ila
r with
Context Sen
s
itive Gram
m
a
r [12].
Pseud
ocode
Rule
Defining
Low Level E
x
ecution i
s
explaine
d in Fig
u
re 12.
Figure 12. Ru
le of Low Lev
el Execution
3. Results a
nd Analy
s
is
This research use
s
do
u
b
le time-sta
mped event
log contai
n
i
ng more than one
orga
nization
executin
g a
c
tivities in eve
n
t log.
The in
formation
attributes
co
ntai
ned in eve
n
t log
are the n
u
m
ber of case id, the activity in proc
ess,
time-stam
p
of activity executio
n, and
the
orga
nization
executin
g the
activity.
Figure 13. YAWL mo
del of stand
ard b
u
si
ness process
Busine
ss p
r
o
c
e
s
s on
Figu
re 13
exe
c
ute
d
a
s
in
event
log i
s
a p
r
o
c
ess to
buy g
ood
s for
prod
uctio
n
from suppli
e
r
by a comp
an
y. There
ar
e
two
kind
of supplie
r. The
first
sup
p
lier i
s
in
bond
ed
zon
e
of Cu
stom
s a
nd Exci
se
an
d the
se
con
d
is o
u
t of it. T
he
comp
any
sen
d
s pu
rcha
se
0
1
2
3
a
b
c
a
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Determ
inin
g Process Mod
e
l Usi
ng Tim
e
-Based Pro
c
e
ss Mini
ng an
d… (Ri
y
a
n
a
r
to Sarno
)
355
orde
r
numb
e
r
to
su
pplie
r to o
r
de
r g
o
od fo
r p
r
od
u
c
tion
(a
ctivity A). Th
en, t
he first
sup
p
lie
r
prod
uces
(a
ctivity
B), packag
e
s
(a
ctivity C), and
se
nds the
goo
d
orde
r to co
mpany (a
ctivity I)
;
whe
r
ea
s the
se
con
d
su
ppl
ier send
s pe
rmitting doc
u
m
ent to Cu
st
oms a
nd Exci
se for
req
u
e
s
ting
transactio
n
a
pprove
m
ent betwe
en su
p
p
lier and
co
mpany
(a
ctivity D). T
hen,
Cu
stom
s
a
n
d
Excise dete
r
mines tax a
nd app
rove
the transac
ti
on (a
ctivity
E). Furtherm
o
re, the se
cond
sup
p
lier p
a
ys the tax (acti
v
ity F), produ
ce
s (a
ct
ivity
G), pa
ckage
s (activity H), and send
s g
ood
orde
r (a
ctivity I). Finally, company
re
ceiv
es the go
od (activity J).
The event log
which is u
s
in this
experi
m
ent is prese
n
ted in Table
1.
Table 1. Singl
e time stamp
ed event log
of busin
ess p
r
ocess
We provide a
n
example of
Proce
s
s Di
sc
overy u
s
ing
Timed-ba
sed
Proce
s
s Min
i
ng and
Control-Fl
ow
Patterns a
s
written belo
w
.
3.1.
Step 1:
Disc
ov
ering Proc
ess Mod
e
l
The first step
is cl
assify the se
quen
ce
a
nd pa
rallel rel
a
tion
a
c
tivities
from every trace
i
n
eventlog. After
cla
ssifying
relatio
n
a
c
tivities fr
o
m
eve
r
y tra
c
e, me
rge it into
a
seque
nce relat
i
on
activities an
d a parall
e
l rela
tion activities.
T
he merging
relation
s are descri
bed in
Figure 14.
Figure 14. Mergin
g rel
a
tions
After mergi
n
g the relatio
n
activities,
then
cl
assify parallel rela
tions to
be
AND
or
con
d
itional O
R
. The cl
assif
i
es are de
scri
bed in Figu
re
15.
Figure 15. AND o
r
Co
nditi
onal O
R
relati
ons
⊕
⊕
,
⊕
• = { F •
G
}
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 349 – 3
5
9
356
The third
ste
p
is form a g
r
aph b
a
sed o
n
the
relation
activities in Figure 14 an
d Figure
15. The grap
h can b
e
mod
e
led a
s
Figu
re 16.
Figure 16. Final gra
p
h
3.2.
Step 2:
Disc
ov
ering Standard Proce
s
s
Model usin
g Control-flo
w
P
a
tterns
There are 3
orga
nization
s partici
pated i
n
whol
e bu
si
ness p
r
o
c
e
s
s :
α
,
β
, and
γ
Ac
tivity
A, I, J is
don
e
by organi
zati
on
α
, a
c
tivity B and
C i
s
do
ne by
org
ani
zation
β
, and
ac
tivity D, E, F
,
G, H is d
o
n
e
by orga
nization
γ
.
α
i
s
the main o
r
gani
zation
re
spo
n
si
ble for whole
bu
sin
e
ss
p
r
oc
es
s
.
