TELKOM
NIKA
, Vol. 11, No. 8, August 2013, pp. 46
0
9
~4
615
e-ISSN: 2087
-278X
4609
Re
cei
v
ed Ma
rch 7, 2
013;
Re
vised
Ma
y 18, 2013; Accepted Ma
y 29
, 2013
Study on Software Quality Improvement based on
Rayleigh Model and PDCA Model
Ning Jingfen
g*, Hu Ming
Coll
eg
e of Co
mputer Scie
nc
e and En
gi
neer
i
ng, Ch
an
gchu
n Univ
ersit
y
of T
e
chnolog
y
Cha
ngch
un 1
3
001
2, Chi
n
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: ning
jin
gf
en
g
@
mail.ccut.e
d
u
.cn, humin
g_c
cut@16
3.com
A
b
st
r
a
ct
As the softw
are in
dustry gra
d
ually
bec
o
m
es
matur
e
, softw
are q
ual
ity is r
egar
ded
as th
e life
of a
softw
are
enter
prise.
T
h
is article discuss
es how
to impr
ov
e the
q
ual
ity of
softw
are, ap
pl
ies
Rayl
eig
h
mod
e
l
and PDC
A
mo
del to the softw
are qual
ity mana
ge
me
nt,
comb
in
es w
i
th th
e defect remo
val effectiven
e
s
s
ind
e
x, exerts PDCA mod
e
l
to solve the
prob
le
m
of qu
ality man
a
g
e
ment obj
ectives
w
hen usin
g t
h
e
Rayle
i
g
h
mo
de
l in
bi
directi
o
n
a
l
qu
al
ity i
m
pr
ov
ement
strate
gi
es of softw
are
qua
lity
ma
nag
e
m
e
n
t, an
d p
u
ts it
into the a
ppl
ica
t
ion to achi
eve
goo
d results
.
Ke
y
w
ords
:
qu
ality man
age
ment, raylei
gh
mode
l,
PDCA mode
l, defect an
alysis
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1.
Introduc
tion
With co
ntinuo
us adva
n
ce o
f
the economi
c
gl
ob
alizatio
n, the era of kno
w
le
dge e
c
on
omy
has come. Software h
a
s
been the ra
pi
d rise a
s
an
emerging hig
h
-tech indu
st
ry [1]. The rapid
developm
ent
of inform
atio
n technol
ogy
ena
ble
s
soft
ware to
be
a
pplied
to va
ri
ous soci
al fie
l
ds,
and
software
quality b
e
comes the
fo
cu
s a
c
co
rdin
gly [2].
Dr.
Jo
sep
h
M.
Juran,
a fam
ous
American q
u
a
lity manage
ment scienti
s
t, pointed out
"T
he 21
st ce
ntury is the
century of qu
a
lity;
quality is the
mo
st effe
ctive
weapo
n
for
the
pea
ce
ful occu
patio
n of the market”. The
r
e is no
exceptio
n for the
softwa
r
e
indu
stry’s g
r
adual
matu
rity, softwa
r
e
q
uality is
bein
g
con
s
ide
r
ed
as
the life of the softwa
r
e ind
u
stry. Softwa
r
e qu
ality managem
ent ha
s develo
ped
in an all roun
d
way in
the
software
o
r
ga
nizatio
n
, st
ro
ng q
uality co
nsciou
sn
ess
is g
r
a
dually t
a
kin
g
root i
n
the
heart
s
of the
softwa
r
e te
chni
cal a
nd
manag
eme
n
t person
nel, till the formin
g of the wh
ole
orga
nization
quality cultu
r
e. The a
r
ticl
e co
mbine
s
usin
g reli
abili
ty model-Ray
leigh mo
del
and
quality impro
v
ement mod
e
l-PDCA mo
del, on the
b
a
si
s of u
s
ing
the existed
d
a
ta, analyzes the
defect, monit
o
rs a
nd evalu
a
tes the soft
ware’s
q
uality, and gives ju
dgment ba
si
s on whethe
r the
prod
uct can b
e
relea
s
e
d
or
not
.
2.
