Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
6, N
o
. 4
,
A
ugu
st
2016
, pp
. 19
07
~
1
919
I
S
SN
: 208
8-8
7
0
8
,
D
O
I
:
10.115
91
/ij
ece.v6
i
4.9
438
1
907
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
An Optimizing Approach for Multi Constraints
Reassign
ment
Probl
e
m
of
Hum
a
n Res
o
urces
Said Tk
atek
1
, Otm
a
n Ab
dou
n
2
, J
a
a
f
a
r
Ab
ouch
ab
ak
a
1
, Na
ja
t R
a
f
a
lia
1
1
LaRI
T,
F
acul
t
y
of S
c
i
e
nc
es
, Ibn
T
ofai
l Univers
i
t
y
K
e
nitr
a,
M
o
rocco
2
Pluridisciplin
ar
y
Labor
ator
y
,
Po
ly
d
i
sciplinar
y
F
a
cu
lt
y,
Abde
lm
al
k Es
s
aadi
Unive
r
s
i
t
y
,L
ara
c
he,
M
o
rocco
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Nov 16, 2015
Rev
i
sed
Ap
r 4, 20
16
Accepted
Apr 27, 2016
This
pap
e
r pr
es
ents
an
eff
e
c
tiv
e appro
ach
to
o
p
tim
ize
the
re
as
s
i
gnm
ent
of
Hum
a
n Resources in th
e
enter
p
rise th
at is fo
rm
ed b
y
s
e
vera
l units
o
f
productions to
take in
to considerati
on the hu
man characteris
tics. Th
is
approach
cons
is
ts
of two s
t
eps
;
the f
i
rs
t s
t
ep
i
s
to form
aliz
e
t
h
e s
t
udied
problem
tha
t
is
prac
tic
all
y
t
a
k
e
th
e form
of
the g
e
nera
li
zed
as
s
i
gnm
ent
problem
(GAP) known as NP-h
ard problem
. A
dditionall
y
,
th
e
variab
les in
the formulation of our problem are inte
rlinked b
y
certain constr
aints. Th
ese
two proprie
ties
can
to just
if
y th
e im
portan
t
com
p
lexit
y
of
this p
r
oblem
. Th
e
second step
is f
o
cused to solv
e
this complex pr
oblem b
y
using
the gen
e
tic
algorithm
.
W
e
p
r
es
ent th
e exp
e
ri
m
e
ntall
y
r
e
sult f
o
r justif
ying
the
valid
it
y o
f
the proposed ap
proach. So
, the s
o
lution obt
ain
e
d
allowed us to g
e
t an optimal
assignment of personnel th
at can lead to
improve the
averag
e pr
oductivity
or
ratab
ili
t
y
or
at
l
e
ast ensure
i
t
s eq
uilibra
tion
withi
n
sites of
en
terpr
i
se.
Keyword:
Gen
e
tic al
g
o
rith
m
Mu
lti co
n
s
t
r
aints
Produ
ctiv
ity
R
eassi
gnm
ent
pr
o
b
l
e
m
Copyright ©
201
6 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
Otm
a
n Ab
do
u
n
,
Depa
rtem
ent of Com
puter Sci
e
nce,
Pol
y
disc
iplin
ar
y Facult
y,
Abdel
m
alek Essaadi
Uni
v
ersity,
B.P 745
, Main
Po
st
9
2004
- Lar
ach
e -
M
o
ro
cco
.
Em
a
il: ab
d
o
u
n
@fp
l
.m
a
1.
INTRODUCTION
Th
e
hu
m
a
n
reso
urces can b
e
seen
as a source of
su
stain
e
d co
m
p
etitiv
e ad
v
a
n
t
ag
e for
org
a
n
i
zatio
n
s
an
d
a m
a
j
o
r asset in
th
eir strateg
i
c reorientatio
n
.
Fu
rt
he
r
m
ore, for a m
o
re dy
nam
i
c
m
a
nagem
e
nt
of t
h
ese
reso
u
r
ces,
t
h
e m
a
nagers
of
t
h
e or
ga
ni
zat
i
ons of
p
r
o
d
u
ctio
n atte
m
p
t to
id
en
tify su
itab
l
e p
r
ofiles to
m
a
in
tain
and
dev
e
l
o
p t
h
ei
r p
r
o
d
u
ct
i
v
i
t
y
. Gi
ven t
h
a
t
t
h
e t
h
eory
base
d o
n
h
u
m
a
n res
o
u
r
ces
sho
w
t
h
at
t
h
e
go
od
p
r
od
u
c
tiv
ity in
the en
terp
ri
ses can
b
e
resu
lted
i
n
t
h
e
way how t
h
e
assig
n
m
en
t of hu
m
a
n
reso
urces is
per
f
o
r
m
e
d [1]
-
[
2]
. T
h
e
pe
rso
n
n
el
reassi
gnm
ent
can
be c
o
ns
i
d
ere
d
as
an important
f
actor to
dispose the hum
a
n
resources qu
ality th
at
can
b
e
cap
ab
le to
contrib
u
t
e fo
r m
e
e
tin
g
th
e d
e
sired
ob
j
ectiv
e
b
y
th
e
m
a
n
a
g
e
r an
d
its
serve
s
to
reduce the rec
r
uit
m
ent cost
s. I
n
t
h
e l
a
rg
e ent
e
rp
ri
ses o
r
a
d
m
i
ni
st
rat
i
ons
de
com
posed
by
several
pr
o
duct
i
o
n u
n
i
t
s and di
st
ri
but
ed acr
oss m
u
l
t
i
p
l
e
geo
g
ra
p
h
i
cal
si
t
e
s, t
h
i
s
r
eassi
gnm
ent
pr
ocesses o
r
p
r
o
cesses
m
o
b
ili
ty can
t
o
lead
to
im
p
r
o
v
e
o
r
t
o
m
a
in
tain
th
e pr
oductiv
ity w
ith
in
th
e pr
odu
ction
u
n
its an
d
t
o
mo
tiv
ate
the age
n
ts.