α
shares
its resp
onsi
b
ilities wi
th
β
and
γ
to
do the
same
goal : supply
good
s
whi
c
h
is
requ
este
d in
activity A. Th
e form of pa
rtnership b
e
tween o
r
ga
niza
tion
α
an
d
β
,
γ
is d
e
fined
as
sub
s
titutive partne
r
ship. Thus
we can o
b
tain Top Lev
el Abstra
ction
model as Fi
g
u
re 17.
Figure 17. To
p Level Abstraction mo
del
with OR g
a
te
Activity A (purch
asi
ng o
r
d
e
r) i
s
a
sta
r
ting a
c
tivity in busi
n
e
ss
pro
c
e
ss
and foll
owe
d
b
y
activity B (first su
pplie
r) a
nd a
c
tivity D (se
c
o
nd
sup
p
lier). Activity B and a
c
tivity D sha
r
e t
h
e
same
inp
u
ts
and o
u
tputs,
and
com
pany
can
option
a
ll
y choo
se
bet
wee
n
exe
c
uti
ng a
c
tivity B or
D
or both
(si
n
ce
activity B and activity D are sam
e
, to ch
oose su
pplie
r type). Thus
activity B and D
use
Multi
Cho
i
ce
pattern
(see Fig
u
re
18
). Activity
B followe
d
by a
c
ti
vity C re
sultin
g a
c
tivity B and
C use Sequ
e
n
ce p
a
ttern
(see Fi
gure 1
9
). Activity D
is followed by
activity E,
those two a
c
tivitie
s
define a
s
Se
quen
ce
patte
rn. Next a
c
tivities are
a
c
tivity F (paying t
a
xes)
and
G(prod
uce go
o
d
s).
Activity F
and activity G
must be do
n
e
to co
mplet
e
the orde
r, and both a
c
tivities could
be
execute
d
sim
u
ltaneo
usly, resultin
g a
c
tivity F and
G
a
s
Parall
el p
a
ttern
(see
Fig
u
r
e
20). A
c
tivity I
(se
nd goo
ds)
and activity
J
(re
c
eive go
ods)
a
r
e
exe
c
uted
se
que
n
t
ially. We can
define th
e L
o
w
Level Executi
on model fo
r each org
ani
zation as b
e
lo
w.
Figure 18. Lo
w Level Execution model o
f
α
with OR gate
Figure 19. Lo
w Level Execution model o
f
β
without any parallel gat
e
α
β
γ
α
V
V
A
β
γ
I
V
V
J
B
C
α
α
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Determ
inin
g Process Mod
e
l Usi
ng Tim
e
-Based Pro
c
e
ss Mini
ng an
d… (Ri
y
a
n
a
r
to Sarno
)
357
Figure 20. Lo
w Level Execution model o
f
γ
with AND
gate
The form
alization of this busi
n
e
ss p
r
o
c
e
ss m
odel i
n
Figure 13
will be prese
n
ted as
follow.
F
o
rm
a
lization
for st
anda
rd
proce
dure
of
pro
c
e
s
s is ne
ede
d
as
rig
h
t gui
da
nce
in
process
executio
n. We rep
r
e
s
ente
d
its
stand
ard
formalization
as belo
w
:
W = {
S, A, N,
∑
,
α
, E
} where:
S is
a Start ac
tiv
i
ty
in
workflow proc
ess
A is
a finite s
e
t of Ac
tiv
i
ties
N
A is
a finite s
e
t of Nodes
∑
is
finite s
e
t of tas
k
inputs/outputs
α
is the Nod
e
transition fun
c
tion
E is an End a
c
tivity in
workflow pro
c
e
ss
S = {A}
E =
{J
}
A =
{A, B, C,
D, E, F,G, H,
I, J
}
N = {n
A
, n
B
, n
C
, n
D
, n
E
, n
F
, n
G
, n
H
, n
I
, n
J
}
∑
=
{a, b, c
,
d, e, f
,
g, h, i
,
j,
k
}
α
= {
Sequen
ce
[n
B
, c
,
n
C
],
[n
D
, d, n
E
],
[n
I
, i
,
n
J
],
Parallel
split & synchroni
zation
[n
E
, e, n
F
n
G
],
[n
F
n
G
, f, n
H
]
Multi Choi
ce
[n
A
, a, n
B
]
,
[n
A
, b, n
D
],
[n
A
, j,
n
B
n
D
]
Synchroni
zin
g
Merg
e
[n
C
, g, n
I
]
,
[n
H
, h,
n
I
], [n
C
n
H
, k
,
n
I
]
}
Figure 21. Fo
rmali
z
ed g
r
ap
h of depen
de
ncy gra
p
h
In step
1
we
obtaine
d relat
i
on b
e
twe
en
activity as
sh
own
in Fi
gu
re
21, thu
s
we
derived
a
stand
ard
formalizatio
n a
n
d
comp
are
th
is fo
rmali
z
ati
on a
nd fo
rma
lization
of
sta
ndard p
r
o
c
e
d
u
re
to verify whe
r
eas the
r
e’
s n
o
ise
in
obtain
ed m
odel
or
not. Late
r
in
t
h
is
step
n
o
ise is d
e
leted
from
obtaine
d mod
e
l.