Ra
y
l
eig
h
Mo
del
Raylei
gh
model i
s
a
comm
on reli
ability model
that can p
r
edi
cate the
defect
s
distrib
u
tion of
the whole life cycle of sof
t
ware
d
e
velo
pment [3]. Rayleigh model
is a membe
r
of
the Weib
ull di
stributio
n fam
ily. It has bee
n severa
l de
cade
s sin
c
e
Weibull [4] distribution u
s
ed in
reliability anal
ysis in differe
nt engine
erin
g fields,
whi
c
h is one of th
e three famo
us extrem
e value
distrib
u
tion
s
(Tobi
as, 1
9
8
6
). On
e of its symb
olic
f
eature
s
i
s
th
at the tail of its pro
babili
ty
approa
che
s
zero gradu
ally, but canno
t reach
ze
ro.
In 1982, Trachte
nbe
rg [5] obse
r
ved
the
monthly defe
c
t data of a set of softwa
r
e proj
ect
s
, and found th
at the comprehen
sive def
ect
mode
of the
proj
ect
s
m
eet Rayleigh
cu
rve.
In
1
982, G
a
ffney of IBM Fe
deral
Syste
m
s
Dep
a
rtme
nt reporte
d, the
time di
strib
u
tion of th
e
software life
cy
cle of d
e
fect f
o
unde
d in
the
6
publi
c
d
e
fect
dete
c
ting
ph
ase
s
used
b
y
IBM, along
with
the
s
e
s
pha
se
s al
so
meet
Raylei
gh
curve. Its cumula di
stribution function (CDF
)
and
probability density fu
nction (PDF)
are as
follows
:
CDF:
m
c
t
e
t
F
)
/
(
1
)
(
PDF:
m
c
t
m
e
c
t
t
m
t
f
)
/
(
)
(
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 8, August 2013: 4609 –
4615
4610
Whe
r
e m i
s
the shape
pa
rameter,
c is t
he scale
pa
ra
meter, an
d t is time. Wh
en
applying
to software,
PDF often
indicates th
e defe
c
t de
nsity (rate)
cha
nge
s ove
r
time o
r
d
e
fect
occurre
n
ce
mode
(d
efect
i
ve data),
wh
ile CDF
indi
cates th
e o
ccurren
ce m
o
d
e
of
cum
u
lati
ve
defect.
In Wei
bull family
,
the two
mod
e
ls alrea
d
y used i
n
so
ftware
reli
abil
i
ty are th
e m
odel
s of
sha
pe pa
ram
e
ters
m
=
1 a
nd m=2. Ray
l
eigh mod
e
l is the sp
eci
a
l
case of Wei
bull distri
buti
on
m=2, where CDF a
nd PDF are a
s
follo
ws
:
CDF:
2
)
/
(
1
)
(
c
t
e
t
F
PDF:
2
)
/
(
2
2
)
(
c
t
e
c
t
t
t
f
PDF of Rayl
eigh first increa
se
s to pe
ak value, th
en de
crea
se
s at de
cre
a
sing rate.
Parameter
c
is
the func
tion of t
m
, and t
m
i
s
the tim
e
tha
t
the cu
rve
re
ach
e
s th
e p
e
a
k valu
e. Ta
ke
derivative of t from f (t), and make it be
ze
ro, solve th
e simultan
eo
us eq
uation t
o
get t
m
.
2
c
t
m
After
t
m
is est
i
mated
,
the shape
of the
whol
e curve
s
can
be dete
r
mined. Th
e a
r
ea of t
m
portion b
e
lo
w the curve i
s
39.35% of the total area.
The above fo
rmula in
dicates sta
nda
rd d
i
stributio
n; in particula
r, the total area bel
ow the
PDF cu
rve is
1. In practi
cal
application, the formul
a is
multiplied by the co
nsta
nt K (K is the total
number of defects or the total cumu
lative defect rate). If we still make
substitution in the formul
a,
2
m
t
C
We’ll get the
followin
g
formula. In orde
r to determi
n
e
model fro
m
one data p
o
i
n
t set,
K
and t
m
are pa
rameters to be
estimated.
