Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
15
,
No.
1
,
Febr
uary
20
25
, pp.
669
~
676
IS
S
N:
20
88
-
8708
, DO
I: 10
.11
591/ij
ece.v
15
i
1
.
pp
669
-
676
669
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om
Hybrid
long sho
rt
-
term m
emor
y
and de
cisi
on tree
model fo
r
optimizi
ng patie
nt volum
e pre
dictions in e
mergen
cy
departm
ents
Ah
med
Aba
t
al
1
, Mou
r
ad
M
zi
li
2
, Za
ka
ri
a B
enlali
a
2
, Ha
j
ar
K
ha
ll
ouki
3
,
Toufi
k
Mzil
i
2
,
Moham
med E
l Kaim Bi
ll
ah
4
, La
it
h
Ab
u
aligah
5,6,7,8
1
Facu
lty
of Scienc
e L
eg
al
Econ
o
m
ic
an
d
Social, Ch
o
u
aib
Dou
k
k
ali Univ
ersity
,
EL
Jad
id
a,
M
o
rocco
2
Facu
lty
of Scienc
e,
Ch
o
u
aib
Dou
k
k
ali Univ
ersity
,
El
J
ad
id
a,
Moro
cco
3
Facu
lty
of Scienc
e and
T
echn
o
lo
g
y
,
Un
iv
ersite Ha
ss
an
Prem
i
er,
Settat
M
o
rocco
4
Dep
artm
en
t of
Co
m
p
u
ter
Sci
en
ce E
S
TSB,
EL
I
TE
S L
ab
,
Ch
o
u
aib
Dou
k
k
al
i Univ
ersity
,
El
Ja
d
id
a,
Moro
cco
5
Co
m
p
u
ter
Scien
ce
Depart
m
en
t,
Al al
-
Bay
t Univ
ersity
,
Maf
raq
,
Jo
rdan
6
Cen
tre
for Rese
ar
ch
I
m
p
act
an
d
Ou
t
co
m
e,
Ch
itk
ara
Un
iv
ersity
I
n
stitu
te of
E
n
g
in
eering
and
T
echn
o
lo
g
y
,
Ch
itk
ara
Un
iv
ersity
,
Raj
p
u
ra,
Pu
n
jab
,
Ind
ia
7
Ap
p
lied
scien
ce res
earc
h
cente
r,
Ap
p
lied
scien
ce priva
te un
iv
ersity
,
Am
m
an
,
Jo
rdan
8
Artif
icial
I
n
telli
g
e
n
ce a
n
d
Sens
in
g
T
echn
o
lo
g
ies (AI
ST
)
Res
ear
ch
Cen
ter,
Un
iv
ersity
of T
ab
u
k
,
Tabu
k
,
Sau
d
i Ar
ab
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
M
a
y 10,
2024
Re
vised
A
ug 13, 2
024
Accepte
d
Aug 20,
2024
In
th
is
study
,
we
addr
ess
critical
oper
at
ion
al
i
nef
ficien
c
ie
s
in
em
erg
enc
y
depa
rt
me
nts
(
E
Ds
)
by
deve
loping
a
hybrid
pr
e
dic
ti
v
e
mod
el
th
at
in
te
gra
te
s
long
short
-
te
r
m
me
mory
(
LSTM)
net
works
with
dec
ision
trees
(
DT)
.
Th
is
mode
l
significan
tl
y
enha
nc
es
th
e
pre
di
ct
ion
of
p
a
ti
ent
volu
me
s,
a
key
fa
ct
or
in
red
uc
ing
wa
it
times,
optimi
zi
ng
resour
ce
a
ll
oc
at
ion
,
and
i
mprovi
ng
over
all
serv
ic
e
q
ual
it
y
in
hospit
a
ls.
By
accuratel
y
fore
ca
sting
th
e
n
umbe
r
of
inc
omi
ng
p
atien
ts,
our
model
f
ac
i
li
t
at
es
th
e
ef
fic
i
ent
distr
ibut
i
on
of
both
huma
n
and
m
ateri
a
l
resour
ce
s,
ta
il
or
ed
spe
ci
fi
c
al
ly
to
ant
i
ci
pa
t
ed
de
ma
nd
.
Furtherm
ore
,
th
i
s
pre
dictive
acc
ura
cy
ensure
s
t
hat
EDs
ca
n
mainta
in
high
servic
e
stand
ard
s
eve
n
dur
ing
p
e
ak
t
im
es,
ultim
ately
le
ad
ing
to
be
tt
er
patient
outc
om
es
and
more
eff
e
ct
iv
e
use
of
h
ealth
ca
re
fa
ci
l
it
i
es.
Thi
s
pap
er
dem
onstra
te
s
ho
w
adva
nc
ed
da
ta
ana
ly
ti
cs
ca
n
b
e
le
ver
age
d
to
sol
ve
some
of
the
most
pr
essing
challe
ng
es
fa
c
ed
by
emerge
n
c
y
me
d
ic
a
l
serv
ices t
oday
.
Ke
yw
or
d
s
:
Bi
g
data
Data anal
ys
is
Decisi
on tree
models
Emer
gen
c
y de
par
tme
nts
Healt
hcar
e
syst
ems
Lo
ng sho
rt
-
te
r
m mem
ory
model
M
ac
hin
e
l
ea
rn
i
ng
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
BY
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
Ahmed
Abata
l
Faculty
of Scie
nce Le
gal Ec
onomi
c an
d Soc
ia
l
,
Chouaib
Douk
kali U
niv
e
r
sit
y
El Jadida
, M
orocco
Emai
l:
a.ab
at
al
@uh
p.
ac.
ma
1.
INTROD
U
CTION
Emer
gen
c
y
d
e
par
tme
nts
(E
D
s)
a
re
crit
ic
al
com
pone
nts
of
healt
hcar
e
s
yst
ems
[
1],
[
2]
,
ta
sk
e
d
with
the
chall
en
ging
du
t
y
of
pro
vid
in
g
imme
diate
care
unde
r
unpre
dicta
ble
a
nd
oft
en
c
ha
otic
co
nd
it
io
ns
.
E
ff
ic
ie
nt
mana
geme
nt of
t
hese units
is cru
ci
al
,
as
the
v
aria
bili
ty
in
pa
ti
ent
volu
mes can
sig
nificantl
y
im
pact
wait
t
imes,
resou
rce
al
loc
at
ion
,
a
nd
ov
erall
qu
al
it
y
of
healt
hca
re
[
3]
se
rv
ic
es
.
H
igh
var
ia
bili
ty
le
ads
to
pe
riod
s
of
ov
e
rcro
wd
i
ng,
increase
d
wait
ti
mes,
and
ca
n
ulti
mate
ly
co
mpro
mise
patie
nt
care
qual
it
y
w
he
n
res
our
ces
are
stret
ched to
o
t
hin
[
4]
.
Trad
it
io
nally,
EDs
ha
ve
us
e
d
basic
sta
ti
sti
cal
methods
and
li
near
f
oreca
sti
ng
mode
ls,
su
c
h
a
s
auto
regressive
integr
at
e
d
m
oving
a
ver
a
ge
(AR
IMA)
[
5]
m
odel
s
an
d
e
xpon
entia
l
smoothi
ng,
to p
re
dict
pa
ti
ent
vo
l
um
es
.
T
hes
e
meth
ods
of
te
n
fail
to
ca
ptur
e
com
plex
,
nonlinear
patte
r
ns
an
d
sea
sonal
va
riat
ion
s
i
n
patie
nt
arr
ivals
,
le
adi
ng
to
sub
op
ti
mal
operati
on
al
decisi
ons.
Fo
r
e
xam
ple,
ARI
M
A
m
ode
ls,
w
hile
cap
a
ble
of
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
669
-
676
670
handlin
g
ti
me
series
,
are
l
imi
te
d
by
t
he
ir
ass
umpti
on
of
li
near
relat
ion
s
hip
s
a
nd
t
heir
ina
bili
ty
to
accomm
odat
e
t
he
non
-
li
near
dynamics
of
E
D
patie
nt
fl
ows
[
6],
[
7]
.
Simi
la
rly,
e
xpone
ntial
smoothin
g,
wh
il
e
us
ef
ul f
or shor
t
-
te
rm
f
or
eca
sti
ng, does
not
ta
ke
s
uffici
ent
a
c
count o
f
lo
ng
-
t
erm
t
rends
an
d
seas
on
al
va
riat
ion
s,
le
ading
to
inac
cur
at
e
f
or
ecast
s
du
rin
g
pea
k
per
i
od
s
.
In
a
ddit
ion
,
t
rad
it
io
na
l
for
ecast
in
g
models
ge
ner
al
ly
fail
to
inc
orp
or
at
e
di
ver
se
ty
pes
of
data,
s
uc
h
as
patie
nt
de
mogra
phic
s,
m
edical
histo
r
y
and
oth
er
c
on
t
extual
inf
or
mati
on
,
w
hich
a
re
esse
nt
ia
l
fo
r
ma
king
inf
or
me
d
res
ource
al
locat
io
n
decisi
ons
[8]
.
Dem
ogra
ph
ic
detai
ls
of
the
patie
nt
and
me
dical
history
ca
n
pro
vi
de
valua
ble
in
formati
on
on
pote
ntial
patie
nt
nee
ds
an
d
res
ourc
e
requireme
nts,
bu
t
the
se
el
ements
are
of
t
en
overl
ooke
d
by
sim
pler
models.