G
e
n
e
r
a
lly,
each
ag
en
t
o
c
cupy
a
po
st
i
n
the o
r
ig
in
al site an
d
w
a
n
t
to
ch
ang
e
th
is site an
o
t
h
e
r
vol
unt
a
r
i
l
y
or by
rede
pl
oy
m
e
nt
, i
s
i
d
ent
i
f
i
e
d by
i
n
di
vi
d
u
a
l
wei
ght
cal
cu
l
a
t
e
d by
basi
n
g
soci
o-
pr
ofe
s
si
onal
criteria fu
n
c
ti
o
n
. Th
is
weig
h
t
can
b
e
group
ed th
e co
m
p
eten
cy, the profitab
ility
and
th
e indiv
i
d
u
a
l
p
e
rform
a
n
ce …etc. In
o
r
d
e
r to
o
p
tim
ize th
e b
e
st
rep
a
rtitio
n of
p
e
rsonn
el an
d to
reach
i
n
g
th
e
ob
j
ecti
v
es fix
e
d
,
t
h
e
m
a
nager
s
m
u
st
t
o
di
spo
s
e an o
p
t
i
m
i
zing t
o
ol
s base
d
on a f
o
rm
al
m
odel
by
i
n
t
e
grat
i
n
g t
h
ese s
p
eci
fi
c
weigh
t
s an
d mo
b
ility co
nstrai
n
t
s [3
].
Th
e wo
rk
[4
] ad
dresses a theo
retical
m
o
d
e
l b
a
sed
fo
r em
p
i
rical
facts
o
n
rep
a
rtitio
n
o
f
work
ers
acros
s cl
ust
e
rs
wi
t
h
di
f
f
ere
n
t
l
a
bor
pr
o
duct
i
vi
t
y
. It
s
k
e
y id
ea is to
assu
m
e
th
at
there are restrictions on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
19
07
–
1
919
1
908
cap
acity o
f
wo
rk
ers fo
r clusters with
h
i
gh
produ
ctiv
ity. Th
e wo
rk
[5] is fo
cu
sed
to
find
a so
lu
t
i
o
n
for
assi
gnm
ent
p
r
obl
em
by
res
p
ect
i
ng t
h
e
di
ff
erent
c
o
nst
r
ai
n
t
s rel
a
t
i
v
e t
o
l
a
bo
r
reg
u
l
a
t
i
o
ns a
n
d a c
o
ns
t
r
ai
nt
relativ
e to
m
u
ltip
le sites, b
a
lan
ce th
e
wo
rk
load
o
v
e
r em
p
l
o
y
ees b
u
t
no
t add
r
essi
ng
in
t
h
e p
r
ob
lem
th
e i
m
p
act
t
h
ese di
s
p
l
ace
m
e
nt
s of
h
u
m
a
n re
so
ur
ces on
th
e
produ
ctiv
ity in
th
e sit
e
s and
n
o
t
takin
g
i
n
to
accoun
t th
e
co
nstrain
t
s
related
to
cyclic as
si
gnm
ent
an
d c
ons
er
vat
i
o
n
o
f
post
s
.
Gene
ral
l
y
, Ou
r
wo
rk i
s
t
h
e c
o
m
p
l
e
m
e
nt
s of ou
r p
r
evi
ous
wo
rk
s [
1
]
-
[
3
]
,
i
s
foc
u
sed
on
di
ve
rsi
t
y
o
f
p
o
s
sib
l
e form
s an
d
relation
s
h
i
p
b
e
t
w
een
t
h
e reassi
g
n
m
en
t wh
ich
also we lab
e
l th
e
m
o
b
ility
o
f
h
u
m
an
resources, and h
i
s
p
o
sitiv
ely i
m
p
act o
n
th
e p
r
od
uctiv
ity in
th
e en
terprise b
a
si
n
g
on
t
h
e sp
ecific
wei
g
h
t
b
y
resp
ecting
o
f
sev
e
ral co
nstrain
t
. Th
e fo
rm
u
l
atio
n
dev
e
l
o
p
e
d
in
th
is
wo
rk illu
strate th
at th
is p
r
ob
lem
c
a
n
b
e
cl
assi
fi
ed as const
r
ai
ne
d
c
o
m
b
i
n
at
o
r
i
a
l
opt
i
m
i
zat
i
on pro
b
l
e
m
s
and can be t
a
ken a ge
neral
i
zed assi
g
n
m
e
nt
p
r
ob
lem
fo
rm
k
nown
as
NP-h
ard
pro
b
l
em
[6
] an
d
its
p
r
act
ically si
mi
lar to
m
u
lti
in
d
e
x
assig
n
m
en
t p
r
ob
lems
(
M
I
A
Ps) [7
].
To
reso
lv
e t
h
is p
r
o
b
l
em
, we are in
terested to
i
m
p
l
e
m
en
t
th
e g
e
n
e
tic alg
o
rith
m
s
wh
ich
can
b
e
characte
r
ized
by an efficienc
y
to
so
lv
e NP-h
ard
prob
lem
s
. Add
itio
n
a
lly, th
e op
ti
m
a
l so
l
u
tio
n
ob
tain
ed
m
u
st
be ens
u
re
d the
im
provem
e
nt the avera
g
e productivity fo
r each unit of production th
rough the
right choice of
param
e
t
e
rs of t
h
i
s
p
r
o
b
l
e
m
.
For t
h
i
s
, F
o
r t
h
i
s
, t
h
e
genet
i
c
o
p
erat
ors
use
d
f
o
r st
a
r
t
i
ng t
h
e
genet
i
c
al
g
o
ri
t
h
m
are
uni
fo
rm
sel
ect
ion
,
ope
rat
o
r M
a
t
r
i
x
cr
ossi
ng
UX
(U
ni
f
o
rm
C
r
oss
o
ver
)
[8]
,
O
p
erat
or M
a
t
r
i
x
m
u
t
a
t
i
on
HPRM
th
at is co
m
b
in
ed
b
e
tween
unifo
rm
m
u
tatio
n
[9
] an
d
H
PRM
operat
or
[10].