]
1
[
)
t
(
2
2
)
2
/
1
(
1
t
t
m
e
K
F
2
m
2
)
2
/
1
(
2
t
1
)
(
t
t
m
te
K
t
f
Rayleig
h
mo
del involve
s
conte
n
ts
su
ch a
s
defe
c
t
preventio
n a
nd p
r
o
pha
se
defe
c
t
removal
relat
ed to
the
proj
ects in
ea
rly
pha
se. O
n
th
e ba
si
s of
thi
s
m
odel, if
redu
cing
the fi
lling
rate
of erro
r, the a
r
e
a
bl
ow th
e
Rayl
eigh
cu
re
wil
l
be
come
sm
aller,
re
sulting in
a sm
al
ler
predi
ction fiel
d defect rate, as sho
w
n in
Figure 1:
Figure 1.
Ray
l
eigh Mod
e
l I
Similarly, if more defe
c
ts a
r
e remove
d in the
early pha
se of develo
p
ment pro
c
e
ss, the defect
rate will be lo
wer in the lat
e
r pha
se of te
sting an
d mai
n
tenan
ce, a
s
sho
w
n in Fig
u
re 2.
Figure 3 de
scrib
e
s the
strategy to make qua
lity improvement fro
m
two dire
cti
ons. From
Figure 3, we
can see ou
r current qu
al
ity improv
ement target is
to redu
ce th
e height of the
curv
e
s
a
s
m
u
ch a
s
p
o
s
s
i
b
le,
mean
whi
l
e mov
e
t
he
cre
s
t
of
t
he
Ray
l
eig
h
cu
r
v
es t
o
t
he le
ft.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Study on Software Qualit
y Im
provem
ent based on
Ra
yleig
h
Model
and PDCA…
(Nin
g Jin
g
fen
g
)
4611
Among them,
I0: high level
desig
n review; I1: low lev
e
l de
sign
review; I2: cod
e
i
n
sp
ectio
n
; UT:
unit tes
t; CT: c
o
mponent tes
t; ST: s
y
s
t
em tes
t.
Figure 2. Ray
l
eigh Mod
e
l II
Figure 3.
Dire
ctional
Diag
ra
m of Develop
m
ent Quality Improveme
n
t
3.
PDCA M
ode
l Analy
s
is
PDCA is the
acro
nym for Plan, Do, Check
an
d Action. PDCA is also called
Demin
g
cycle, which is a cla
s
sic q
u
a
lity manage
ment m
odel p
r
omote
d
and
pra
c
tice
d in Japan by Dr. W.
Edwa
rd
s De
ming, an American qu
ality manage
ment
expert
,
as
shown in Figu
re 4.
Figure 4. PDCA Cycle Pro
c
e
s
s
Demin
g
cycl
e
is used a
s
a model to make cont
inu
o
u
s
pro
c
e
ss imp
r
ovement usi
n
g CMM by SEI,
whi
c
h wa
s called
I
D
EAL (Initiating,
Di
agno
sin
g
,
Le
veragin
g
) [7]. ISO9001:
2
000
stated
in
its
introdu
ction
that: PDCA
m
e
thod i
s
avail
able fo
r al
l
p
r
oce
s
se
s. Whi
l
e all
pro
d
u
c
ts a
r
e
the
re
sults
of process, t
he p
r
o
d
u
c
ts
quality is rel
a
ted to
the
p
r
ocess of
set
t
ing up
the
p
r
odu
cts [8].
The
improve
d
PDCA theo
ry h
a
s
been
wi
d
e
ly use
d
in
quality man
a
gement
of the ente
r
p
r
ise.
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 8, August 2013: 4609 –
4615
4612
Mean
while,
PDCA also become
s
a logical wo
rki
ng pro
c
e
s
s that enable
s
any effective
ac
tivities
[9]
.
3.1. Plan
It include
s th
e deci
s
io
n of prin
ciple
s
an
d aims,
an
d the formul
atio
n of activities pro
c
e
ss.
The
plan
re
q
u
ire
s
“5
W1
H”, that is,
wh
at, why
who,
when,
whe
r
e
a
nd h
o
w [8], find p
r
o
b
lem
s
and
find out the
reason
s an
d t
he mai
n
rea
s
ons,
set u
p
q
uality prin
cipl
es, aim
s
, lett
er of i
n
tention
and
manag
eme
n
t princi
ple
s
, etc. For exa
m
ple, t
he manag
ement
prin
ciple
s
ha
ve “pro
ce
ssi
ng
method
s”, “m
anag
ement system metho
d
s” a
nd “co
n
tinuou
s imp
r
ov
ements”, etc.