Fail
ure
to
inte
gr
at
e
this
heter
og
e
ne
ou
s
data
le
ads
to
a
la
ck
of
forecast
accu
rac
y,
f
urt
her
c
ompli
cat
ing
the
mana
geme
nt
of
E
D
resou
rces
a
nd
operati
ons.
I
n
a
dd
it
io
n,
t
ra
diti
on
al
met
ho
ds
a
re
gen
e
ral
ly
reacti
ve
rat
her
t
ha
n
pr
oa
ct
ive,
offer
i
ng
li
tt
le
fo
resi
gh
t
into
pote
ntial
wa
ves
of
patie
nt
ar
riv
al
s.
T
his
reacti
ve
a
ppr
oach
ha
mp
e
rs
t
he
a
bili
ty
of
emer
gen
c
y
de
pa
rtments
to
plan
a
nd
al
locat
e
resou
rces
e
ffec
ti
vely,
oft
en
re
su
lt
ing
in
ov
e
r
crow
ding,
inc
r
eased
wait
ing
ti
mes,
and
a
po
or
qua
li
ty
of
ca
re.
R
ecent
stu
dies
hi
gh
li
ght
the
ne
ed
f
or
ad
va
nce
d
pre
dicti
ve
m
od
el
s
capab
le
of
bette
r
ha
ndli
ng
the
c
omplexit
y
an
d
var
ia
bili
t
y
of
e
mer
ge
nc
y
dep
a
rtme
nt
data,
s
uggestin
g
tha
t
machine
lear
nin
g t
ech
niques
cou
l
d offer
m
ore r
obust
so
l
ution
s
.
The
a
dvent
of
big
data
a
nalyt
ic
s
[9]
a
nd
a
dv
anced
mac
hin
e
le
arn
i
ng
[
10]
,
[11]
te
ch
nique
s
prese
nt
a
new
opport
un
i
ty
t
o
overc
ome
these
c
halle
ng
e
s.
I
n
pa
rtic
ular,
the
i
nteg
rati
on
of
lo
ng
s
hort
-
te
r
m
m
emo
ry
(LST
M)
[12
]
netw
orks
a
nd
decisi
on
t
rees
(
DT
)
i
n
a
hybri
d
m
odel
offe
rs
a
pro
misi
ng
s
olu
ti
on.
LST
M
netw
orks
are
r
enow
ned
for
t
heir
ef
ficacy
i
n
anal
yzin
g
ti
me
-
se
ries
data,
captur
i
ng
the
te
mp
oral
de
pe
nd
e
ncies
essenti
al
f
or
pr
edict
ing
patte
r
ns
i
n
patie
nt
ar
rivals.
O
n
t
he
oth
e
r
hand,
D
T
are
ade
pt
at
processi
ng
str
uc
ture
d
data,
pro
vid
in
g
crit
ic
al
insi
gh
ts
i
nto
patie
nt
pro
file
s
a
nd
histo
rical
medical
data
,
wh
ic
h
ar
e
cr
uc
ia
l
fo
r
resou
rce
plan
nin
g
[
13]
.
The
main
c
on
t
rib
ution
of
thi
s
pa
pe
r
is
the
de
velo
pme
nt
of
a
no
vel
hybr
i
d
pr
e
dicti
ve
m
od
el
that
sign
ific
a
ntly
e
nh
a
nces
t
he
predict
io
n
of
pa
ti
ent
vo
l
um
es
.
By
s
yner
gizing
t
he
te
m
por
al
data
proces
sing
capab
il
it
ie
s
of
LST
M
[
14]
with
t
he
str
uctu
red
data
analysis
st
rength
of
D
Ts
,
our
m
odel
of
fer
s
a
com
pr
e
he
ns
ive
too
l
f
or
m
or
e
accurate
a
nd
dyna
mic
f
or
ecas
ti
ng
.
T
his
a
ppr
oach
facil
it
at
es
eff
ic
ie
nt
al
loc
at
ion
of
re
source
s,
both
huma
n
an
d
mate
rial
,
ba
se
d
on
a
nt
ic
ipate
d
patie
nt
volu
mes,
th
us
ai
mi
ng
to
reduce
pa
ti
ent
wait
ing
ti
mes
and
im
prov
e
s
erv
ic
e
qual
it
y
in
hos
pital
s.
T
he
im
pleme
ntati
on
of
this
m
od
el
co
uld
le
ad
to
trans
formati
ve
impro
veme
nts
in
ED
operati
on
s
,
op
ti
mizi
ng
res
ource
us
a
ge
an
d
e
nhanc
ing
patie
nt
out
comes
by ali
gn
i
ng op
erati
on
al
st
rategies
with actua
l dema
nd p
at
te
rn
s
[
15]
.
The
r
emai
ni
ng
par
t
of
t
his
pa
per
is
orga
nize
d
as
fo
ll
ows:
s
ect
ion
2
prese
nt
s
the
main
c
ontrib
utio
n
of
so
me
of
the
r
el
at
ed
work
s
.
Sect
ion
3
pro
vid
es
the
met
hodolo
gy,
w
hile
sect
ion
4
de
scribes
the
pro
posed
hybri
d
lo
ng
s
hort
-
te
rm
mem
ory
-
decisi
on
tre
es
(
LST
M
-
DT
)
model,
al
on
g
with
it
s
imple
mentat
io
n
t
o
predict
patie
nt
volume
in
E
Ds.
Sect
i
on
5
pre
se
nts
resu
lt
s
an
d
s
hows
t
he
superi
or
acc
ur
ac
y
of
the
hybr
i
d
m
od
el
in
patie
nt
volu
me
pr
e
dicti
on,
whil
e
hav
in
g
the
le
ast
RMSE
bet
ween
al
l
models.
Sect
io
n
6
ex
po
s
es
li
mit
at
ion
s
of
the
pro
pose
d
model.
Her
e
w
e
co
nclu
de
ou
r
wor
k
in
t
his
s
ect
ion
a
nd
sug
gest
some
f
ut
ure
exte
ns
io
ns
of
t
his
top
ic
in
secti
on
7
.
2.
RELATE
D
W
ORKS
In
E
Ds,
le
veragin
g
big
da
ta
to
predict
pa
ti
ent
volume
s
has
at
tract
e
d
e
xtensi
ve
resea
rch
ef
forts
.
Var
i
ou
s
m
ode
ls
ha
ve
bee
n
exp
l
or
e
d,
f
r
om
tra
diti
on
al
sta
ti
sti
cal
meth
ods
to
a
dv
a
nc
ed
mac
hin
e
le
ar
ni
ng
te
chn
iq
ues
.
T
hi
s
sect
ion
e
xa
mines
the
r
obust
ness
a
nd
li
mi
ta
ti
on
s
of
thes
e
existi
ng
methods
a
nd
em
phasi
zes
the n
ee
d f
or
a
novel a
ppr
oach, the
LST
M
de
ci
sion
tree
m
odel
, to o
ve
rcome i
de
ntifie
d s
hortco
min
gs
.
Time
series
models
s
uc
h
as
ex
pone
ntial
sm
oo
t
hing
(
ETS)
[
16]
an
d
ARI
MA
[5]
hav
e
been
commo
nly
util
iz
ed
to
pr
e
dict
patie
nt
vo
l
um
e
s
in
E
Ds.
W
hile
ef
fecti
ve
i
n
c
aptu
rin
g
li
nea
r
tre
nd
s
a
nd
patte
rn
s,
their
li
mit
at
ions
bec
om
e
ap
pa
ren
t
w
hen
ha
ndli
ng
no
n
-
li
nea
r
patte
rn
s
an
d
seaso
nal
va
riat
ion
s
,
of
te
n
fail
ing
t
o
integrate
c
r
ucial
struct
ur
e
d
da
ta
li
ke
dia
gnosi
s
co
des
a
nd
pa
ti
ent
dem
o
-
gr
aph
ic
s
[
17]
–
[
19]
.
Gafni
-
Pa
pp
as
an
d
Kh
a
n
[
20]
f
urt
her
de
monstr
at
ed
that
mac
hin
e
le
ar
ning
models,
pa
rtic
ularly
ra
ndom
f
or
est
s
an
d
gradie
nt
boos
te
d
mac
hin
es,
ou
t
perfor
med
thes
e tradi
ti
on
al
ti
me series meth
od
s i
n
te
rms of
acc
ur
a
cy
in
pr
e
dicti
ng d
ai
ly
ED
visit
s
[
21]
.
The
ad
opti
on
of
machi
ne
le
arn
in
g
[22
]
al
gorith
ms,
in
cl
ud
in
g
ra
ndom
f
or
est
s
(RF
s)
a
nd
su
pp
or
t
vecto
r
machine
s
(SV
M
)
,
has
sho
wn
promise
in
ha
nd
li
ng
str
uctu
r
ed
data
e
ff
ect
i
vely.
H
ow
e
ve
r
,
these
models
oft
en
strug
gle
with
ti
me
-
se
ries
data d
ue
to
t
heir
ina
bili
ty
to
ca
ptur
e
long
-
te
rm
de
pende
ncies, r
e
qu
i
rin
g
extensi
ve
fe
at
ure e
ng
i
neer
i
ng
and care
fu
l
re
gula
rizat
ion t
o
a
vo
i
d ov
e
rf
it
ti
ng
[23],
[24]
.
Trad
it
io
nal
ne
ur
al
netw
orks
hav
e
been
ap
pl
ie
d
with
s
ome
su
cce
ss
i
n
predict
in
g
patie
nt
vo
l
um
es,
prof
ic
ie
nt
in
mana
ging
non
-
li
near
data
pa
tt
ern
s.