W
e
call again that the
operators
UM
a
nd
HPR
M
h
a
v
e
prov
ided
en
cou
r
ag
ing
resu
lts in
so
lv
ing
th
e trav
elin
g
salesm
an
p
r
o
b
l
em
(
TSP
) w
h
ich
i
s
an
NP-
c
om
pl
et
e p
r
o
b
l
e
m
[10]
.
Th
is p
a
p
e
r is
o
r
g
a
n
i
zed
as fo
llo
ws: firstly we presen
t th
e
m
a
th
e
m
at
ical
form
al
is
m
mo
d
e
ling
th
i
s
ap
pro
ach and
ev
alu
a
te its com
p
lex
i
t
y
. Secon
d
l
y
we
will im
p
l
e
m
en
t g
e
netic alg
o
r
it
h
m
s
to
find
t
h
e
o
p
ti
m
a
l
rep
a
rtitio
n
of
h
u
m
an
resou
r
ces to im
p
r
o
v
i
ng
t
h
e av
er
ag
e
p
r
od
u
c
tiv
ity an
d to satisfying
it. C
o
m
p
u
t
atio
n
a
l
resu
lts
ob
tain
ed
o
n
rand
o
m
ly
g
e
n
e
rated in
stan
ces are
repo
rt
ed
to ev
al
u
a
te t
h
e
v
a
lid
ity appro
a
ch
.
2.
FORMULAT
ION OF
T
H
E PROBLEM
2.
1.
Represe
n
tati
on
of the Problem
Th
is
n
e
w work is con
s
id
ered
as th
e co
m
p
lemen
t
th
e
work p
r
esen
ted
in
p
a
p
e
r
[1
]-[2
]. In ad
d
ition
,
we
co
nsid
er th
at t
h
e en
terp
rise is d
e
co
m
p
o
s
ed in
to
sev
e
ral prod
u
c
tion
un
its
with
∈
1,
ge
og
ra
ph
i
cal
l
y
di
st
ri
b
u
t
e
d.
T
h
e n
u
m
b
er o
f
e
m
pl
oy
ee
wis
h
i
n
g to
dis
p
lace from
a unit
t
o
a
not
her
is sy
m
b
o
lized
b
y
and is
possess
ed the
s
p
ecific
weight
,
∈
1
,
. Th
e set o
f
th
is
g
r
ou
p
is:
,
,
,….,
Ket:
∑
is the
num
ber
of the ca
ndidat
e’
s em
ployees
to reas
sign from
E
j
to
E
k
. (
∀
∈
1,
).
i
s
t
h
e
bi
val
e
nt
va
ri
abl
e
s
,
t
h
e
em
pl
oy
ee i
s
re
assi
gne
d
f
r
o
m
one
uni
t
E
j
t
o
anot
her
E
k
if
= 1 and
0
ot
he
rwi
s
e.
In
itially, th
e all cand
i
d
a
tes
wish
ing
to m
o
b
ilize m
u
st b
e
v
e
rified
th
e
ob
jectiv
e co
nstrain
t
giv
e
n
b
y
:
1
(1
)
α
is th
e t
o
leran
c
e co
efficien
t
param
e
trizin
g
this con
s
train
t
.
W
is the a
v
era
g
e
weight enge
ndered
by
ca
n
d
i
d
at
es m
oved t
o
t
h
e
uni
t
.
W
is th
e av
erag
e
weigh
t
of candid
a
tes in
itially
worked in
t
h
e
u
n
i
t
.
Fo
r th
e prob
lem
stu
d
i
ed
,
we
can
list sev
e
ral
con
s
train
t
s:
2.
2.
Cos
t
c
o
ns
tr
aint
We ass
u
m
e
t
h
at
t
h
e post
s
ca
n be
occ
upi
e
d
by
an a
g
ent
1
w
h
i
c
h c
ons
um
es a vari
et
y
of t
h
e
enterprise re
s
o
urces
. The
global cons
umed resources
is
li
mited
b
y
th
e cap
acity resou
r
ce
. T
h
e
expressi
on of
resource c
o
nstraint as follows:
∑∑
,
2
(2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
An
Op
timizin
g App
r
oa
ch
for Mu
lti
Co
nstra
i
n
t
s
Rea
ssi
gmen
t Pro
b
l
em
o
f
Huma
n Reso
u
r
ces (S
a
i
d
Tka
t
ek)
1
909
2.
3.
Equilibrium co
nstra
i
nt
o
f
po
sts
This constrai
nt explain t
h
at the num
b
er of
oc
c
upi
e
d
p
o
st
s by
t
h
e c
a
ndi
dat
e
s real
l
o
cat
ed
fr
om
depa
rt
m
e
nt
E
j
t
o
un
it
E
k
m
u
st
be equal the num
b
er of vacat
ed post
s by candi
dates which leave the sa
me unit.
The e
x
pressi
on
o
f
t
h
i
s
c
o
nst
r
a
i
nt
i
s
ex
pre
sse
d
by
:
∑∑
∑∑
3
(3
)
2.
4.
Capaci
ty cons
traint
The ca
paci
t
y
const
r
ai
nt
req
u
i
r
es
whi
c
h t
h
e
num
ber
N
k
of
agent
reassi
gn
ed t
o
t
h
e
de
pa
rt
m
e
nt
is
li
mited
b
y
an
i
m
posed
by
t
h
e
m
a
nagers;
t
h
e
exp
r
essi
on
i
s
g
i
ven a
s
f
o
l
l
o
ws
:
∑∑
4
(4
)
whe
n
t
h
e
resources
of the
posts are i
d
entical,
th
e res
o
urce
constraint ca
n
be
reduce
d ca
pacity.
2.
5.
Priority c
o
nstraint
These weights are
the
elem
en
t
s
of th
e su
b-m
a
trix
,
:
1
1
;
;
1
;
Or:
∑
l
0
∀
∈
1
,
5
(5
)
2.
6.