3.2. Do
The second
p
hase “Do
”
is
not the sim
p
l
e
“do
”
. Do i
s
t
he implem
ent
ation and
pra
c
tice
o
f
the plan, an
d
it is mainly to do a
c
cordin
g to t
he plan,
to implement
the pra
c
tical
measures, a
n
d
control the
proce
s
s, ena
bli
ng the
activities to
go fo
rward
as
expe
ctation and
fin
a
lly rea
c
h th
e
plan an
d the
target set. The imple
m
e
n
t of meas
ures
shall in
cl
ude 3 p
a
rts
of content
s: Do,
control and
re
gulate.
3.3. Check
Che
c
k is an
evaluation fo
r the effect
afte
r impl
eme
n
tation. Ch
eck
is a
c
comp
ani
ed the
impleme
n
tation process from begi
nning
to end, it is
the process of
contin
u
o
u
s
ly colle
cting d
a
t
a,
and getting i
n
formatio
n, and com
p
lete
the che
c
k by data analysi
s
and results measu
r
em
e
n
t.
Che
c
k shall
unde
rgo
sufficient pl
anni
n
g
even
at th
e
begin
n
ing
of the imple
m
e
n
tation, so
a
s
to
make g
ood e
v
aluation for the re
sults. Int
e
rnal a
udit is
a major
che
c
k.
3.4. Action
The key lie
s in that the measu
r
e
s
sh
all be take
n after che
c
king th
e results, i.e., to sum
up succe
s
sful
experie
nce a
nd lea
r
n from
failure
s,
to i
m
pleme
n
t sta
ndardization
as b
a
si
s fo
r the
future. Actio
n
is the
su
b
limation p
r
o
c
ess of
the PDCA cycle.
Without
a
c
ti
on, there
is no
improvem
ent.
As the
ba
sic method
of q
uality mana
g
e
ment, PDCA cycle
is no
t only suitabl
e for the
whol
e soft
wa
re en
gine
erin
g, but also fo
r t
he
whole software ente
r
prises and e
a
ch depa
rtm
ent
and even in
dividual in the softwa
r
e e
n
terp
rises.
Each d
epa
rtm
ent has its o
w
n PDCA cycle
according to the policy aim
of the software enterp
r
is
e, cyclin
g layer upon laye
r, in the form of big
ring li
nki
ng
with small
ri
ng
, and
sm
all ri
ng lin
kin
g
with smalle
r ri
n
g
. Big ri
ng i
s
the m
a
trix a
nd
basi
s
of the
small ri
ng, while small
rin
g
is t
he de
co
mpositio
n an
d gua
rantee
of big ring. P
DCA
cycle is like climbing
stairs, when finishi
ng one
cycle,
the quality of t
he producti
on
will improve
one
step, the
n
setting up
the next cy
cl
e, re
runni
ng
and
reimp
r
ov
ing, thu
s
goi
ng forwa
r
d
a
nd
improvin
g co
ntinuou
sly, a
s
sho
w
n in
Figure 5.
Co
ntinuou
s
stu
d
y is the b
a
s
is
of co
ntin
uous
improvem
ent
[9].
Figure 5. PDCA Cycle A
s
cendin
g
Diag
ram
1) Ori
g
inal level, 2) New level
As the
ba
sic method
of q
uality mana
g
e
ment, PDCA cycle
is no
t only suitabl
e for the
whol
e soft
wa
re en
gine
erin
g, but also fo
r t
he
whole software ente
r
prises and e
a
ch depa
rtm
ent
and even in
dividual in the softwa
r
e e
n
terp
rises.
Each d
epa
rtm
ent has its o
w
n PDCA cycle
according to the policy aim
of the software enterp
r
is
e, cyclin
g layer upon laye
r, in the form of big
ring li
nki
ng
with small
ri
ng
, and
sm
all ri
ng lin
kin
g
with smalle
r ri
n
g
. Big ri
ng i
s
the m
a
trix a
nd
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Study on Software Qualit
y Im
provem
ent based on
Ra
yleig
h
Model
and PDCA…
(Nin
g Jin
g
fen
g
)
4613
basi
s
of the
small ri
ng, while small
rin
g
is t
he de
co
mpositio
n an
d gua
rantee
of big ring. P
DCA
cycle is like climbing
stairs, when finishi
ng one
cycle,
the quality of t
he producti
on
will improve
one
step, the
n
setting up
the next cy
cl
e, re
runni
ng
and
reimp
r
ov
ing, thu
s
goi
ng forwa
r
d
a
nd
improvin
g co
ntinuou
sly, a
s
sho
w
n in
Figure 5.