N
onet
heless,
the
y
a
r
e
pla
gu
e
d
by
issues
li
ke
va
nish
i
ng
gr
a
dients,
wh
i
ch
diminis
h
th
ei
r
ca
pab
il
it
y
t
o
mainta
in
acc
ur
ac
y
over
e
xt
end
e
d
pe
rio
ds
[25]
.
To
a
ddre
ss
the
li
mit
at
ion
s
in
he
ren
t
i
n
th
ese
netw
orks,
LST
M
netw
orks
ha
ve
bee
n
util
ized
,
sho
wing
i
mpro
ve
d
handl
ing
of
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
Hy
br
id
lo
ng s
ho
rt
-
te
rm
m
e
m
or
y
and decisi
on tree
model f
or
…
(
A
hm
e
d Ab
ata
l
)
671
long
-
te
r
m
de
pe
nd
e
ncies
esse
ntial
f
or
ti
me
-
series
data,
al
though
their
c
ompu
ta
ti
onal
dema
nds
li
mit
their
app
li
cabil
it
y
i
n real
-
ti
me
pr
e
di
ct
ion
sce
nar
i
os
[
26],
[
27]
.
Chen
g
an
d
K
uo
[28]
resear
ch
i
nvest
igate
s
inte
rn
et
of
thin
gs
(
IoT)
a
nd
big
data
i
nteg
rati
on
in
healt
hcar
e
to
e
nh
a
nce
proacti
ve
ca
re
models
an
d
a
ddres
s
da
ta
fr
a
gm
e
ntati
on
a
nd
ine
ff
ic
ie
ncies.
It
highli
gh
ts
the
us
e
of
L
S
TM
m
od
el
s
f
or
pr
e
dicti
ng
E
D
wait
ti
mes,
sh
owcasi
ng
im
pro
ved
acc
ur
a
cy
c
ompa
red
t
o
li
nea
r
regressio
n
(
LR)
models.
H
oweve
r,
the
stu
dy
overlo
oks
po
te
ntial
adv
a
nceme
nts
be
yo
nd
IoT
an
d
bi
g
data
,
la
cks
th
oroug
h
exp
l
or
at
io
n
of
data
secu
rity
and
s
cal
abili
ty
con
ce
r
ns
,
a
nd
may
no
t
fu
ll
y
consi
der
sta
ke
ho
l
der
per
s
pecti
ves
on
te
ch
nolo
gy
a
ccepta
nce
a
nd
impact
in
healt
hcar
e
set
ti
ng
s
emer
ging
re
sea
rch,
s
uch
a
s
th
e
w
ork
by
S
har
a
fat
a
nd
Ba
yati
[
29]
,
i
ntr
oduces
a
dv
a
nce
d
dee
p
le
arn
in
g
f
rame
wor
ks
li
ke
Pat
ie
ntFlowNet,
wh
ic
h
employ
c
onvoluti
on
al
neural
netw
orks
to
e
nhance
patie
nt
f
low
pre
dicti
ons
in
E
Ds,
dem
on
st
rati
ng
s
up
erior
accurac
y
a
nd
offe
rin
g
val
uab
l
e
op
e
rati
onal
insig
hts
[
29]
.
S
imi
la
rly,
stu
die
s
co
nducted
i
n
un
i
qu
e
set
ti
ngs
li
ke
the
Bra
zi
li
an
opht
halmol
ogic
al
hos
pital
highli
gh
t
the
fea
sibil
it
y
of
imple
mentin
g
mac
hin
e
le
a
rn
i
ng
m
odel
s
b
y
cl
inici
ans
with
ou
t c
odin
g
e
xperience t
o fore
cast
ED
visit
s
and tra
um
a ca
s
es accu
ratel
y
[
30]
.
In
par
al
le
l
,
J
ose
ph
et
al.
[
31]
exp
l
or
es
dee
p
le
arn
in
g’s
pot
entia
l
to
pr
e
di
ct
emerg
e
nc
y
dep
a
rtme
nt
work
l
oa
d
at
a
patie
nt
le
vel,
s
howing
that
ne
ural
netw
orks
,
e
sp
eci
al
ly
th
os
e
anal
yzi
ng
unstruct
ur
e
d
data,
su
bst
antia
ll
y
r
edu
ce
er
r
or
rat
es
in
est
imat
in
g
wor
k
relat
iv
e
valu
e
un
it
s
(
wRV
Us),
unde
rscorin
g
t
he
pract
ic
al
app
li
cat
io
ns
of these tec
hnologies i
n real
-
ti
m
e ED
sett
ing
s
.
Coll
ect
ively,
t
hese
st
ud
ie
s
un
der
sc
ore
the
ne
cessi
ty
f
or
c
on
ti
nu
ed
in
novation
in
pr
e
dicti
ve
modeli
ng
for
E
Ds.
Our
r
esearch
c
ontrib
utes to
this
body of
work
by intr
oducin
g
a
hybri
d
L
ST
M
-
DT
model,
desig
ned
t
o
sy
nt
hesize
the
te
m
poral
proc
essing
ca
pa
bili
ti
es
of
L
ST
M
with
t
he
anal
yt
ic
al
pr
eci
sio
n
of
D
Ts
.
T
his
nov
el
appr
oach
ai
ms
to
ad
dress
t
he
com
plex
an
d
dynamic
na
ture
of
patie
nt
flo
w
in
emer
gen
c
y
de
par
t
ments,
pro
vid
in
g
a
ro
bu
st
and
pr
eci
s
e meth
od to
pr
edict
p
at
ie
nt
volume
.
3.
METHO
D
In
this
s
ect
ion,
we
desc
ribe
i
n
detai
l
the
va
r
iou
s
sta
ges
of
our
meth
odology,
fr
om
data
colle
ct
ion
t
o
model
e
valuati
on.
Data
colle
ct
ion
was
ca
rri
ed
ou
t
in
c
ollaborat
io
n
with
a
Mo
ro
cca
n
ho
s
pital
,
f
ocu
s
ing
on
op
e
rati
onal
paramet
ers
an
d
pa
ti
ent
flow
i
n
th
e
emer
gen
c
y
de
par
tme
nt.
I
n
l
ine
with
ri
gor
ous
et
hical
sta
ndar
ds
and
l
ocal
re
gu
l
at
ion
s
on
me
di
cal
data
confid
entia
li
ty,
we
e
ns
ure
d
that
the
data
colle
ct
ion
proce
ss
gua
r
anteed
patie
nt
pri
vac
y.
T
he
data
c
ollec
te
d
inclu
de
s
an
onymo
us
inf
ormat
ion
on
patie
nts
,
ar
rival
ti
mes,
t
ypes
of
treat
ment a
nd
ou
tc
om
es
, pr
ovidin
g
a
s
olid
basis
for our a
nalys
is.
Data pr
e
-
proce
ssing
was
a
n
e
ssentia
l st
ep
in
en
surin
g
the i
nteg
rity and qu
al
it
y
of
the
data u
sed
i
n
our
analysis.
We
t
ook
ste
ps
to
c
le
an
the
data
of
missi
ng
val
ues,
no
rmali
ze
the
data
to
a
scal
e
ap
pro
pr
i
at
e
f
or
analysis,
an
d
e
ncode
cat
e
gori
cal
var
ia
ble
s
t
o
ma
ke
t
hem
com
patible
with
mac
hin
e
le
ar
ning
al
go
rithm
s.
T
o
impleme
nt
the
model,
we
opte
d
f
or
a
hybri
d
appr
oach
c
ombinin
g
lo
ng
s
hort
-
te
rm
me
mory
netw
orks
(L
STM)
and
decisi
on
tr
ees.
T
his
co
mbi
nation
e
na
bles
us
to
ex
plo
it
both
te
m
poral
a
nd
str
uctur
e
d
da
ta
from
e
mer
gen
c
y
dep
a
rtme
nts, w
hich
is
cr
ucial
for
acc
ur
at
e
pr
edict
ion
of
patie
nt volu
mes a
nd e
ff
ic
ie
nt
res
ource
all
ocati
on.
On
ce
t
he
mod
el
was
in
plac
e,
we
trai
ne
d
it
on
a
segm
e
nted
par
t
of
the
dataset
,
us
i
ng
rig
oro
us
cro
ss
-
validat
io
n
t
o
opti
mize
par
a
mete
rs
an
d
a
void
ove
r
-
fitt
ing
.
We
the
n
eval
uated
the
model’s
pe
rfo
r
mance
on
a
se
pa
rate
dataset
reserve
d
f
or
this
pur
pose,
us
in
g
mea
su
res
su
c
h
as
r
oo
t
mea
n
s
qu
a
re
erro
r
(R
M
S
E)
an
d
area
un
der
the
ROC
cu
rv
e
(AUROC)
t
o
ass
ess
it
s
accurac
y
an
d
reli
abili
t
y
in
pr
e
dic
ti
ng
patie
nt
volu
m
es
an
d
resou
rce
re
quir
ements.
Finall
y,
al
l
proce
dur
es
in
t
his
stu
dy
wer
e
co
nduc
te
d
in
acco
rd
a
nce
with
t
he
hi
gh
est
et
hical
sta
ndar
ds
,
with
the
a
ppr
oval
of
the
r
el
evan
t
i
ns
ti
tuti
on
al
re
view
boar
ds
an
d
et
hic
s
co
mmit
te
es.
Pati
ent
data
we
re
an
onym
iz
ed
a
nd
de
-
ide
ntifie
d
pri
or
to
anal
ys
is
to
ensure
c
onfid
entia
li
ty
and
c
ompli
ance
wit
h
loca
l
and inter
natio
na
l data p
r
otect
ion re
gula
ti
on
s.