Unique
ness c
o
nstr
aint
An
ot
he
r c
onst
r
ai
nt
can
be ad
ded t
o
t
h
e
ot
h
e
r co
nst
r
ai
nt
s
t
h
at
i
s
t
h
e u
n
i
que
ness c
o
nst
r
ai
nt
whi
c
h
expl
ai
n
s
a ca
n
d
i
d
at
e ca
n’t
ha
ve t
w
o
p
o
st
s si
m
u
lt
aneou
s
l
y
i
n
t
w
o
de
part
m
e
nt
s
E
k
:
∑
1
∀
∈
1,
,
∀
∈
1,
6
(6
)
Based
o
n
th
e
wo
rk
[
1
], th
e
num
b
e
r
o
f
cand
idate r
eassign
ed
to
th
e
un
it of
pr
odu
ctio
n
E
k
ca
n be gi
ve
n by
:
∑∑
7
(7
)
The
gl
o
b
al
wei
ght
e
n
gen
d
e
r
ed
by
t
h
e
assi
g
n
e
d
a
g
ent
s
i
s
gi
v
e
n
by
:
∑∑
8
(8
)
The a
v
era
g
e
weight ass
o
ciate
d
with
N
k
assi
g
n
ed
age
n
t
s
i
s
g
i
ven
by
:
9
(9
)
Let :
∑∑
10
(1
0)
Th
e co
m
b
in
atio
n of equ
a
tio
n
(1) an
d (9
) is
giv
e
n
b
y
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
19
07
–
1
919
1
910
0
If we p
u
t:
, we
will co
nstru
c
t
a sub
m
a
trix
assig
n
m
en
t no
ted
,
or
:
/
1
;
1
;
Si
m
ilarly, we co
n
s
t
r
u
c
t a su
b-matrix
X
k
su
ch
th
at:
,
/
1
;
1
;
We ass
u
m
e
that the tolera
nce
coefficient
ass
o
ciated
with
un
it
pr
odu
ctio
n
E
k
are e
q
uals.
So, the
g
l
ob
al fo
rm
u
l
atio
n
o
f
th
is
p
r
ob
lem
is written
as fo
llows:
∑
11
(1
1)
Un
de
r t
h
e
co
ns
t
r
ai
nt
s:
0∀
∈
1,
∑
0
∀
∈
1
,
∑∑
∀
∈
1,
∑∑
∀
∈
1,
∑
1
∀
∈
1,
,
∀
∈
1,
(1
2)
The wei
g
ht
ge
nerat
e
d by
candidates
reassigned to t
h
e
unit
is:
∑∑
13
(1
3)
The wei
ght
ge
nerat
e
d by
t
h
e
sam
e
nu
m
b
er
of can
di
dat
e
s m
oved from
t
h
e uni
t
is:
∑∑
14
(1
4)
The im
provement wei
ght
by t
h
e
reassign
m
e
nt processes a
s
sociated t
o
the
unit
is:
∆
∆
∑
∑
15
(1
5)
Gene
ral
l
y
, wi
t
h
i
n
u
n
i
t
o
f
pr
o
duct
i
o
n,
t
h
i
s
ga
p e
x
p
r
esses
t
h
e
im
pro
v
i
n
g
gl
o
b
al
wei
ght
i
f
∆
0
an
d h
e
exp
l
ain
th
e equ
ilib
ri
u
m
of th
is
g
l
ob
al
weigh
t
if
∆
.
3.
COMPLE
X
I
TY OF THE
PROBLEM
The assi
g
n
m
e
nt
s of can
di
dat
e
s t
o
t
h
e uni
t
s
o
f
pr
o
duct
i
o
n ar
e l
i
nked. I
n
ad
di
t
i
on, t
h
e
fo
r
m
ul
at
i
on of
th
e p
r
ob
lem is
p
r
actically si
milar to
th
e
m
a
t
h
em
at
ical
m
odel
for t
h
e Ge
ne
ral
i
zed Assi
gn
m
e
nt
Pro
b
l
e
m
(G
AP
)
th
at it’s NP-h
ard [1
1
]
and
can b
e
con
s
id
ered
su
ch
as m
u
lti in
d
e
x
assign
m
e
n
t
pro
b
l
em
s (MIAPs).
The ge
ne
ralized assignm
e
nt problem
(GAP)
exam
ines
t
h
e m
i
nim
u
m
cost
assi
gnm
ent
of
n j
o
bs t
o
m
agents
suc
h
t
h
at each
job is
assigne
d t
o
e
x
actly one a
g
en
t subject to ca
pacity restrictions
on t
h
e a
g
e
n
ts. T
h
e
gene
ral
i
zed ass
i
gnm
ent
pr
obl
e
m
i
s
an NP-
h
a
r
d c
o
m
b
i
n
at
ori
a
l
opt
i
m
i
zat
i
on pr
obl
em
[12]
.
The f
o
rm
ul
at
i
on o
f
th
is prob
lem
is
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
An
Op
timizin
g App
r
oa
ch
for Mu
lti
Co
nstra
i
n
t
s
Rea
ssi
gmen
t Pro
b
l
em
o
f
Huma
n Reso
u
r
ces (S
a
i
d
Tka
t
ek)
1
911
(G
AP
) :
∑∑
∑
∀
∈
∑
1
∀
∈
c
ij
: th
e co
st of assign
ing
j
o
b
j
to
ag
en
t
I
;
a
ij
: the
capacit
y
ab
so
rp
tio
n wh
en job
j
is assig
n
e
d
to ag
en
t
i
;
b
i
: the
availa
ble capa
c
ity of
agent
i
;
x
:
t
h
e assi
gn
m
e
nt
vari
a
b
l
e
equal
s
1
if
ag
en
t
i
is to p
e
rfo
r
m
j
o
b
j
,
0
ot
h
e
r
w
i
s
e.