Co
ntinuou
s
stu
d
y is the b
a
s
is
of co
ntin
uous
improvem
ent
[9].
4.
Defe
cts Re
mo
v
a
l Effectiv
eness Inde
x
Operation def
inition of defe
c
ts removal e
ffectiveness. Definition n
e
e
d
s all defe
c
t data
(incl
udin
g
field defect
)
in the asp
e
ct
s of def
ect source
and in whi
c
h
pha
se to find and re
move
defect.
Make
j=1,2,…k
,
record t
he ph
ases of
the softwa
r
e
life cycl
e. M
a
ke
i
=
1,2,
…k
, record
review or te
st
types of diffe
rent
softwa
r
e
life
cycle
ph
ase
s
in
clu
d
in
g mainte
nan
ce pha
se
(p
ha
se
k
). Th
en the f
o
llowin
g
matri
x
Figure i
s
d
e
f
ect so
ur
c
e
/finding
place
matrix. In this matrix, only Unit
N
ij
(w
here
i
≧
j
, i.e., the unit
s
in
the
lo
wer left trian
g
le
)
has data.
Th
e data
in
the
units
above
t
he
diago
nal indi
cate the n
u
m
ber of defe
c
t
s
inje
cted
a
n
d
detecte
d in
the same
ph
ase; the d
a
ta
in
the units u
n
der the di
ag
onal indi
cate
the numbe
r
of defects i
n
jecte
d
in th
e early ph
ase of
developm
ent but detecte
d in the later ph
ase. Fo
r in
the early pha
se
, it is impossi
ble to detect th
e
defect
s
inject
ed in the later phase, the u
n
its
above th
e diagon
al are empty. The boun
dary ro
w o
f
the matrix (
N
i
.
) indi
cate
the defe
c
ts
removed in t
h
is p
h
a
s
e, a
nd the b
oun
dary
colum
n
(
N.
j
)
indicate the n
u
mbe
r
of defects u
s
in
g this
pha
se a
s
th
e sou
r
ce, as
sho
w
n in Ta
b
l
e 1.
Table 1.
Defe
cts Sou
r
ce/Fi
nding Pla
c
e
Matrix Table
Phase
defe
c
t
s
removal
effectivene
ss (P
DREi
)
can
b
e
ph
ase
in
sp
ection
effecti
v
eness [IE(i)]
or
phas
e
tes
t
effec
t
iveness
[TE(i)]:
PDRE
i
=
1
1
1
.
.
.
i
m
m
i
m
m
i
N
N
N
Phase
Defe
cts Co
ntainme
n
t Effectiveness
(
PDCE
i
):
PDCE
i
=
i
ii
N
N
.
Overall Te
st Effectiveness (
TE
):
TE
=
k
I
i
k
I
i
i
N
N
1
.
i
1
1
.
in the formula
I+
1,I+
2,…k
-1
-
--te
s
t pha
se.
Overall defe
c
ts removal eff
e
ctivene
ss in the developm
ent pro
c
e
ss
(
DRE
):
DRE
=
N
N
k
i
i
1
1
.
5.
Application
of PDCA an
d
Ra
y
l
eigh
in
Solv
ing Softw
a
r
e Qu
alit
y
Manageme
n
t
The ta
rget
of software q
u
a
lity manag
e
m
ent i
s
to
re
duce the
hei
ght of the
cu
rves a
s
muc
h
as
poss
ible, meanwhile move
the c
r
es
t of the Rayleigh c
u
rv
es
to the left,
whic
h trans
f
orm
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 8, August 2013: 4609 –
4615
4614
the p
r
oble
m
i
n
to two
ind
e
xes, i.e., redu
cing the
in
je
cti
on n
u
mbe
r
of defe
c
ts at e
a
ch
ph
ase, a
n
d
increa
sing th
e defe
c
ts ph
a
s
e removal ef
fectivene
ss.