4.
PROP
OSE
D HYBR
ID
LST
M
-
DECISIO
N
TREE
M
O
DEL S
CHE
M
A
The
pro
posed
h
ybrid
LS
T
M
-
DT
m
od
el
in
Fi
gure
1
is
co
nceive
d
t
o
w
ork
t
hro
ugh
a
str
uctu
re
d
schema
i
nvolvi
ng
se
ver
al
c
r
it
ic
al
sta
ges
and
c
omp
on
e
nt
s
de
vised
in
a
man
ner
t
o
help
in
opti
m
iz
ing
pre
d
ic
ti
on
s
f
or
patie
nt
vo
l
ume
s
in
E
D.
T
he
model
sta
rts
a
t
data
colle
ct
io
n
,
i
n
wh
ic
h
c
riti
cal
informati
on
is
dr
a
w
n
from
va
rio
us
s
ource
s,
inclu
ding
detai
ls
of
the
patie
nt
,
el
ect
r
on
ic
he
al
th
rec
ords
,
a
nd
the
histor
ic
al
dat
a
of
the
patie
nt
arr
ivals
an
d
t
r
eat
ments
at
th
e
ED
.
T
his
is
then
f
ollow
e
d
by
data
prep
r
oc
essing
:
t
he
c
ol
le
ct
ed
data is sta
nda
r
dized, t
ake
n
ca
re
of
f
or
missi
ng
values
, a
nd c
le
aned
,
s
o
that
it
’s
fit f
or m
odel
ing
.
T
h
e
n
e
x
t
s
t
e
p
w
i
l
l
b
e
t
o
d
i
v
i
de
t
h
e
d
a
t
a
i
n
t
o
a
t
r
a
i
n
i
n
g
d
a
t
a
s
e
t
a
n
d
a
t
e
s
t
i
n
g
d
a
t
a
s
e
t
.
T
h
e
t
r
a
i
n
i
n
g
d
a
t
a
s
e
t
i
s
u
s
e
d
t
o
h
e
l
p
i
n
t
h
e
b
u
i
l
d
i
n
g
a
n
d
t
r
a
i
n
i
n
g
o
f
t
h
e
m
o
d
e
l
w
h
i
l
e
t
h
e
t
e
s
t
i
n
g
d
a
t
a
s
e
t
w
i
l
l
b
e
u
s
e
f
u
l
t
o
e
v
a
l
u
a
t
e
t
h
e
m
o
d
e
l
s
.
T
h
e
f
i
r
s
t
p
h
a
s
e
o
f
L
S
T
M
M
o
d
e
l
m
a
k
e
s
u
s
e
o
f
t
h
e
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
n
e
t
w
o
r
k
s
t
o
c
a
p
t
u
r
e
t
h
e
t
i
m
e
d
e
p
e
n
d
e
n
c
i
e
s
a
n
d
p
a
t
t
e
r
n
s
o
f
t
h
e
s
e
r
i
e
s
t
i
m
e
da
t
a
r
e
l
a
t
e
d
t
o
t
he
p
a
t
i
e
n
t
a
r
r
i
v
a
l
s
a
n
d
E
D
o
p
e
r
a
t
i
o
n
s
.
This w
il
l be
f
ol
lowed by f
e
at
ur
e e
xtracti
on t
hat seek
s to
e
xt
ract al
l require
d
feat
ur
es i
n
th
e p
re
dicti
on
of
the
patie
nt
volume
f
r
om
t
he
outp
uts
giv
e
n
by
the
LS
T
M
model.
Fo
ll
ow
e
d
by
t
his
will
be
t
he
dec
isi
on
t
ree
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
669
-
676
672
M
odel
,
assessi
ng
str
uc
tu
red
data
s
uch
as
pa
ti
ent
diag
nosis
co
des
a
nd
de
mogra
phic
s,
w
hich
a
re
very
key
to
making
acc
ur
a
te
predict
io
ns
of
the
volume
.
T
he
trai
ning
phase
of
the
m
odel
t
hen
trai
ns
bo
t
h
m
od
el
s
to
a
dap
t
to the i
den
ti
fie
d patt
ern
s
and
data feat
ur
es
.
Af
te
r
the
trai
nin
g
is
c
omplet
e
,
the
m
odel
s’
pr
e
dicti
on
a
cc
ur
ac
y
is
te
ste
d
thr
ough
the
m
od
el
-
te
sti
ng
ph
a
se,
us
in
g
the
te
sti
ng
dat
aset
.
T
he
m
od
el
evaluati
on
will
be
ca
rr
ie
d
ou
t
with
the
help
of
some
of
the
fo
ll
owin
g
met
rics:
Roo
t
me
an
s
qu
a
re
er
r
or
(R
M
SE
)
an
d
area
un
der
t
he
rec
ei
ve
r
ope
rati
ng
c
ha
racteri
sti
c
(AUROC
)
c
urve.
The
se
a
re
ta
ken
to
te
st
the
pr
e
dicti
on
accurac
y
an
d
cl
assifi
cat
ion
s
tren
gth
a
bili
ty
of
t
he
hybri
d
m
odel
.
If
the
model
pro
ves
r
obust
in
performa
nce,
the
d
e
pl
oyment
ph
a
se
fo
ll
ows,
w
he
re
the
dev
el
op
e
d
model
gets
i
m
plemented
at
a
re
al
ED
set
ti
ng
t
o
ai
d,
am
ong
ot
her
s,
i
n
patie
nt
s’
volume
f
oreca
sti
ng
towa
rd
s
imp
roved
res
ource
a
nd
ser
vice
deliver
y.
C
onti
nuous
Im
prov
e
me
nt,
if
re
qu
ire
d,
sh
oul
d
be
car
ried
out
for
ma
king t
he mo
del m
or
e
ef
fecti
ve ov
e
r
ti
me.
Figure
1.
P
r
opos
e
d
hybri
d
LS
TM
-
decisi
on
tr
ee
model
sc
he
ma
5.
RESU
LT
S
AND DI
SCUS
S
ION
This
stu
dy
sho
wed
t
hat
the
hybri
d
L
ST
M
-
de
ci
sion
tree
m
od
el
dem
onstr
at
ed
high
perf
ormance
i
n
pr
e
dicti
ng
pat
ie
nt
vo
l
um
e
i
n
eme
rg
e
nc
y
dep
a
rtme
nts.
The
res
ults
of
the
pe
rfo
rm
anc
e
eval
uatio
n
are
su
m
marized
in
Table
1
a
nd
il
lustrate
d
in
Fig
ur
e
2.
T
o
highl
igh
t
the
s
up
e
ri
or
it
y
of
our
a
ppr
oac
h,
we
c
ompa
red
our
hybri
d
m
odel
to
se
ve
ral
oth
e
r
c
om
m
only
us
e
d
pr
e
di
ct
ion
te
ch
niqu
es.
Ta
ble
1
show
s
co
mp
a
rin
g
the
performa
nce
of
the
dif
fer
e
nt
models
in
te
r
ms
of
R
M
SE
,
AU
R
OC,
t
rainin
g
ti
me,
model
co
mp
le
xity,
a
nd
abili
ty to
ca
pture
non
-
li
nea
riti
es an
d
te
m
pora
l dep
e
ndencies
.
Table
1.
C
omp
ariso
n
of
pr
e
dicti
on
models
Mod
el
RMSE
AUROC
Tr
ain
in
g
tim
e
Mod
el c
o
m
p
lex
ity
Cap
tu
res no
n
-
lin
earities
Cap
tu
res tem
p
o
ral
d
ep
en
d
en
cies
ARIMA
8
.6
0
.72
5
m
in
u
tes
Low
No
Yes
SVM
7
.5
0
.76
1
0
m
in
u
tes
Mediu
m
Partially
No
RF
6
.9
0
.79
1
5
m
in
u
tes
Hig
h
Yes
No
LST
M
6
0
.82
2
0
m
in
u
tes
Very
h
ig
h
Yes
Yes
Hy
b
rid m
o
d
el
5
.4
0
.85
25
m
in
u
tes
Very
h
ig
h
Yes
Yes
5.1.
Disc
ussio
n
The
ARI
M
A
model
pe
rform
s
well
on
c
on
ve
ntion
al
ti
me
s
eries
data,
with
an
R
M
SE
of
8.6
an
d
a
n
AU
R
OC
of
0.72,
but
s
hows
it
s
li
mit
at
ion
s
wh
e
n
face
d
wi
th
c
omplex
non
-
li
nea
r
t
rends
.
Its
trai
ning
ti
me
is
relat
ively
short
at
5
min
utes,
and
it
s
c
omple
xity
is
lo
w.
S
V
M
,
with
a
n
R
M
SE
of
7.5
a
nd
a
n
A
UR
OC
of
0.76,
is
eff
ect
ive
f
or
li
near
or
wea
kl
y
no
n
-
li
nea
r
da
ta
.
H
ow
e
ve
r,
it
per
f
orms
le
s
s
well
f
or
co
m
plex
ti
me
se
rie
s
an
d
ta
kes
ar
ound
10
min
utes
to
trai
n
with
me
diu
m
c
omple
xi
ty.
The
RF
model
has
a
n
RMSE
of
6.9
and
a
n
AU
R
OC
of
0.7
9.
