1
Chap e
Inc easing he E iciency o
Rule-Based Expe Sys ems Applied
on He e ogeneous Da a Sou ces
Juan IgnacioGue e o Alonso, En iquePe sonal,
An onioPa ejo, S.Ga cía, An onioMa ín and Ca losLeón
Abs ac
Nowadays, he p oli e a ion o he e ogeneous da a sou ces p o ided by di e en
esea ch and inno a ion p ojec s and ini ia i es is p oli e a ing mo e and mo e and
p esen s huge oppo uni ies. These de elopmen s c ea e an inc ease in he numbe
o di e en da a sou ces, which could be in ol ed in he p ocess o decision-
making o a speci ic pu pose, bu his huge he e ogenei y makes his ask di -
icul . T adi ionally, he expe sys ems y o in eg a e all in o ma ion in o a main
da abase, bu , some imes, his in o ma ion is no easily a ailable, o i s in eg a ion
wi h o he da abases is e y p oblema ic. In his case, i is essen ial o es ablish
p ocedu es ha make a me ada a dis ibu ed in eg a ion o hem. This p ocess
p o ides a “mapping” o a ailable in o ma ion, bu i is only a logic le el. Thus, on
a physical le el, he da a is s ill dis ibu ed in o se e al esou ces. In his sense, his
chap e p oposes a dis ibu ed ule engine ex ension (DREE) based on edge com-
pu ing ha makes an in eg a ion o me ada a p o ided by di e en he e ogeneous
da a sou ces, applying hen a ma hema ical decomposi ion o e he an eceden o
ules. The use o he p oposed ule engine inc eases he e iciency and he capabili y
o ule-based expe sys ems, p o iding he possibili y o applying hese ules o e
dis ibu ed and he e ogeneous da a sou ces, inc easing he size o da a se s ha
could be in ol ed in he decision-making p ocess.
Keywo ds: ule-based expe sys em, in e ence engine, he e ogeneous da a sou ce
in eg a ion, dis ibu ed da a sou ces
1. In oduc ion
The expe sys ems (ESs) a e one o he mos adi ional a i icial in elligence
echniques [1], p o iding he possibili y o implemen ing sys ems which allow us
o sol e p oblems in a limi ed domain. Howe e , he a ie y and possibili ies o ESs
ha e imp o ed in he las yea s wi h he combina ion o hem wi h o he echnolo-
gies such as uzzy logic [2], Bayesian ne wo k [3], e c. Mo eo e , se e al languages
and ools ha e been de eloped o p o ide as e de elopmen s (e.g., CLIPS, LISP,
PROLOG, e c.) and deploymen s. O he sys ems a e imp o ed in collabo a ion
wi h o he echnologies in o de o expand he domain o p oblems and inc ease
he knowledge base. As example, [4] imp o es he capabili ies o an ES adding ex
mining, neu al ne wo ks, and s a is ical echniques.
Applica ion o Expe Sys ems - Theo e ical and P ac ical Aspec s
2
On he one hand, he scope o ES applica ion is e y ex ensi e, including heal h
[5], educa ion [6], physics [7], chemical [8], mechanics [9], e c. Thus, he appli-
ca ion o ES is no only es ic ed o sol e he p oblem, hey usually include an
explaining engine, which could be used o educa ional pu poses, ope a ing a he
same ime.
The ESs ha e a limi ed domain, and he size o used da a se s is smalle han in
o he a i icial in elligence echniques. Howe e , he imp o emen o a ailabili y
o in o ma ion, new concep s ela ed o senso ne wo ks, and he capabili y o
gene a e and consume in o ma ion in di e en sec o s p o ide a di e en scena io,
in which he ESs adi ionally had a lo o in o ma ion dissemina ed in o di e en
in o ma ion esou ces. Al hough each in o ma ion esou ces could ha e i s own
s uc u e, he s o ed in o ma ion could be e y simila . In his scena io, i is essen-
ial o make he analysis o he dis ibu ed da a se s and apply he di e en ules
and in e ence engines in hese dis ibu ed da a se s possible.
In esponse o his p oblem, an addi ional laye is p oposed in he cu en
chap e which could be added o he ES engines, mainly based on ules and uzzy
logic. This new laye o middlewa e allows he ES o make a me ada a in eg a ion
o he e ogeneous da a se s, wi hou making a eal in eg a ion and eplica ion o
he in o ma ion, allowing by mean edge, and compu ing he applica ion o ules in
he dissemina ed da a se s by means o a logical decomposi ion o ules based on
me ada a in eg a ion esul s. This no el laye o middlewa e, named dis ibu ed
ule engine ex ension (DREE) has an applica ion p o ocol in e ace (API) o allow
he communica ion be ween he ES engine and he di e en da a sou ces. The
DREE enables he ES engines o be applied in he e ogeneous da a se s dissemina ed
in a ne wo k, by means o ins alla ion o edge compu ing daemon (ECD) in each
dis ibu ed node.