This table
sum
m
arizes the c
o
rres
ponde
nce
between
GAP a
n
d MCRPHR :
Tabl
e
1. C
o
r
r
e
s
po
n
d
e
n
ce
of t
h
e c
o
m
p
l
e
xi
t
y
bet
w
ee
n
(G
AP
) a
n
d
(M
C
R
P
H
R
)
Pro
b
le
m :
G
A
P
Pro
b
le
m :
R
P
H
R
C
fixed value
k
Object
i
Agent Candidate
(i,j)
Job
j
Post in Depar
t
m
e
nt
E
k
Maxi
m
i
z
e
the pro
f
ile
M
a
xim
i
ze the weight
,
Resour
ce constr
aint
∑
Capacity constrain
t
,
Priority Constraint:
1
Priority Constraint:
0
X
1
The displace
m
e
nts
between the units
of pr
od
uction ar
e
dependents
We no
te th
at t
h
e reallo
cation p
r
o
b
l
em
can
be d
eco
m
p
o
s
ed
in
to
sev
e
ral in
t
e
r-related
sub
reallo
catio
n
pr
o
b
l
e
m
t
oget
h
er by
a vari
a
b
l
e
k
and
o
t
her con
s
trai
n
t
related
to
th
is
v
a
riab
le su
ch
as th
e capacity
con
s
t
r
ai
nt
,
p
o
s
i
t
i
ons C
o
nse
r
v
a
t
i
on
of c
o
nst
r
ai
nt
an
d st
rai
n
p
o
st
s c
o
st
.
A
su
b
pr
o
b
l
e
m
can
be f
o
rm
ulat
ed as
follo
w fo
r
k = k
0
:
Un
de
r t
h
e
co
ns
t
r
ai
nt
s:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
19
07
–
1
919
1
912
:
:
0
∑
0
,
∀
∈
1
,
∑∑
,∀
∈
1
,
∀
∈
1,
∑∑
,∀
∈
1
,
∀
∈
1,
∑
1
,
∀
∈
1
,
∀
∈
1,
We n
o
t
i
ce t
h
at
t
h
i
s
fo
rm
ul
at
ion i
s
si
m
i
l
a
r to ge
ne
ral
i
zed
assi
gnm
ent
pr
obl
em
(GAP
)
kn
o
w
n i
n
t
h
e
literatu
re as a
NP-h
ard
p
r
ob
le
m
[1
2
]
. Fo
r t
h
is, th
e co
m
p
lex
ity o
f
th
e m
u
lti co
n
s
trai
n
t
s reassign
m
e
n
t
p
r
ob
l
e
m
can
b
e
ev
alu
a
t
e
d
an
d ju
stified
b
y
th
e co
m
p
lex
ity o
f
th
e
(GAP).
Fo
r a g
i
v
e
n
in
stan
ce, o
u
r prob
lem stu
d
i
ed
con
s
ist to
search
th
e op
timal so
lu
tio
n
s
allo
wing
t
o
m
a
xim
i
ze t
h
e ob
ject
i
v
e
f
unct
i
on
F(X)
by
us
i
ng a
m
e
t
a
heur
i
s
t
i
c
al
gori
t
h
m
suc
h
as t
h
e
gen
e
t
i
c
al
gori
t
h
m
.
4.
GENETIC
ALGORIT
HM OF RESOL
U
TION
4.
1.
Genetic algori
thm
The m
e
t
hod
u
s
ed f
o
r s
o
l
v
i
n
g t
h
i
s
pr
o
b
l
e
m
i
s
t
h
e gene
t
i
c
al
gori
t
h
m
s
. It
i
s
pa
rt
o
f
t
h
e fam
i
l
y
of
ev
o
l
u
tio
n
a
ry alg
o
rith
m
s
. Th
ey h
a
v
e
attracted
th
e in
tere
st of
m
a
ny researc
h
ers starting wit
h
Holland [13]
, who
devel
ope
d t
h
e
basi
c p
r
i
n
ci
pl
es t
h
r
o
u
g
h
Gol
dbe
r
g
[
14]
,
w
h
i
c
h
use
d
t
o
s
o
l
v
e
real
p
r
o
b
l
e
m
s
of
opt
i
m
izat
i
on.
Ot
he
r re
searc
h
ers
have
f
o
l
l
o
wed
t
h
i
s
pat
h
i
n
cl
u
d
i
n
g
Davi
s
[
15]
, M
a
hf
o
u
d
[
16]
, M
i
c
h
al
ewi
cz [
1
7]
, et
c
…
In
ou
r ap
pr
oac
h
, we
use
d
as a real
enco
di
n
g
m
e
t
hod
of re
prese
n
t
a
t
i
o
n
,
t
h
e bi
na
ry
rep
r
esent
a
t
i
on
o
f
sub-m
a
trix, each sub m
a
trix is com
pos
ed
by
lines vectors t
h
at can
be
enc
ode
d
by an a
r
ray of Booleans
:
0
or
1
.
T
h
e a
ppl
i
e
d
m
e
t
hod
s i
n
t
h
i
s
ap
pr
oac
h
a
r
e:
Ran
d
o
m
g
e
n
e
ratio
n
o
f
th
e i
n
itial p
o
p
u
l
ation
;
Uni
f
orm
Selection
(US);
Matrix
Cro
s
sov
e
r (MOX);
Matrix
Mu
tation
(MMP);
In
sertion
Metho
d
(i
n
s
erting
el
itis
m
)
.
4.
2.
Gl
ob
al
al
gori
t
hm
The
gl
o
b
al
al
g
o
ri
t
h
m
i
s
com
pose
d
of t
h
e
fol
l
owi
n
g
i
n
st
r
u
ct
i
ons:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
An
Op
timizin
g App
r
oa
ch
for Mu
lti
Co
nstra
i
n
t
s
Rea
ssi
gmen
t Pro
b
l
em
o
f
Huma
n Reso
u
r
ces (S
a
i
d
Tka
t
ek)
1
913
5.
EX
P
E
R
I
M
E
NT
5.
1.
Descripti
o
n
of the
instance
of the s
t
udied problem
In th
is
sectio
n, we presen
t an ex
am
p
l
e to
valid
th
is app
r
oach
b
y
u
s
ing
t
h
e
g
e
n
e
tic algo
rith
m
.