We u
s
e P
D
CA model to so
lve the pro
b
le
ms
of redu
cing th
ese two index
es.
We u
s
e the j
u
st finished P
r
oje
c
t A as hi
stor
y data, th
e scale of Project A is 2
0
000 line
s
of so
urce
co
de; mea
n
whil
e in th
e ne
w starte
d Proj
ect B
(the
scale i
s
21
000
line
s
of
sou
r
ce
cod
e
), we use Rayleig
h
m
odel an
d PDCA model to
make
proj
ect managem
e
n
t. Becau
s
e t
he
scale
s
of Project A and Pro
j
ect B are ba
sically
the sa
me, the work load estim
a
te
d by COCO
M
O
medium
mod
e
l is
ba
sicall
y the sam
e
,
we first ma
ke the follo
win
g
su
mma
ry o
n
the d
e
fect
s of
Proje
c
t A, as sho
w
n in Ta
b
l
e 2:
Table 2. Proj
ect A Defe
cts Source and
Finding Pla
c
e
Example Dat
a
Sheet
In acco
rd
an
ce with
the
T
able,
we fig
u
r
e
out the
d
e
fects in
spe
c
tion effective
ness
at
different pha
ses: IE(I0)=11
6
/(182
+3
52
)=21.
7%;IE(I1)=28
9
/(18
2+3
52-1
1
6
+
46
4)=32.8%;
IE(I2)=598/(1
82+352
+4
64-116-
289
+7
27
)= 45.3%
;TE(UT
)=377/
(182
+3
52
+46
4
+7
27-
116-289
-59
8
+
37
)=49.7%;
TE(CT)
=256/(1
82
+35
2
+4
64
+72
7
+37-1
1
6
-
289
-5
98-3
7
7
+
31
)=62%;
TE(ST)=14
4
/174
=83%
;
th
en o
n
the
ba
sis of the
dat
a, wh
en d
e
velopin
g
Proj
e
c
t B, we
u
s
e
the
origin
al d
e
velopment te
am
; before
devel
oping P
r
oje
c
t
B, we m
a
ke
relevant
sum
m
ary o
n
Proj
ect
A usin
g PDCA model, me
anwhile a
ppl
y the sum
m
a
r
ize
d
exp
e
rie
n
ce i
n
to Proj
ect B, and
u
s
e
PDCA mod
e
l
and Raylei
gh model at
the beginni
ng pha
se of
Project to make p
r
oj
ect
manag
eme
n
t to redu
ce nu
mber of defe
c
ts.
From the
abo
ve Table, we can
see th
ere
are
ma
ny de
fects inj
e
cte
d
in different p
hases,
mean
while
th
e defe
c
ts in
spectio
n
effe
ctiveness i
s
relatively low
in the
pha
se
s of
high
lev
e
l
desi
gn, low l
e
vel desig
n, and coding,
etc. Th
ro
ugh
reason findi
ng, we find the rea
s
o
n
s f
o
r
defect
s
in the high level
desig
n are
the follo
win
g
defect
s
inj
e
ction rea
s
o
n
s: usi
ng wrong
para
m
eters,
invalid or in
corre
c
t scre
en flow, mi
ssing o
r
in
co
rre
ct of high
level flow of
comp
one
nts
passe
d in the
review p
a
ckage, no inp
u
t of the modul
e interfa
c
e
s
, inco
rrect u
s
e
of
the publi
c
d
a
ta stru
cture
,
unreali
z
ed
low level desi
gn for t
he co
de, in
corre
c
t varia
b
le
initialization,
etc. We
can
also
see f
r
om the
abov
e Figu
re, m
o
re d
e
fect
s a
r
e intro
d
u
c
ed
in
pha
se
s such
as u
n
it test,
comp
one
nts t
e
st an
d
syste
m
test, whi
c
h
sho
w
s that t
he devel
ope
rs’
quality of modifying defect
s
have proble
m
s.