Alth
ough
it
handles
hi
gh
-
di
mensional
dat
a
well
a
nd
a
vo
ids
ov
e
r
-
fitt
ing,
it
do
es
n’t
rea
ch
t
he
accurac
y of o
ur
hybr
i
d
m
odel
and take
s
15
minu
te
s
to
t
rain wit
h hig
h
c
omplexit
y.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
Hy
br
id
lo
ng s
ho
rt
-
te
rm
m
e
m
or
y
and decisi
on tree
model f
or
…
(
A
hm
e
d Ab
ata
l
)
673
The
L
ST
M
m
od
el
is
excell
e
nt
f
or
capt
ur
i
ng
lo
ng
-
te
r
m
de
pende
ncies
in
ti
me
series
da
ta
,
with
a
n
RMSE
of
6.0
a
nd
an
AU
R
OC
of
0.82.
Its
tra
ining
ti
me
is
20
mi
nute
s
wit
h
ve
ry
hi
gh
c
omplexit
y.
Fi
nally
,
ou
r
hybri
d
m
odel
(
LST
M
+
DT
)
outpe
rforms
al
l
oth
e
r
m
od
el
s
with
an
RM
SE
of
5.4
a
nd
an
AU
R
OC
of
0.85.
It
com
bin
es
t
he
s
tren
gth
s
of
LS
TM
for
ca
ptu
ri
ng
te
m
poral
de
pende
ncies
a
nd
decisi
on
tre
es
for
non
-
li
ne
arit
ie
s
and
c
omplex
i
nteracti
ons,
de
monstrati
ng
superi
or
it
y
in
te
r
ms
of
acc
uracy
an
d
ov
e
rall
pe
rformance
.
Althou
gh
it
s
trai
nin
g
ti
m
e
is
25
min
utes
and
it
s
co
mp
le
xity
ve
ry
high,
the
performa
nc
e
ben
e
fits
justi
fy
these
c
os
ts.
This
model
validat
e
s
our
a
ppro
ac
h
to
patie
nt
volu
me
predict
io
n,
ou
t
performi
ng
oth
e
r
m
od
el
s
e
valuated
i
n
te
r
ms
of
RMSE
a
nd
A
UROC.
To
il
lu
strat
e
this
supe
rior
it
y,
we
pre
sent
Fig
ure
2
c
ompari
ng
t
he
performa
nce
i
n
te
rm
s
of
R
M
S
E
a
nd
AU
R
OC
f
or
e
ach
m
odel
.
Th
ese
gr
a
phs
cl
e
arly
s
how
th
at
our
hybr
i
d
m
od
el
ou
t
perf
orms
th
e
oth
e
r
m
odel
s i
n
te
r
ms
of both
RMS
E
a
nd AUROC,
v
al
idat
ing
our
appr
oa
ch
to
p
at
ie
nt volume
pre
dicti
on.
Figure
2.
RM
S
E
an
d
A
UROC
c
ompa
rison
be
tween
models
5.1.1.
Adv
an
t
ag
e
s
an
d
bene
fits fo
r
emer
ge
ncy ser
vices
Our
hy
br
id
m
od
el
has
sig
ni
ficantl
y
i
mpr
ov
e
d
res
ource
ma
na
geme
nt
f
or
healt
hca
r
e
pr
ov
i
der
s
,
par
ti
cula
rly
in
the
area
s
of
op
ti
mize
d
be
d
al
locat
io
n,
str
at
egic
sta
f
f
pl
ann
i
ng,
an
d
re
du
ce
d
pa
ti
ent
wait
ing
ti
mes.
Ac
c
urat
e
patie
nt
volu
me
f
oreca
sti
ng
is
esse
ntial
for
the
ef
fici
ent
op
e
rati
on
of
e
mer
gen
c
y
de
pa
rtments,
le
ading
to
bette
r
patie
nt
outc
om
es
a
nd
a
m
or
e
strea
mli
ne
d
w
orkf
l
ow.
T
he
res
ults
of
our
stu
dy
highli
gh
t
th
e
increasin
g
im
portance
of
util
iz
ing
data
a
nd
a
dva
nce
d
analyti
cal
met
hodolo
gies
t
o
ad
dress
the
com
plex
chall
enges
facing
the
he
al
thc
are
sect
or
.
B
y
co
mb
i
ning
L
STM
netw
orks
with
DT
al
gorithms,
the
pro
po
s
ed
model
le
ve
rages
the
stre
ngth
s
of
bot
h
te
ch
ni
qu
es
t
o
e
nh
a
nc
e
the
ef
fici
en
cy
a
nd
res
ponsi
ven
ess
of
eme
rg
e
nc
y
serv
ic
es
.
T
his
appr
oach
no
t
only
en
ha
nces
pa
ti
ent
care
but
al
so
set
s
a
pre
ceden
t
f
or
t
he
app
li
cat
io
n
of
hybri
d
models i
n healt
hcar
e
an
al
ytics t
o dr
i
ve
in
nova
ti
on
a
nd im
prov
e
servic
e
del
iver
y.
6.
CONCL
US
I
O
N
In
c
oncl
us
i
on,
our
stu
dy
em
pl
oy
s
a
hybri
d
model
that
me
rg
es
LST
M
a
nd
de
ci
sion
tree
s
to
pre
dict
patie
nt
volum
e
in
eme
r
gency
de
par
tme
nt
s.
T
his
hy
br
i
d
ap
proac
h
le
ve
rag
es
L
ST
M
’
s
a
bili
ty
to
c
aptu
re
te
mp
oral
dep
e
nd
e
ncies
a
nd
de
ci
sion
tree
s’
a
bili
ty
to
handl
e
non
-
li
near
in
te
racti
on
s,
res
ul
ti
ng
in
a
m
odel
that
ou
t
performs
c
onve
ntion
al
ti
me
-
se
ries
a
nd
mac
hin
e
le
ar
ning
al
gorithm
s
in
te
r
ms
of
accurac
y
an
d
patie
nt
vo
l
um
e
cl
assif
ic
at
ion
.
By
op
ti
mizi
ng
t
he
t
r
ai
nin
g
process
,
we
ens
ure
c
on
sist
e
nt
mod
el
pe
rformanc
e
an
d
eff
ic
ie
nt
re
sou
rce
util
iz
at
ion
.
The
pr
a
ct
ic
al
impleme
ntati
on
of
our
mod
el
has
the
po
t
entia
l
to
sign
i
f
ic
antly
enh
a
nce
the
e
f
fici
ency
of
e
m
erg
e
nc
y
op
e
rat
ion
s
,
le
adi
ng
t
o
im
pro
ved
pa
ti
ent
care,
re
duced
healt
hca
re
costs,
and
m
or
e
e
ff
ic
i
ent
res
ource
al
l
ocati
on. O
ur f
ind
i
ngs
are con
sist
ent
with p
re
vi
ous r
esearc
h
that
emp
hasize
s
the
sign
ific
a
nce
of
a
dvance
d
da
ta
anal
ytics
in
a
ddressi
ng
the
c
omplex
c
halle
ng
es
f
aced
by
eme
r
gen
c
y
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
669
-
676
674
dep
a
rtme
nts.
Our
model
st
ands
ou
t
f
or
it
s
op
ti
mize
d
pr
e
dicti
ve
ca
pa
bili
ti
es,
ena
bl
ing
bette
r
re
so
urce
mana
geme
nt
decisi
ons.
Fu
t
ur
e
resea
rch
will
con
ce
ntra
te
on
inte
gr
at
ing
a
ddit
ion
al
machi
ne
le
a
rn
i
ng
al
gorithms,
s
uc
h
as
ra
ndom
f
or
est
s
an
d
SVMs,
to
f
ur
t
her
enh
a
nce
the
pr
edict
ive
acc
ur
a
cy
a
nd
r
obust
ne
ss
of
patie
nt
volum
e
f
or
eca
sts
i
n
eme
rg
e
nc
y
de
par
tme
nts.
T
his
ongoin
g
de
velo
pm
e
nt
s
e
eks
to
imp
r
ove
th
e
app
li
cabil
it
y
a
nd ef
fecti
ve
nes
s of the
m
od
el
in r
eal
healt
hca
re s
et
ti
ngs.
7.
LIMITATI
O
NS
OF P
ROP
OSED M
ODE
L
A
hybri
d
LST
M
-
DT
m
od
el
s
hows
pro
mise
bu
t
faces
chall
eng
e
s,
nota
bly
in
da
ta
qual
it
y,
over
fitt
ing
,
and
inte
rpretab
il
it
y.
LST
M
s
de
pen
d
on
am
pl
e
data
a
nd
have
hi
gh
c
omp
utati
on
al
dema
nds,
li
mit
ing
rea
l
-
ti
me
pr
e
dicti
on.
De
ci
sion
T
rees’
a
ssu
m
ptio
n
of
f
eat
ur
e
in
de
penden
ce
oft
en
underper
f
or
m
s
in
real
-
w
orl
d
dataset
s
.
Hype
rp
a
ramet
er
tun
i
ng
is
te
dio
us,
c
omp
ounded
by
c
on
ce
pt
dri
ft
a
nd
var
ia
ti
ons
in
EDs.
Ac
hieving
exp
la
ina
bili
ty
for
t
he
co
mb
i
ned
model
is
di
ff
ic
ult.
C
onti
nuous
m
on
i
toring
an
d
a
da
ptati
on
a
re
vi
ta
l
for
real
-
w
orl
d
e
ff
e
ct
iveness.
REFERE
NCE
S
[1]
A.
Ab
at
al,
M.
M
zil
i,
T
.
Mzili,
K
.
Ch
e
rr
at,
A
.
Y
ass
in
e,
a
n
d
L
.