Edge compu ing has some simila i ies wi h og compu ing, cloud compu ing,
e c. [10] p o ides a e iew o di e en simila echnologies, p o iding a e y com-
ple e su ey. In he case o he p oposed sys em, edge compu ing is usually ela ed
o he In e ne o Things (IoT) de ices. In case o he p oposed solu ion, he edge
nodes a e he compu e s o de ices, which has he in o ma ion s o ed, by means o
any ype o da abase managemen sys em.
In he ollowing sec ions, he gene al a chi ec u e is de ailed. Fi s ly, he p ocess o
me ada a in eg a ion om he e ogeneous da a se s is desc ibed. Secondly, he p ocess
o logical decomposi ion and how he edge compu ing con ibu es in he p ocess o
dis ibu ed applica ion o an eceden ule a e desc ibed. Finally, he expe imen al
applica ion o he p oposed laye o middlewa e is shown, wi h he conclusion o
esul s and u u e esea ch lines.
2. A chi ec u e o e iew
The applica ion p ocess o he DREE is pe o med in h ee s ages. In he i s
one, he me ada a in eg a ion is pe o med. In he second, he ules om knowl-
edge base a e logically decomposed, acco ding o he me ada a model. In he hi d
one, he ules a e ca ied ou in he dissemina ed da a se s by means o edge com-
pu ing in he dis ibu ed nodes. In his sense, o ca y ou hese asks, he a chi ec-
u e p oposed o he DREE is shown in Figu e 1.
Each node has an edge compu ing daemon which pe o ms que ies in he
da abase node o ga he all in o ma ion om local da a se s, sending only me ada a
o he DREE.This in o ma ion is in eg a ed and ansla ed in a uni ied me ada a
s uc u e o an engine di ec ly a ailable o DREE.
3
Inc easing he E iciency o Rule-Based Expe Sys ems Applied on He e ogeneous Da a Sou ces
DOI: h p://dx.doi.o g/10.5772/in echopen.90743
The me ada a in eg a ion model da abase p o ides a map abou he dis ibu-
ion o da a be ween he di e en nodes. This in o ma ion is used by logical
decomposi ion engine (LDE) o iden i y he dependences be ween a iables and
he complexi y o ules in he knowledge base. In some cases in which he e exis s a
e y complex logic exp ession wi h a high le el o dissemina ion be ween nodes, he
in o ma ion would be s o ed in his da abase in o de o make he ule applica ion,
bu his op ion needs o be con igu ed manually.
3. He e ogeneous da a se in eg a ion (HDSI)
The undamen als o he e ogeneous da a se in eg a ion we e desc ibed in [11]. In
his e e ence, a he e ogeneous da a sou ce in eg a ion sys em (HDSIS) is desc ibed
and applied o sma g id and heal h. Following his idea, a HDSIS e olu ion is imple-
men ed p o iding also a logic in eg a ion o all in o ma ion a me ada a le el. The
a chi ec u e o he p oposed HDSI is shown in Figu e 2. Speci ically, he me ada a
om di e en sou ces is he only in o ma ion ha is in eg a ed and s o ed.
The HDSI includes a me ada a mining engine (MME), which connec s wi h
EDCs by means o cha ac e iza ion engine, in o de o ex ac in o ma ion om
local da abases in each node, ge ing new and in eg a ing exis ing use ul me ada a.
The dynamic ex ac , ans o m, and load (ETL) engine pe o ms he p ocess o
in eg a e he me ada a, p e iously in e ed by he MME, acco ding o he speci ica-
ions ga he ed by me ada a mining engine and he ules s o ed in he knowledge
base. All hese modules de ine he que y engine and he ule-based expe sys em.
The que y engine is ocused on pe o ming he di e en asks equi ed o he que-
ies in he dis ibu ed esou ces. The ule-based expe sys em (RBES) is included
in he HDSI and implemen s he ules ha pe o m an in eg a ion o all me ada a
om all dissemina ed esou ces. The e o e, i is an RBES o ien ed o in o ma ion
in eg a ion. Some e e ences show he applica ion o HDSIS in di e en p oblems:
non echnical losses de ec ion [12], elec ic ehicle and consump ion modeling in
sma g ids [13], e c.