Aiming at th
e
above
facto
r
s fo
r d
e
fect
s i
n
ject
io
n,
the proje
c
t revie
w
ers
ma
ke summary,
and m
a
ke pl
an to p
r
eve
n
t the ab
ove p
r
oblem
s. As
k
the expe
rts
e
x
perien
c
e
d
in
de
sign to
tra
i
n
the revie
w
e
r
s, be st
rict in
re
view proce
ss;
set u
p
d
e
fects data
b
a
se,
analyze
the rea
s
on
s for
these
defe
c
ts, summ
ari
z
e
experie
nce, set up
ch
e
ck li
st, ch
eck the
desi
gn
co
nte
n
ts o
ne
by o
ne
in the form
of che
c
k list
and cro
s
s review m
e
th
od to prevent the simila
r mista
k
e
s
from
happ
ening
a
gain; b
e
fore
modifying
defect
s
, th
e
develo
pers first
rea
d
defect
s
d
a
ta
base,
mean
while im
prove the d
e
velope
rs’
self test wo
rk
after modifying de
fects, to prev
ent the defe
c
ts
that have o
c
curred fro
m
ha
ppeni
ng ag
ai
n. As pe
r th
e
plan, after
we ca
rry o
u
t Proje
c
t B, we
get
data as
sho
w
n in Table 3.
From th
e abo
ve Figure, we
can
se
e the
overa
ll n
u
mb
er of def
ect
s
redu
ce
d by n
early a
third, in pa
rti
c
ula
r
, the n
u
m
ber
of def
ects i
n
hig
h
level de
sign
and
codi
ng
pha
se
s red
u
ce
obviou
s
ly. In Proje
c
t B, the peak val
ue o
f
defects findi
ng is in th
e in
spe
c
tion p
h
a
s
e of lo
w lev
e
l
desi
gn, due t
o
the obviou
s
increa
se of d
e
fects
re
m
o
val rate, the p
eak valu
e al
so cha
nge
s, when
the pea
k valu
e red
u
ces, it cha
nge
s from
codin
g
ph
ase to the low l
e
vel desi
gn.
The field d
e
fe
cts
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Study on Software Qualit
y Im
provem
ent based on
Ra
yleig
h
Model
and PDCA…
(Nin
g Jin
g
fen
g
)
4615
numbe
r is al
so le
ss tha
n
that of Project A. So
it achieve
s
o
u
r targ
ets in
software qu
ality
manag
eme
n
t.
Table 3.
Proj
ect B Defe
cts Source and
Finding Pla
c
e
Example Dat
a
Sheet
IE(I0)=28
%
;;
;
;
;
;
IE(I
1)=51.2%
IE
(I2)
=46.6%
TE
(UT
)
=52.8%
TE
(CT
)
=57%
TE(S
T)=
81%
6.
Conclusion
As a
reli
abili
ty model, Ra
yleigh mo
del
applie
s to
the d
e
fect
s a
nalysi
s
in th
e whole
softwa
r
e
dev
elopme
n
t cy
cle, and
its
re
mained
def
e
c
ts qu
antitative evalu
a
tion
mainly relies
on
the corre
c
tne
ss of the ea
rl
y phase d
a
ta; in qualitat
ive analysi
s
of cross-ph
ase testing a
c
tivities,
singl
e-p
h
a
s
e
testing evalu
a
tion ca
nnot
be mad
e
. Its significan
c
e
lies in e
m
ph
asi
z
ing the t
w
o
prin
ciple
s
of defect
s
prevention an
d e
a
rly defect
s
removal. Th
e
y
are the m
a
in dire
ction
s
o
f
improvem
ent
strategy
of d
e
v
elopment
qu
ality.
PDCA
cycle-b
ased
software
q
uality mana
geme
n
t
pro
c
e
s
s control an
d imp
r
o
v
ement mo
d
e
l u
s
e
pro
c
e
s
s-o
r
iente
d
p
r
oje
c
t pla
n
m
e
thod, tra
n
sf
orm
the stand
ard
pro
c
e
ss of software o
r
ga
nizatio
n
in
to the tasks of relev
ant perso
nnel of software
proje
c
t, which
effectively en
sure the
execution
of the
q
uality manag
ement p
r
o
c
e
s
s. The
analy
s
is
of the measu
r
eme
n
t data enabl
es the
obje
c
tive
decision on h
o
w to control a
nd improve t
he
quality man
a
gement
pro
c
ess. Combin
e Raylei
gh m
odel
with PDCA mod
e
l in
software
qu
ality
manag
eme
n
t, and
use PDCA cy
cle in
bi-di
r
e
c
tional
quality impro
v
ement st
rat
egy of Raylei
gh
model, thus a
c
hievin
g our
quality mana
gement go
al better.