For
this, we consi
d
er an e
n
terpri
se co
m
p
o
s
ed
of
th
r
ee pr
odu
ctio
n
un
its (
E1
,
E2
and
E3
). O
u
r w
o
rk ap
p
r
o
ach is
b
a
sed
o
n
th
e follo
wing
h
ypo
theses:
Each
un
it in
clu
d
e
s a nu
m
b
er o
f
ag
en
ts wan
tin
g
t
o
ch
ang
e
th
eir
o
r
ig
i
n
al p
o
s
t of wo
rk
b
y
tak
i
ng
into
account their c
hoices
.
The
data ass
o
ciated with individual
weights of ca
ndidate
s
are ra
ndom
l
y g
e
n
e
rated
i
n
th
e in
terv
al [5
0,
10
0]
.
The all
posts
are possesse
d t
h
e
sam
e
cost value
.
The t
o
lera
nce
coefficients
value is i
d
entical for each unit of
production: (
).
The Ta
ble (2, 3 and 4) summarize the value
s
of the
i
ndi
vidual’s weight of candi
dates in each unit
of
pr
o
duct
i
o
n a
n
d
fo
r seve
ral
of
t
o
l
e
rance c
o
e
f
f
i
ci
ent
val
u
es
α
.
M
o
re
ove
r, t
h
e
Tabl
e 5 s
u
m
m
a
ri
ze t
h
e
const
r
ai
nt
val
u
es
C
k
and t
h
e a
v
era
g
e
wei
ght
value
s
.
/1
STEP 1
Define the tolera
nce coeffici
ent
α
Define the number of units
NU
Define the candi
date’s
num
ber in each unit
E
j
Define the precis
i
on
Constructing the global m
a
trix:
/1
STEP 2
Generating the i
n
itial population
com
p
o
s
ed of
N
binar
y
m
a
trix assignm
ent :
STEP 3
Calculate
∑
and
ch
eck
constraints,
Arran
g
in
g
the solutions in the list
N
STEP 4
I
t
eration=
1
Repeat
S
e
lect
two solutions X1 and X2 from
the list N
If
Max
1
,
2
M
ax
1
,
2
and
ch
eck
cons
traints
The
n
Cross
X
1
and X
2
:
X1
XC1 and X2
XC2
If
1
2
then
Mutate
,
XC1
XM
Else
Mutate
2
,
XC2
XM
End if
If
F(
XM)
> Max
(
F
(
XC1)
,F
(
X
C2))
an
d
verifies
th
e cons
traints
then
In
se
r
t
the solution in the list
N
I
t
eration
=I
teration
+
1
Els
e
R
e
ject
XM
End if
Until
0
/
0
END
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
19
07
–
1
919
1
914
Tabl
e
2. C
h
a
n
g
e
o
f
post
t
o
t
h
e
u
n
i
t
E1
(
k
=1
)
Original site
(
α
=0)
(
α
=1/4)
(
α
=1/3)
(
α
=1/2)
(
α
=2/3)
(
α
=3/4)
(
α
=1)
E2
97
76
70,
67
60
49,
33
44
28
E2
75
67
61,
67
51
40,
33
35
19
E2
69
43
37,
67
27
16,
33
11
-
5
E2
67
43
37,
67
27
16,
33
11
-
5
E2
64
29
23,
67
13
2,
33
-
3
-
19
E2
57
22
16,
67
6
-
4
,
67
-
10
-
26
E2
54
20
14,
67
4
-
6
,
67
-
12
-
28
E2
51
16
10,
67
0
-
10,
67
-
16
-
32
E2
50
13
7,
67
-
3
-
13,
67
-
19
-
35
E3
96
80
74,
67
64
53,
33
48
32
E3
50
39
35
26,
5
18
13,
75
1
Tabl
e
3. C
h
a
n
g
e
o
f
post
t
o
t
h
e
u
n
i
t
E2
(
k
=2
)
Original site
(
α
=0)
(
α
=1/4)
(
α
=1/3)
(
α
=1/2)
(
α
=2/3)
(
α
=3/4)
(
α
=1)
E1
90
83
79
70,
5
62
57,
75
45
E1
89
78
74
65,
5
57
52,
75
40
E1
83
76
72
63,
5
55
50,
75
38
E1
82
69
65
56,
5
48
43,
75
31
E1
77
58
54
45,
5
37
32,
75
20
E1
63
21
17
8,
5
0
-
4
,
25
-
17
E1
62
9 5
-
3
,
5
-
12
-
16,
3
-
29
E3
66
86
82
73,
5
65
60,
75
48
E3
54
74
70
61,
5
53
48,
75
36
E3
51
60
56
47,
5
39
34,
75
22
E3
51
35.
25
53
44,
5
36
31,
75
19
E3
50
39
35
26,
5
18
13,
75
1
Tabl
e
4. C
h
a
n
g
e
o
f
post
t
o
t
h
e
u
n
i
t
E3
(
k
=3
)
Original site
(
α
=0)
(
α
=1/4)
(
α
=1/3)
(
α
=1/2)
(
α
=2/3)
(
α
=3/4)
(
α
=1)
E1
90
83
79
70,
5
62
57,
75
45
E1
89
78
74
65,
5
57
52,
75
40
E1
83
76
72
63,
5
55
50,
75
38
E2
82
69
65
56,
5
48
43,
75
31
E2
77
58
54
45,
5
37
32,
75
20
E2
63
21
17
8,
5
0
-
4
,
25
-
17
Table
5. T
h
e
e
x
am
ple of a
v
e
r
age
weig
ht
a
n
d ca
paci
t
y
co
n
s
t
r
ai
nt
val
u
es
Average
weight
W
96,
2
79,
72
114,
71
Capacity constr
aint C
k
90
83
79
5.
2.
Results
and
discussion
Th
e fo
llo
wi
n
g
tab
l
e su
mmarizes th
e
m
a
in
resu
lts of
o
u
r
t
e
st
. They
wer
e
obt
ai
ne
d o
n
a com
put
er
havi
ng C
P
U
P
e
ntium i
5
2.
5
GHz
with
4GB
of
RAM
. Th
e genet
i
c
al
go
r
i
t
h
m
used was
enco
de
d by
t
h
e
C++
l
a
ng
uage
.