Figu e 1.
DREE a chi ec u e and low o e iew.
Applica ion o Expe Sys ems - Theo e ical and P ac ical Aspec s
4
The ECD includes some modules o pe o m he necessa y que ies o e he
da abase. Thus, he a chi ec u e o ECD is desc ibed in he nex sec ion.
4. Logical decomposi ion engine
The logical decomposi ion engine included wo pa s. The main pa is in he
LDE and is shown in Figu e 3. The second pa is loca ed in he ECDs, i s s uc u e
being shown in Figu e 4.
The LDE has a i s componen , named logic pa se . This componen makes i
possible o pa se logic exp ession, which could be based on uzzy logic o ype-2 uzzy
logic, oo. The logic pa se wo ks wi h di e en ma hema ical ep esen a ion s anda d
languages: Ma hML (ma hema ical ma kup language 3.0 [14]) and he OpenMa h
s anda d [15]. The logic pa se se es as a REST ul web se ice in e ace, which
suppo s ex ensible ma kup language (XML) o ma messages based on Ma hML and
Ja aSc ip Objec No a ion (JSON) o ma messages based on OpenMa h.
The LDE has egis e ed some logic exp essions, which a e ypical equa ions and
o he p e iously pe o med exp essions. LDE da abase s o es all in o ma ion abou
he decomposi ion o logic exp ession. In his sense, i he exp ession is no ye in
he da abase, he decomposi ion engine checks he dependency be ween a iables
and he numbe o s eps necessa y o calcula e he esul . To educe he numbe o
s eps, he decomposi ion engine applies a pa icle swa m op imiza ion (PSO) [16]
algo i hm. The objec i e o using a gene ic algo i hm is o ind ou an equi alen
logic exp ession wi h a small numbe o s eps, educing he dependency be ween
a iables and ope a ions.
Thus, i he numbe o s eps is s ill high o he dependence be ween a iables
could in ol e di e en da a se s in dissemina ed nodes, he subschedule plans
he message exchange in o de o es ablish he message sequence. This message
sequence could be p o ided by he esul o pa ial o comple e logic exp ession.
Thus, i is possible ha some logic exp essions ake a long ime o ge he esul ,
because o he high numbe o messages exchanged. In his case, he use could
manually con igu e he HDSI o in eg a e he anonymized in o ma ion in a se e
in o de o educe he ime and consump ion o edge compu ing nodes. This
con igu a ion and speci ica ion abou decomposi ion a e s o ed in exp ession
da abase. When he LDE de ec s an exp ession o his ype, he schedule manages
he decomposi ion acco ding o he manual con igu a ion.
In he ECD, he messages om he LDE a e in e p e ed and ans e ed o
local SQL engine and logic in e ence engine. The local da abase engine ex ac s
he equi ed in o ma ion om local da abase and, his in o ma ion is ga he ed by
Figu e 2.
He e ogeneous da a se in eg a ion module o e iew.
5
Inc easing he E iciency o Rule-Based Expe Sys ems Applied on He e ogeneous Da a Sou ces
DOI: h p://dx.doi.o g/10.5772/in echopen.90743
he logic in e ence engine which pe o ms he exp essions and e u ns he esul s.
Simul aneously, he ex ac ed in o ma ion and esul s a e checked by he anony-
miza ion checke , which is esponsible o es s i he in o ma ion con ains da a,
which would be anonymized.
The local da abase engine is o med by wo mechanisms. One engine is based on
simple s anda d s a emen que y language (SQL). The e a e se e al da abase man-
agemen sys ems ha include imp o emen s in he SQL language. The p oposed
engine only applies que ies on s anda d o ma , using minimized SQL que ies wi h
simple s a emen s o ensu e he compa ibili y. The o he engine is based on NoSQL
language, and i was designed o ope a e wi h uns uc u ed da abases.
Addi ionally, he local da abase engine no i ies any modi ica ion in he da a
se o he LDE, in o de o upda e he me ada a model. In his case, he LDE
Figu e 4.
Func ional a chi ec u e o he edge compu ing daemon.
Figu e 3.
The main pa o LDE.

Applica ion o Expe Sys ems - Theo e ical and P ac ical Aspec s
6
ecalcula es he decomposi ion o di e en exp essions ha in ol es he da a om
he upda ed o modi ied da a se .
The logic in e ence engine suppo s logic, uzzy logic, and ype-2 uzzy logic
exp essions. This engine does no pe o m any decomposi ions bu only pe o ms
he exp essions p o ided by LDE.