Ackn
o
w
l
e
dg
ements
Than
ks for th
e supp
ort of
the Provinci
al
Natural
Scientific Fu
nd
Proje
c
t: The
study of
the Key Tech
nology of the
Web S
o
ftwa
r
e
Develop
m
ent Platform
Oriente
d
to t
he Servi
c
e
(
No.
2010
1525
) a
nd The
Proje
c
t of Ji Li
n
Provinci
al
De
partme
n
t of Educatio
n: T
he Developm
ent
Platform based on OSGi (No. 201
085
).
Referen
ces
[1]
Ying
lan F
a
ng,
Bing H
an, W
e
i
f
u Z
hou. Res
e
arch
a
nd An
al
ysis
of CMMI Process Impro
v
ement Bas
e
d
on SQCS S
y
st
em.
TEL
K
OMNIKA
. 2012; 10(
7).
[2]
SJ Sun, J Xia
o
. A Soft
w
a
re
Reli
ab
ilit
y GE
P Mode
l Base
d on Us
ag
e P
r
ofile.
TEL
K
O
M
NIKA
. 2012;
10(7): 17
56-
17
64.
[3]
Xi
e M. Soft
w
a
r
e
Reli
ab
ilit
y Mo
deli
ng. Sin
g
a
p
o
re
: Word Scie
ntific Publ
ishi
n
g
Co.Pte.Ltd., 199
1.
[4]
L
y
u M. H
a
n
d
b
ook
of Re
lia
bil
i
t
y
E
ngi
ne
erin
g
.
Singa
por
e: McGra
w
-
Hil
l 1
9
9
6
cov
e
rin
g
h
o
w
t
o
e
n
sur
e
soft
w
a
re reliabilit
y
.
RCA En
gi
n
eer
. 198
2; 27(1
)
: 53-57.
[5]
Paulk, Mark C. et al.
T
he Capa
bil
i
t
y
M
a
turit
y
Mo
de
l:
Guidel
ines
f
o
r
Improvin
g
the
Soft
w
a
r
e
Process (
R
ea
d
i
ng: A
ddis
o
n
W
e
sle
y
Lo
ngm
a
n
, Inc., 19
95).
Addis
on-W
e
sl
e
y
P
u
b
lish
i
n
g
C
o
mpa
n
y
,
Inc.
Repri
n
ted b
y
p
e
rmissio
n
of Addis
on-W
e
sl
e
y
Longm
an. 19
9
5
: 81.
[6]
DERN INGW
E. Out of
the Crisis. Cambrid
g
e
:
M IT
Center for Advanc
ed E
ngi
neer
in
g Stud
y
.
1
982.
[7]
W
ang Q, Li M
S
, Liu
X. An ac
tive meas
urem
ent
mod
e
l for s
o
ft
w
a
r
e
pr
oces
s control
and
i
m
provem
ent.
Journ
a
l of Softw
a
r
e. 20
05; 16
(3): 407
−
41
8.
h
ttp://
w
w
w
.
jos.o
r
g.cn/100
0-9
8
2
5
/16/40
7.htm
(i
n Chi
nese)
[8]
Hua
ng F
e
i
x
ue,
Li Z
h
i
hao, S
u
n
Xia
o
li. Stu
d
y
on Qua
lit
y G
uara
n
tee M
o
d
e
l Bas
ed
on
PDCA Ind
i
an
soft
w
a
re. J
o
u
r
nal
of H
a
rbi
n
I
n
stitute of T
e
c
hno
logy
.
20
05;
11:
15
83-1
5
8
5
. Gingr
ich, N
e
w
t
R
ene
w
i
n
g
America
n
Civ
i
lizatio
n, Pil
l
ar
Five: Qualit
y
as
defi
n
e
d
b
y
Dem
i
ng.
In: Qualit
y Progress
(Decemb
e
r,19
95).
Evaluation Warning : The document was created with Spire.PDF for Python.