Fo
r
b
e
tter in
t
e
rpretatio
n, the o
b
t
ain
e
d
n
u
merical resu
lts are graph
i
cally v
i
su
alized
in
th
e n
e
x
t
figures
.
On the prim
ary experience,
we interest to
sho
w
t
h
e i
n
fl
ue
nce
of n
u
m
b
er o
f
can
di
dat
e
s o
n
t
h
e
gene
rat
i
o
n t
i
m
e
and
co
n
v
er
ge
nce t
i
m
e. In or
der t
o
d
o
t
h
is
,
we take t
h
ree i
n
stances c
o
m
pose
d
consec
utively of
29
,
58
and
11
6
candidates
. In th
is test, th
e gen
e
tic alg
o
rithm is started
with
40
i
ndi
vi
d
u
a
l
s
consi
d
ere
d
as t
h
e
in
itial p
o
p
u
l
atio
n
size, and
50
iteratio
n
s
as th
e sto
p
p
i
ng
co
nd
itio
n
app
lied
[1
]. Th
ese
two
param
e
ters were
expe
ri
m
e
nt
al
ly j
u
st
i
f
i
e
d
by
t
h
e fi
g
u
re
s a
n
al
y
s
i
s
.
Oth
e
r resu
lts
are
p
r
esen
ted,
th
ey con
cern
t
h
e st
u
d
y
o
f
the ti
m
e
co
n
s
umed
to
g
e
nerate th
e in
itial
populations
for diffe
re
nt sizes accordin
g on num
ber of
ca
ndidates wishi
n
g
to
c
h
a
nge t
h
e
i
r positions a
n
d these
resu
lts are illustrated
in
th
e
Fig
u
re 1. By reg
a
rd
ing
th
is fig
u
re, we can
o
b
s
erv
e
th
at the g
e
n
e
ratio
n
ti
me i
s
in
creased
with
th
e effectiv
e of can
d
i
d
a
tes. In ad
d
itio
n, if we elev
ate th
e n
u
m
b
e
r o
f
cand
i
d
a
tes b
y
a p
o
r
t
i
o
n
2
or
3
, t
h
e
ge
ner
a
t
i
on t
i
m
e can al
so
be asses
s
e
d
by
t
h
e
sam
e
po
rt
i
o
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISS
N
:
2088-8708
An
Op
timizin
g App
r
oa
ch
for Mu
lti
Co
nstra
i
n
t
s
Rea
ssi
gmen
t Pro
b
l
em
o
f
Huma
n Reso
u
r
ces (S
a
i
d
Tka
t
ek)
1
915
Fi
gu
re
1.
Ge
ne
rat
i
o
n
t
i
m
e
dep
e
ndi
ng
o
n
n
u
m
b
er
o
f
i
n
di
vi
du
al
s
Th
e
m
a
in
g
o
a
l o
f
th
is research
is to
carry
o
u
t
in
v
e
stig
ati
o
n
of th
e in
fluen
ce o
f
on
e o
f
th
e k
e
y GA
param
e
t
e
rs – p
o
p
u
l
a
t
i
on si
ze
(num
ber of chr
o
m
o
so
m
e
s) –
on t
h
e al
gori
t
h
m
perform
ance for i
d
ent
i
f
i
cation o
f
an o
p
t
i
m
a
l
m
a
tri
x
of
reassi
gn
m
e
nt
of candi
d
a
t
e
s.
On
Fi
g
u
re 2, th
e precision
issu
e th
e fitn
ess fun
c
tio
n
v
a
lues, ob
tain
ed
du
ri
n
g
th
e
60
iteratio
n
s
b
y
usi
n
g a
di
ffe
re
nt
p
o
pul
at
i
o
n
si
ze:
10
, 20
,3
0
an
d
40
are
sh
own
.
Each t
e
st con
s
ists to d
e
term
in
e th
e go
od
p
opu
latio
n
size to
lead to
t
h
e
b
e
st so
lu
tion
.
Fi
gu
re
2.
Preci
si
on
fact
o
r
de
p
e
ndi
ng
o
n
e
x
ec
ut
i
o
n
Ti
m
e
(
W
i
t
hout
Ge
nerat
i
on
t
i
m
e
) fo
r a
n
i
n
st
ance
com
pose
d
of
29
can
di
dat
e
s
The
preci
si
on
f
act
or cal
c
u
l
a
t
e
d
by
t
h
i
s
e
x
pre
ssi
on:
Whe
r
e
S
opt
is th
e av
erag
e so
lu
tion
op
ti
m
a
l ob
tain
ed fo
r
10
tests a
n
d
S0
is
th
e b
e
st so
lu
tion
issu
e
fro
m
th
e in
itial p
opu
latio
n
.
Th
e
g
r
aph
i
cal
resu
lts sho
w
th
at th
e
GA cou
l
d
no
t fi
n
d
accu
rate so
lu
ti
on
u
s
ing
sm
al
l
popul
at
i
on si
ze u
n
d
er
20
chrom
o
somes. That nee
d
s at least
30
chr
o
m
o
som
e
s i
n
po
pul
at
i
o
n
fo
r
achieving
a better
sol
u
tion of the
gene
ration
ti
m
e
according to the
num
ber
of
genetic population.
We
can
also
co
n
c
l
u
d
e
th
at t
h
e
p
r
ob
ab
ility to
o
b
t
ain a
p
r
ecisio
n
factor app
r
ox
im
a
t
ely eq
u
a
l to
0
.