Ab
u
alig
ah
,
“
Intellig
en
t interco
n
n
ected h
ealth
care
s
y
stem:
in
teg
rating
IoT
an
d
b
ig
d
ata
f
o
r
p
erso
n
alized
p
a
tien
t
car
e,”
Inter
n
a
tio
n
a
l
Jo
u
rnal
o
f
On
lin
e
a
n
d
Biom
ed
ica
l
Eng
in
eerin
g
,
v
o
l.
2
0
,
n
o
.
1
1
,
p
p
.
4
6
–
6
5
,
2
0
2
4
,
d
o
i: 10
.39
9
1
/ijo
e.v20i1
1
.4
9
8
9
3
.
[2]
A.
Ab
atal,
H.
Khal
lo
u
k
i,
an
d
M
.
Bah
aj,
“
A
sm
a
rt
in
te
rc
o
n
n
ected h
ealth
care
sy
stem
us
in
g
clo
u
d
compu
tin
g
,”
in
ACM
Inter
n
a
tio
n
a
l
Co
n
feren
ce P
ro
ceedin
g
Ser
ies
,
2
0
1
8
,
do
i: 10
.11
4
5
/3
2
3
0
9
0
5
.3230
9
3
6
.
[3]
T.
Mzil
i
et
a
l.
,
“E
n
h
an
cin
g
COVID
-
1
9
v
accinatio
n
a
n
d
m
ed
icatio
n
d
istrib
u
tio
n
ro
u
tin
g
stra
teg
ies
in
rural
regio
n
s
o
f
Moro
cco:
a
co
m
p
arative
m
etah
eu
ristics an
aly
sis
,”
Info
rma
tics in
Me
d
icin
e Unlo
cked
,
v
o
l.
4
6
,
2
0
2
4
,
d
o
i: 1
0
.10
1
6
/j.im
u
.20
2
4
.10
1
4
6
7
.
[4]
R.
Pasto
rino
et
a
l.
,
“Ben
efits
an
d
c
h
allen
g
es
o
f
Big
Data
in
h
ealth
car
e
:
an
o
v
ervie
w
o
f
th
e
Euro
p
ean
in
iti
ativ
es,”
Eur
o
p
ea
n
Jo
u
rn
a
l of Pub
lic
Hea
lth
,
v
o
l.
2
9
,
p
p
.
2
3
–
2
7
,
2
0
1
9
,
d
o
i:
10
.10
9
3
/eu
rp
u
b
/c
k
z1
6
8
.
[5]
H.
Bo
u
sq
ao
u
i,
I.
Slima
n
i,
an
d
S.
A
ch
ch
ab
,
“Co
m
p
ara
tiv
e
an
aly
sis
o
f
short
-
ter
m
d
em
an
d
p
redictin
g
m
o
d
els
u
sin
g
ARIMA
an
d
d
eep
learnin
g
,”
I
n
tern
a
tio
n
a
l
Jo
u
r
n
a
l
o
f
Electrica
l
a
n
d
Co
mp
u
ter
Eng
in
eerin
g
,
v
o
l.
1
1
,
n
o
.
4
,
p
p
.
3
3
1
9
–
3
3
2
8
,
2
0
2
1
,
d
o
i: 10
.1159
1
/ijec
e.v1
1
i4
.pp
3
3
1
9
-
3
3
2
8
.
[6]
B.
Gay
e,
D.
Zhan
g
,
an
d
A.
W
u
lam
u
,
“I
m
p
rov
em
en
t
o
f
su
p
p
o
rt
v
ecto
r
m
achi
n
e
alg
o
rithm
in
b
ig
d
ata
b
ackg
ro
u
n
d
,”
Ma
th
ema
tica
l
Pro
b
lems in
E
n
g
in
eerin
g
,
2
0
2
1
,
d
o
i:
1
0
.11
5
5
/2
0
2
1
/5
5
9
4
8
9
9
.
[7]
H.
He,
S
.
Gao
,
T
.
Jin
,
S.
Sato
,
an
d
X.
Zhan
g
,
“A
seas
o
n
al
-
trend
d
ecomp
o
sitio
n
-
b
ased
d
e
n
d
ritic
n
eu
ron
m
o
d
el
for
finan
ci
al
tim
e
series p
redictio
n
,”
App
lied
So
ft Co
mp
u
tin
g
,
v
o
l.
1
0
8
,
2
0
2
1
,
d
o
i: 10
.1016
/j.
aso
c.20
2
1
.1074
8
8
.
[8]
H.
Ran
so
m
an
d
J.
M.
Olss
o
n
,
“Allo
c
atio
n
o
f
h
ealth
car
e
reso
u
rces:
p
r
in
cip
les
for
d
ecisio
n
-
m
ak
in
g
,”
Pedia
tri
cs
In
Re
view
,
v
o
l.
3
8
,
n
o
.
7
,
p
p
.
3
2
0
–
3
2
9
,
J
u
l.
2
0
1
7
,
d
o
i: 10.
1
5
4
2
/p
ir.
2
0
1
6
-
0
0
1
2
.
[9]
S.
R
.
Salk
u
ti,
“
A
su
rvey
o
f
b
ig
d
a
ta
an
d
m
achi
n
e
le
arnin
g
,”
Inter
n
a
tio
n
a
l
Jo
u
rn
a
l
o
f
El
ectrica
l
a
n
d
Co
m
p
u
ter
Eng
in
eerin
g
,
v
o
l.
1
0
,
n
o
.
1
,
p
p
.
5
7
5
–
5
8
0
,
2
0
2
0
,
d
o
i: 10
.11
5
9
1
/ijece.v1
0
i1
.pp57
5
-
5
8
0
.
[10
]
M.
Sa
rker,
“Re
v
o
lu
tio
n
izin
g
h
ealth
care
:
th
e
role
o
f
m
achi
n
e
learnin
g
in
th
e
h
ealth
secto
r,
”
Jo
u
rn
a
l
o
f
Artif
icia
l
Intellig
en
ce
Gen
era
l scien
ce (
J
AI
GS
)
,
v
o
l.
2
,
n
o
.
1
,
p
p
.
3
5
–
4
8
,
2
0
2
4
,
d
o
i: 10
.60
0
8
7
/jaig
s.v
2
i1
.p47.
[11
]
N.
Jiw
an
i,
K.
Gu
p
ta,
an
d
P
.
W
h
ig
,
“
Machin
e
lea
rnin
g
ap
p
roach
es
for
an
aly
sis
in
sma
rt
h
ealt
h
care
in
for
m
ati
cs,”
Ma
ch
in
e
Lea
rn
in
g
a
n
d
A
rtificia
l I
n
tellig
en
ce in Health
ca
re S
ystems
:
To
o
ls a
n
d
Techn
iq
u
es
,
p
p
.
1
2
9
–
1
5
4
,
2
0
2
2
,
d
o
i: 10
.1201
/9
7
8
1
0
0
3
2
6
5
4
3
6
-
6.
[12
]
M.
Ban
sal,
A.
G
o
y
al,
an
d
A.
Ch
o
u
d
h
ary,
“A
co
m
p
arative
an
aly
sis
o
f
k
-
n
eare
st
n
eig
h
b
o
r,
g
en
etic,
su
p
p
o
rt
v
ecto
r
m
achi
n
e
,
d
ecisio
n
tree,
an
d
lo
n
g
sh
o
rt
ter
m
m
e
m
o
ry
alg
o
ri
th
m
s
in
m
achi
n
e
lea
rnin
g
,”
De
cis
io
n
Ana
lytics
Jo
u
rn
a
l
,
v
o
l.
3
,
2
0
2
2
,
d
o
i: 10
.1016
/j.dajour.20
2
2
.1000
7
1
.
[13
]
N.
V.
Ch
awl
a
an
d
D.
A.
Dav
is,
“Brin
g
in
g
b
i
g
d
ata
to
p
e
rso
n
alized
h
ealth
care
:
a
p
atien
t
-
c
en
te
red
fr
a
m
ewo
rk,”
J
o
u
rn
a
l
o
f
Gen
era
l
Inter
n
a
l Med
icin
e
,
v
o
l.
2
8
,
n
o
.
S3
,
p
p
.
6
6
0
–
6
6
5
,
Sep
.
2
0
1
3
,
d
o
i: 10
.1007
/s
1
1
6
0
6
-
013
-
2
4
5
5
-
8.
[14
]
A.
Bo
u
k
h
alfa,
A.
Ab
d
ellao
u
i,
N.
H
m
in
a,
an
d
H.
Ch
a
o
u
i,
“LST
M
d
eep
learnin
g
m
eth
o
d
f
o
r
n
etwo
rk
in
trus
i
o
n
d
etectio
n
sy
stem,
”
Jo
u
rn
a
l of Eng
in
ee
rin
g
an
d
A
p
p
lied
S
cien
ces
,
v
o
l.
1
5
,
n
o
.
1
,
p
p
.
2
2
7
–
2
3
2
,
2
0
1
9
,
d
o
i: 1
0
.36
4
7
8
/jeasci.2
0
2
0
.22
7
.
2
3
2
.
[15
]
G.
Ma
ragath
am
an
d
S.
Dev
i,
“LT
S
M
m
o
d
el
for
p
re
d
icti
o
n
o
f
h
eart
failure
in
b
ig
d
ata,”
Jo
u
rnal
o
f
Med
ica
l
S
ystems
,
v
o
l.
4
3
,
n
o
.
5
,
2
0
1
9
,
d
o
i: 1
0
.10
0
7
/s1
0
9
1
6
-
019
-
1
2
4
3
-
3.
[16
]
H.