W = {
S, A, N,
∑
,
α
, E
} where:
S = {A}
E =
{J
}
A =
{A, B, C,
D, E, F,G, H,
I, J
}
N = {n
A
, n
B
, n
C
, n
D
, n
E
, n
F
, n
G
, n
H
, n
I
, n
J
}
∑
=
{a, b, c
,
d, e, f
,
g, h, i
,
j,
k
,
x}
α
= {
Sequen
ce
[n
B
, c
,
n
C
],
[n
D
, d, n
E
],
[n
I
, i
,
n
J
], [n
c
, x
,
n
d
],
D
E
α
F
G
H
V
V
α
A
B
D
E
F
G
H
I
J
a
b
C
c
d
e
e
f
f
h
g
i
x
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 349 – 3
5
9
358
Parallel
split & synchroni
zation
[n
E
, e, n
F
n
G
],
[n
F
n
G
, f, n
H
]
Multi Choi
ce
[n
A
, a, n
B
]
,
[n
A
, b, n
D
],
[n
A
, j,
n
B
n
D
]
Synchroni
zin
g
Merg
e
[n
C
, g, n
I
]
,
[n
H
, h,
n
I
], [n
C
n
H
, k
,
n
I
]
}
Relatio
n
of [
n
C
, x, n
D
] is not availabl
e
in rel
a
tion
set
of stan
dard procedu
re,
thus thi
s
relation i
s
lab
e
led a
s
noi
se
and delete
d
from obtai
ned
model a
s
se
e
n
in Figure 22
.
Figure 22. Di
scovere
d
mo
del with pa
rall
el gate
The form
aliza
t
ion of busin
e
ss p
r
o
c
e
ss o
b
tained in Fig
u
re 22 i
s
written as b
e
lo
w,
W = {
S, A, N,
∑
,
α
, E
} where:
S = {A}
E =
{J
}
A =
{A, B, C,
D, E, F,G, H,
I, J
}
N = {n
A
, n
B
, n
C
, n
D
, n
E
, n
F
, n
G
, n
H
, n
I
, n
J
}
∑
=
{a, b, c
,
d, e, f
,
g, h, i
,
j,
k
,
x}
α
= {
Sequen
ce
[n
B
, c
,
n
C
],
[n
D
, d, n
E
],
[n
I
, i
,
n
J
]
Parallel
split & synchroni
zation
[n
E
, e, n
F
n
G
],
[n
F
n
G
, f, n
H
]
Multi Choi
ce
[n
A
, a, n
B
]
,
[n
A
, b, n
D
],
[n
A
, j,
n
B
n
D
]
Synchroni
zin
g
Merg
e
[n
C
, g, n
I
]
,
[n
H
, h,
n
I
], [n
C
n
H
, k
,
n
I
]
}
We
com
pare
the form
alization in thi
s
step
wi
th formalizatio
n of
stand
ard
pro
c
ed
ure
in
step 2 a
nd o
b
taining the
same result. If there’
s a differen
c
e
betwe
en form
alizati
on of stan
da
rd
pro
c
ed
ure a
nd obtai
ned
model, the
n
the form
aliz
a
t
ion of sta
n
d
a
rd
pro
c
e
d
u
r
e mod
e
l is the
pre
c
ise one.
4. Conclusio
n
In this pa
per, we
propo
sed m
e
thod
s
to
dete
r
mine
process
mo
del by
timed
-
ba
se
d
pro
c
e
ss
mini
ng an
d control-flow
patterns. First,
we
discover se
quen
ce relati
on
a
c
tivities and
parall
e
l relati
on (AND
o
r
OR co
ndition
al) activi
ties
by timed-ba
sed p
r
o
c
e
s
s
mining. After we
discover relat
i
ons, we mo
d
e
l the relatio
n
s to gr
ap
h. Secon
d
, we re-mo
del the grap
h to cont
rol-
flow p
a
ttern.
Third,
we
co
mpare formal
ization
of
con
t
rol-flo
w
p
a
tterns in
discov
ered
g
r
aph
a
nd
formali
z
ation
of control
-
flo
w
pattern
s in
the busin
ess process mo
d
e
l to confirm
the right activ
i
ty
relation
s in
graph. Confirm
a
tion is
used
to clea
r up th
e noi
se relati
on which ha
s same f
r
eq
ue
ncy
with non
-noi
se relation. Fin
a
lly, we obtai
n parall
e
l rela
tion of discovered p
r
o
c
e
s
s model is
sam
e
as the pa
ralle
l relation of b
u
sin
e
ss process model.
Evaluation Warning : The document was created with Spire.PDF for Python.