On Fi
gu
re 3, t
h
e fi
t
n
ess f
u
nc
t
i
on val
u
e
s
, o
b
t
ai
ned d
u
ri
ng t
h
e
60
al
go
ri
t
h
m
runs f
o
r
40
indi
vidual are
sho
w
n. It
ca
n b
e
obse
r
ved i
n
t
h
i
s
fi
g
u
re t
h
at
t
h
e i
m
provem
e
nt
of t
h
e fi
t
n
es
s fu
nct
i
o
n (
F
) i
s
rapi
d an
d bec
o
m
e
s
slo
w
er un
til the stag
n
a
tion
of th
is fun
c
tio
n. Th
is st
ag
n
a
ti
o
n
is t
o
sh
ow th
e conv
erg
e
n
ce of th
e
objectiv
e
fu
nct
i
o
n t
o
a
n
opt
i
m
al
val
u
e
Wop
t
=106
1.7
fo
r
α
. Th
e o
p
t
i
m
al
so
lu
tion
X
ob
tain
ed corresp
ond
s to g
l
o
b
al
weigh
t
ed
m
a
trix
th
at is
p
r
esen
ted
as fo
llow:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
JECE
Vo
l. 6
,
N
o
. 4
,
Au
gu
st 2
016
:
19
07
–
1
919
1
916
1
2
3
and
1
2
3
Wi
t
h
:
1
1
11100000
110000000
2
1111000
1100000
3
110
110
The value
1
indicates that the
em
ployee is
re
assi
gne
d,
h
o
w
e
v
er
t
h
e
val
u
e
0
indicates t
h
at the
e
m
ployee is not reassigne
d
.
Fi
gu
re 3.
Va
ri
at
i
on of
fi
t
n
ess
f
unct
i
o
n de
pe
nd
i
ng o
n
num
ber of
i
t
e
rat
i
ons
an
i
n
st
ance
com
pose
d
of 2
9
candi
dat
e
s
Th
e algor
ith
m
d
e
scr
i
b
e
d above h
a
s
v
a
r
i
o
u
s
p
a
r
a
m
e
ter
s
, inclu
d
i
ng
t
h
e influ
e
n
tial v
a
lu
es str
o
n
g
l
y
o
n
its q
u
a
lity. I
ndeed
,
w
e
co
nducted
an
em
p
i
r
i
cal stu
d
y
o
f
t
h
e adj
u
stm
e
n
t
of
th
ese
p
a
r
a
m
e
ter
s
to
o
b
t
ain
g
ood
resul
t
s
of
pe
rf
orm
a
nce. I
n
t
h
i
s
ap
p
r
oac
h
,
we h
a
ve i
n
t
e
g
r
at
ed t
h
e
o
p
er
at
or
uni
fo
rm
m
a
t
r
i
x
cros
so
v
e
r a
n
d
m
u
t
a
t
i
on m
a
t
r
ix (m
at
ri
x HPR
M
) [
3
]
,
[
10]
. T
h
ese t
w
o
ope
ra
tors a
r
e operated altern
ately followi
ng a
so-called
"Flip
-Fl
o
p
crossov
e
r-m
u
tatio
n
.
"
[3
].
Th
is i
m
p
l
ies th
at th
e m
u
t
a
ti
on o
p
e
r
at
or i
s
a
b
l
e
t
o
be reac
he
d
onl
y
i
n
t
h
e
case wh
en
th
e
so
lu
tion
ob
tained
b
y
th
e cro
s
sin
g
p
r
o
cedure d
o
e
s no
t satisfy all
th
e co
n
s
t
r
ain
t
s. In
add
itio
n, the
choi
ce o
f
st
o
p
p
i
n
g cri
t
e
ri
o
n
base
d o
n
t
h
e st
abi
l
i
zat
i
on of
t
h
e fitn
ess fun
c
tio
n
streng
thens the converge
nce of
th
is
ap
pro
ach
to
ward
s g
ood
so
lu
tion
.
Th
ese
n
e
w p
r
oce
d
ures al
s
o
al
l
o
w re
du
ce t
h
e com
putation time and
i
m
p
r
ov
es so
lu
t
i
o
n
t
h
e so
lu
tion
[3
].
In
o
r
de
r t
o
bet
t
er j
u
dge
t
h
e
v
a
l
i
d
i
t
y
of
ou
r a
p
p
r
oach
,
we m
a
ke a
com
p
ari
s
on
bet
w
ee
n
ou
r a
p
p
r
oa
c
h
and
ge
net
i
c
a
p
pr
oac
h
used
t
o
sol
v
e t
h
e
fi
r
s
t
wo
rk
“
A
Meta-h
eu
ristically App
r
o
a
ch
o
f
the Sp
atial
Assign
m
e
n
t
Problem
of Hum
a
n Resourc
e
s in Mu
lti-sit
e
s Enterprise”
[1] according to
genetic operators and the stop
co
nd
itio
n.
In t
h
is
work
we ap
p
lied th
e m
a
x
i
m
u
m
n
u
m
b
e
r
o
f
iteratio
n
s
as a stopp
ing criterion
wh
ich
we
p
r
ed
efi
n
ed
i
n
t
h
e on
set run
n
i
n
g
. Howev
e
r,
th
is criterion
is in
sufficien
t t
o
falling
in
to
t
h
e b
e
st so
lu
ti
on
.
In
ad
d
ition
,
th
e ap
p
lication
of classical cro
ssov
e
r an
d
m
u
ta
ti
o
n
o
p
erators i
n
th
e work
[1
]
serv
es
o
n
l
y serv
es t
o
p
r
ov
id
e a so
l
u
tio
n
in
m
o
re ti
m
e
an
d
lo
wer qu
ality co
mp
ared
to
th
e so
lu
tion
ob
tained
b
y
our appro
a
ch
descri
bed
i
n
t
h
i
s
pa
per.
Th
rou
g
h
a synth
e
sis of th
e literatu
re
o
n
human
resource
allo
catio
n
p
r
ob
lem
s
, th
ere is n
o
i
n
stan
ce
targeted to
numerically co
mpare the
result
s associated
of th
ese p
r
ob
lem
s
with
th
e resu
lts o
b
t
ained
b
y
our
approach. Inde
ed, we are int
e
rested
to
fo
cu
s on
th
e wo
rk
in
treatin
g
certain
term co
n
s
train
e
d
assign
m
e
n
t
pr
o
b
l
e
m
sol
v
i
n
g
by
genet
i
c
al
go
ri
t
h
m
s
. The
com
m
on p
o
i
n
t
f
o
cus
e
s
on
t
h
e use
o
f
c
o
nv
e
n
t
i
onal
st
o
p
c
r
i
t
e
ria
Evaluation Warning : The document was created with Spire.PDF for Python.