Liu
et
a
l.
,
“Fo
r
ecast
o
f
th
e
trend
in
in
cid
en
ce
o
f
acut
e
h
em
o
rr
h
ag
ic
co
n
ju
n
ctiv
itis
in
C
h
in
a
fr
o
m
2
0
1
1
–
2
0
1
9
u
sin
g
t
h
e
seas
o
n
al
au
to
regressiv
e
in
te
g
rated
m
o
v
in
g
av
erage
(SARI
MA
)
a
n
d
ex
p
o
n
en
tial
smoo
th
in
g
(E
T
S)
m
o
d
els,”
Jo
u
rn
a
l
o
f
I
n
fectio
n
a
n
d
Pub
lic
Hea
lth
,
v
o
l.
1
3
,
n
o
.
2
,
p
p
.
2
8
7
–
2
9
4
,
2
0
2
0
,
d
o
i: 10
.1
0
1
6
/j
.jiph
.20
1
9
.1
2
.00
8
.
[17
]
A.
G.
Sal
m
an
an
d
B.
Kan
ig
o
ro,
“
Visib
ility
forecast
in
g
u
sin
g
au
to
reg
ressiv
e
in
teg
rated
m
o
v
in
g
av
erage
(ARI
MA)
m
o
d
els,”
Pro
cedi
a
Co
mp
u
te
r S
cien
ce
,
v
o
l.
1
7
9
,
p
p
.
2
5
2
–
2
5
9
,
2
0
2
1
,
d
o
i: 10
.10
1
6
/j.p
rocs
.20
2
1
.01
.0
0
4
.
[18
]
B. Wag
n
er,
“Class
ifyin
g
em
ergen
cy
d
ep
artm
en
t data
to
im
p
rov
e sy
n
d
romi
c su
rveillan
ce:
fr
o
m
m
ix
ed
d
ata
typ
es to
I
CD
Co
d
es an
d
sy
n
d
romes,”
MS
.
Thes
is, Faculty
of
Econ
o
m
ics, Univ
e
rsität Biele
feld, 20
2
3
.
[19
]
S.
Kim
,
P.
Y
.
Le
e,
M.
Le
e,
J.
Ki
m
,
a
n
d
W
.
Na,
“Im
p
ro
v
ed
state
-
of
-
h
ealth
p
redictio
n
b
ased
o
n
au
to
-
regressiv
e
in
teg
rated
m
o
v
in
g
av
erage
with
ex
o
g
en
o
u
s
v
ariables
m
o
d
el
in
o
v
ercoming
b
attery
d
eg
radatio
n
-
d
ep
en
d
en
t
in
ter
n
al
p
aram
et
er
v
ari
atio
n
,”
Jo
u
rn
a
l
o
f
Ener
g
y Sto
ra
g
e
,
v
o
l.
4
6
,
2
0
2
2
,
d
o
i: 1
0
.10
1
6
/j.est.20
2
1
.1
0
3
8
8
8
.
[20
]
G.
Gafni
-
Pap
p
as
an
d
M.
Kh
an
,
“Pre
d
ictin
g
d
aily
em
ergen
cy
d
ep
artm
en
t
v
isits
u
sing
m
achi
n
e
learnin
g
co
u
ld
i
n
crea
se
accura
cy
,”
American
Jou
rn
a
l of Emerg
en
cy Med
icin
e
,
v
o
l.
6
5
,
p
p
.
5
–
1
1
,
2
0
2
3
,
d
o
i: 1
0
.10
1
6
/j.ajem
.20
2
2
.12
.01
9
.
[21
]
M.
A
.
C.
Vo
llm
er
et
a
l.
,
“A
u
n
ifie
d
m
achi
n
e
lea
rnin
g
ap
p
roach
to
tim
e
series
forecastin
g
ap
p
lied
to
d
em
an
d
at
em
e
rgen
cy
d
ep
artm
en
ts,”
BM
C Emerg
en
cy Med
icin
e
,
v
o
l.
2
1
,
n
o
.
1
,
p
p
.
1
–
1
4
,
Dec.
2
0
2
1
,
d
o
i: 10
.1
1
8
6
/
s1
2
8
7
3
-
020
-
0
0
3
9
5
-
y.
[22
]
A.
Z
.
A
.
Zainu
d
d
in
,
W
.
Mans
o
r,
K
.
Y.
L
ee,
an
d
Z.
Ma
h
m
o
o
d
in
,
“Machin
e
learnin
g
an
d
d
ee
p
learnin
g
p
erf
o
r
m
an
ce
in
class
ifyin
g
d
y
slex
ic
ch
ild
ren’s
electroen
ceph
alo
g
ram
d
u
ring
writ
in
g
,”
Inter
n
a
tio
n
a
l
Jo
u
rn
a
l
o
f
Electrica
l
a
n
d
Co
mp
u
ter
Engineer
in
g
,
v
o
l.
1
2
,
n
o
.
6
,
p
p
.
6
6
1
4
–
6
6
2
4
,
2
0
2
2
,
d
o
i: 1
0
.1
1
5
9
1
/ijece.v1
2
i
6
.p
p
6
6
1
4
-
6
6
2
4
.
[23
]
S.
Tuli,
G.
Cas
ale
,
an
d
N.
R.
Jen
n
in
g
s,
“Tr
an
AD:
d
ee
p
trans
form
e
r
n
et
wo
rks
for
an
o
m
al
y
d
etectio
n
in
m
u
ltiv
ariate
ti
m
e
seri
es
d
ata,”
a
rXiv pr
ep
ri
n
t ar
Xiv:
2
2
0
1
.07
2
8
4
,
2
0
2
2
,
d
o
i: 1
0
.1
4
7
7
8
/
3
5
1
4
0
6
1
.35
1
4
0
6
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
Hy
br
id
lo
ng s
ho
rt
-
te
rm
m
e
m
or
y
and decisi
on tree
model f
or
…
(
A
hm
e
d Ab
ata
l
)
675
[24
]
D.
Zhu
an
g
,
V.
J.
L.
Gan
,
Z.
Du
y
g
u
Tekler
,
A
.
Ch
o
n
g
,
S.
Tian,
an
d
X.
Sh
i,
“Dat
a
-
d
riven
p
redictiv
e
co
n
trol
for
s
m
art
HVAC
sy
stem
in
IoT
-
in
teg
rated
b
u
ild
in
g
s
with
tim
e
-
series
f
o
reca
stin
g
an
d
rei
n
forcem
en
t
l
earni
n
g
,”
App
lied
Ener
g
y
,
v
o
l.
3
3
8
,
2
0
2
3
,
d
o
i: 10
.1016
/j.ape
n
ergy
.20
2
3
.1
2
0
9
3
6
.
[25
]
S
.
Bas
h
eer
,
S
.
Bh
a
tia,
an
d
S
.
B
.
Sak
r
i,
“Co
m
p
u
tatio
n
al
m
o
d
elin
g
o
f
d
em
e
n
tia
p
redictio
n
u
sin
g
d
eep
n
eu
ral
n
et
wo
rk:
an
aly
sis
o
n
OASIS
datas
et,”
I
EE
E
A
cc
ess
,
v
o
l.
9
,
p
p
.
4
2
4
4
9
–
4
2
4
6
2
,
2
0
2
1
,
d
o
i: 10
.110
9
/ACC
ESS.
2
0
2
1
.30
6
6
2
1
3
.
[26
]
B.
B.
Sah
o
o
,
R.
Jh
a,
A.
Sin
g
h
,
a
n
d
D.
Ku
m
a
r,
“Lo
n
g
sh
o
rt
-
term
m
e
m
o
ry
(
LST
M)
re
c
u
rr
en
t
n
eu
ral
n
etwo
rk
for
lo
w
-
flo
w
h
y
d
rological tim
e
series for
ecastin
g
,”
Acta Geop
h
ysica
,
v
o
l.
6
7
,
n
o
.
5
,
p
p
.
1
4
7
1
–
1
4
8
1
,
2
0
1
9
,
d
o
i: 10
.1007
/s11600
-
019
-
0
0
3
3
0
-
1.
[27
]
H.
Lin,
C
.
Sh
i,
B
.
W
an
g
,
M.
F.
Ch
an
,
X.
Tan
g
,
an
d
W
.
Ji,
“Towa
rds
real
-
t
im
e
resp
iratory
m
o
tio
n
p
redictio
n
b
a
sed
o
n
lo
n
g
sh
o
rt
-
term m
em
o
ry n
eu
r
al netwo
rks
,”
Phys
ics in
Med
icin
e an
d
B
io
lo
g
y
,
v
o
l.
6
4
,
n
o
.
8
,
2
0
1
9
,
d
o
i: 1
0
.10
8
8
/
1
3
6
1
-
6
5
6
0
/
ab
1
3
fa.
[28
]
N.
Ch
en
g
an
d
A.
Ku
o
,
“Usin
g
lo
n
g
sh
o
rt
-
term
m
em
o
r
y
(
LT
SM
)
n
eu
ral
n
etwo
rks
to
p
redic
t
em
ergen
cy
d
ep
ar
tm
en
t
w
ait
ti
m
e,
”
S
tu
d
ies in
hea
lth
te
ch
n
o
lo
g
y a
n
d
info
r
ma
tics
,
v
o
l.
2
7
2
,
p
p
.
1
9
9
–
2
0
2
,
2
0
2
0
,
d
o
i: 10
.3233
/SHTI
2
0
0
5
2
8
.
[29
]
A.
R.
Sh
arafat
an
d
M.
B
ay
ati,
“P
atie
n
tflown
et:
a
d
eep
l
earnin
g
ap
p
roach
to
p
atien
t
fl
o
w
p
re
d
ictio
n
in
em
ergen
cy
d
ep
artm
en
ts.,
”
IE
E
E
A
ccess
,
v
o
l.
9
,
p
p
.
4
5
5
5
2
–
4
5
5
6
1
,
2
0
2
1
,
d
o
i: 10
.1
1
0
9
/ACC
ESS.
2
0
2
1
.30
6
6
1
6
4
.
[30
]
L.
F.
Nak
ay
a
m
a,
L.
Z.
Rib
ei
ro,
an
d
C.
V.
S.
Reg
atier
i,
“An
em
e
rgen
cy
roo
m
in
flux
an
d
tr
au
m
a
cases
p
redic
tio
n
in
a
Brazilian
o
p
h
th
al
m
o
lo
g
ical
h
o
sp
ital
b
y
an
o
p
h
th
alm
o
lo
g
ist
with
o
u
t
co
d
e
ex
p
erience
,”
Arq
u
ivo
s
Bra
sil
eiro
s
d
e
Ofta
lmo
lo
g
ia
,
v
o
l.
8
7
,
n
o
.
3
,
2
0
2
4
,
d
o
i: 1
0
.59
3
5
/0
0
0
4
-
2
7
4
9
.2
0
2
2
-
0130.
[31
]
J.
W
.
Jo
sep
h
et
a
l.
,
“Machin
e
lea
rnin
g
m
eth
o
d
s
fo
r
p
redictin
g
p
atien
t
-
lev
el
em
e
rgen
cy
d
ep
a
rtm
en
t
wo
r
k
lo
ad
,”
Jo
u
rnal
o
f
Emerg
en
cy Medici
n
e
,
v
o
l.
6
4
,
n
o
.
1
,
p
p
.
8
3
–
9
2
,
2
0
2
3
,
d
o
i: 10
.10
1
6
/j.jem
er
m
ed
.20
2
2
.10
.00
2
.
BIOGR
AP
HI
ES OF
A
UTH
ORS
Ah
me
d
Ab
atal
ob
ta
in
ed
a
b
a
che
lor’s
degr
ee
in
co
mput
e
r
sc
ience
and
industri
a
l
engi
ne
eri
ng
fro
m
the
Fa
cul
ty
of
Scie
nc
es
and
T
e
chni
ques
of
Set
t
at
(FS
TS)
in
201
1,
a
m
aste
r’s
degr
ee
in
d
istri
b
ute
d
info
rmatio
n
sys
te
ms
from
the
Moham
me
d
i
a
Higher
Norma
l
School
of
Te
chn
ic
a
l
Edu
cation
in
2013,
an
d
a
do
ct
ora
te
in
com
put
er
sc
ie
n
ce
fro
m
th
e
sam
e
fa
cul
ty
in
2021.
His
rese
a
rch
intere
sts
include
m
ac
hin
e
a
nd
dee
p
le
arn
in
g,
he
al
th
ca
r
e
ar
chi
t
ec
tur
e
in
per
vasive
com
p
uti
ng,
and
SQ
L
t
o
SP
ARQ
L
conve
rsion.
He
has
publi
shed
num
er
ous
scie
nti
f
ic
art
i
cl
es
in inde
x
ed
journ
al
s.
He c
an
be
con
ta
c
te
d
at
em
a
il
:
Abat
al.
ahm
ed@gm
ai
l
.
c
om
.
Mourad
M
zi
l
i
is
a
r
ese
ar
c
her
in
the
Dep
a
rtm
ent
of
Ma
th
em
a
ti
cs
at
Choua
i
b
Doukkali
Univ
e
rsity,
spec
i
alizi
n
g
in
opt
im
i
zatio
n
and
ma
th
emat
ic
s.
With
a
robu
st
a
ca
d
em
i
c
bac
kground
and
expe
rt
ise
in
his
f
ie
ld
.
He
con
tri
b
ute
s
significantl
y
to
adva
n
ci
ng
knowledge
in
opti
mization
.
He
ca
n
be
con
ta
c
ted a
t
e
ma
i
l:
mour
adm
z
il
i2023@g
ma
il.c
o
m
.
Zak
ari
a
B
enlalia
ob
ta
in
ed
a
b
ac
he
lor’s
d
e
gre
e
in
m
at
he
m
at
i
ca
l
sci
ences
an
d
com
put
er
sc
ie
nc
e
fro
m
th
e
Faculty
of
Scie
n
ce
s
o
f
El
Jadid
a
in
2
010,
a
spe
ci
a
liz
ed
m
aste
r’s
degr
ee
in
engi
n
ee
ring
,
co
mput
e
r
sci
ence
and
i
nte
rne
t
from
th
e
F
ac
ul
ty
of
S
ci
en
ce
s
Ai
n
Choc
of
C
asa
b
la
nc
a
in
2012,
and
a
doct
or
ate
in
com
put
er
scie
nc
e
from
th
e
Faculty
of
Scie
nc
es
of
El
J
adi
da
in
2021.
His
rese
arc
h
intere
sts
inc
lud
e
o
pti
mization,
cl
o
ud
com
puti
ng
.
He
has
publ
ished
nume
rous
sc
ie
nt
ifi
c
ar
ti
c
le
s
in
ind
exe
d
journa
ls
.
H
e
c
an
be
contac
t
ed
at
email:
benl
a
li
a
.
z
aka
ri
a
@gma
il.c
o
m
.
Hajar
Khall
ou
k
i
born
in
19
89,
she
com
pl
eted
her
ma
st
er’
s
degr
ee
in
com
pu
t
e
r
scie
nc
e
fro
m
Fa
c
ult
y
of
Sci
ences,
Hass
an
II
Univ
ersit
y,
Casab
la
n
ca
,
Moroc
co.
In
pursuit
of
a
Ph.D,
she
joi
n
ed
the
Depa
r
tm
en
t
of
Mathe
m
at
i
cs
and
Comput
er
S
ci
en
ce
s,
Fa
cul
ty
of
Scie
nc
es
and
T
ec
hno
logy
,
Hass
an
I
Univ
e
rsity
Set
tat,
Mor
occ
o,
in
2013
.
I
n
2019,
she
jo
in
ed
La
keh
ea
d
Univer
sity
Can
a
da
as
a
Pos
tdoctora
l
Fe
ll
ow.
H
er
actua
l
m
ai
n
rese
arc
h
intere
st
s
conc
ern
mul
timedia
do
cu
me
nts
ad
aptati
o
n,
con
te
x
t
awa
r
e
ness,
smart
ci
t
ies
,
and
sem
ant
i
c
web
.
She
ca
n
be
con
tacte
d
a
t
e
ma
il:
Hkha
ll
ou
@la
keheadu.ca
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
669
-
676
676
Tou
fik
M
zi
l
i
is
a
disti
ngu
ished
profe
ss
or
a
nd
rese
ar
che
r
at
Chouai
b
Doukkal
i
Univer
sity,
whe
re
he
spe
ci
a
lize
s
in
op
ti
m
izati
o
n
and
ar
ti
fi
cial
intelligen
ce.
H
is
extensive
publi
c
at
ion
rec
o
rd
inc
lud
es
nu
me
rous
ar
ti
c
le
s
in
pre
st
igi
ous
Q1
and
Q2
journ
al
s,
ma
rk
ing
signifi
c
ant
con
tributions
to
th
ese
f
ie
lds
.
He
is
a
lso
highl
y
reg
a
rde
d
as
a
se
asone
d
re
vie
wer
and
edi
tor
for
respe
c
te
d
a
ca
d
em
i
c
jo
urna
ls,
and
he
h
as
edi
t
ed
multip
le
ind
exe
d
book
s
withi
n
his
doma
in
.
He
c
an be
con
tacte
d
a
t
e
ma
il:
mz
i
li
.
t@uc
d.
ac.ma
.
Mohamme
d
El
Kai
m
B
il
lah
is
a
rese
arc
h
er
i
n
th
e
Dep
artme
n
t
of
Inform
at
i
cs
a
t
Chouai
b
Doukka
li
Univer
sity
,
spec
i
al
i
zi
ng
in
mo
del
ing
,
applied
ma
th
em
a
ti
cs,
d
ata
sc
ie
n
ce
and
stat
isti
c
s.
With
a
strong
a
ca
de
m
ic
b
ac
kground
a
nd
experti
se
in
t
hese
fi
el
ds
.
He
signifi
c
ant
ly
cont
ributes
to
ad
vanc
ing
knowle
dge
in
applied
sc
ie
nc
es
and
r
elated
areas.
He
ca
n
be
con
ta
c
te
d
at
em
a
il
:
e
lka
i
m
_bil
la
h
.
moh
am
m
ed@uc
d.
ac.m
a
.
Laith
Ab
ualigah
is
the
Dire
c
tor
of
the
Dep
artme
nt
of
In
te
rn
at
i
onal
R
el
a
ti
ons
an
d
Affai
rs
a
t
Al
Al
-
Bayt
Univer
sity
,
Jordan.
He
is
an
associ
at
e
prof
essor
at
th
e
Comp
ute
r
Scie
n
ce
Depa
rtment,
Al
Al
-
Bayt
Unive
rsity,
Jordan.
He
is
a
lso
a
d
isti
nguished
res
ea
r
che
r
at
ma
ny
pre
stigi
ous
univ
ersit
i
es.
He
c
an be
con
tacte
d
a
t
e
ma
il:
al
ig
ah.
202
0@gma
il.c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.