4.1 Pa icle swa m op imiza ion
The p io i iza ion algo i hm wo ks as a swa m in elligence algo i hm. The
applica ion o he algo i hm is pe o med a e a p ep ocessing o in o ma ion.
The p io i iza ion algo i hm is based on he pa ame ic op imiza ion un il a
solu ion is ob ained. This op imiza ion is execu ed depending on he capabili ies
o a sys em. The canonical PSO model consis s o a swa m o pa icles, which a e
ini ialized wi h a popula ion o andom candida e solu ions. The candida e solu-
ions a e gene a ed by he applica ion o di e en p ope ies o ien ed o educe he
dependence and ope a ions in ol ed in he logic exp ession o lead o he exp ession,
which minimizes he numbe o messages and dis ibu ed exp ession. They i e a i ely
mo e h ough he d-dimension p oblem space o sea ch o he new solu ions, whe e
i ness can be calcula ed as he ce ain quali y measu e. Each pa icle has a posi ion
ha is ep esen ed by he posi ion- ec o xid (i is he index o he pa icle, and d is
he dimension) and a eloci y ep esen ed by he eloci y- ec o id. Each pa icle
emembe s i s bes posi ion in he ec o xi#, and i s j- h dimensional alue is x#ij. The
bes posi ion- ec o among he swa m is s o ed in he ec o x*, and i s j- h dimen-
sional alue is x*j. A he i e a ion ime , he upda e o he eloci y om he p e ious
eloci y o he new eloci y is de e mined by Eq. (1). The new posi ion is de e mined
by he sum o he p e ious posi ion, and he new eloci y is de e mined by Eq. (2):
id
(
+ 1
)
= w ∙ id
(
)
+ c 1 ∙ ψ 1 ∙
(
p id
(
)
− x i
(
)
)
+ c 2 ∙ ψ 2 ∙
(
p g
(
)
− x id
(
)
)
(1)
x id
(
+ 1
)
= x id
(
)
+ id
(
+ 1
)
(2)
whe e c1 and c2 a e cons an weigh ac o s, pi is he bes posi ion achie ed by
pa icle i, pg is he bes posi ion ob ained by he neighbo s o pa icle i, ψ1 and ψ2 a e
andom ac o s in he [0,1] in e al, and ω is he ine ia weigh . Some e e ences
deno e c1 and c2 as he sel - ecogni ion componen and he coe icien o he social
componen , espec i ely.
Di e en cons ain s can be applied o ensu e he con e gence o he algo i hm.
In his case, he ope a ions a e o ien ed o op imize he i ness, and he i ness is
calcula ed based on he numbe o dis ibu ed ope a ions and numbe o exchanged
messages.
PSO algo i hm:
1. Ini ialize pa icles
2. Repea
a. Calcula e he i ness alues o each pa icle.
b. Is he cu en i ness alue be e han pi?
i. Yes. Assign he cu en i ness as he new pi.
ii. No. Keep he p e ious pi.
7
Inc easing he E iciency o Rule-Based Expe Sys ems Applied on He e ogeneous Da a Sou ces
DOI: h p://dx.doi.o g/10.5772/in echopen.90743
c.Assign he bes pa icle’s pi alue o pg.
d.Calcula e he eloci y o each pa icle.
e.Use each pa icle’s eloci y alue o upda e i s da a alues.
3. Un il s opping c i e ia a e sa is ied
5. Tes ing DREE
The p oposed sys em DREE was es ed wi h di e en RBESs. These es s we e
pe o med inse ing he DREE be ween RBES and da a se s, ede ining he in e -
ence engines o eplace he calls o ule execu ion by he DREE in e ace. [17]
p o ides he desc ip ion o di e en ESs o ien ed o heal h, elecommunica ion,
powe supply, e c. The p oposed a chi ec u e is shown in Figu e 5. The in o ma ion
used in he di e en ESs was manually dissemina ed among i e di e en nodes.
The DREE was able o wo k wi h di e en ESs a he same ime. The applica ion
o all ules om all ESs a he same ime may in ol e in o ma ion om di e en
domain knowledge, because he DREE only applies he ules ha he co esponding
in e ence sys em p o ides. Thus, he s a e and he decision suppo ed is pe o med
inside o he ES.The DREE only e u ns he esul o consequen om he i ed
ule. Addi ionally, he usage o ECD makes independen om he local da a se
managemen sys ems. The e o e, in hese cases, he ESs imp o ed hei ope a ional
capabili ies, p o iding he possibili y o ope a e in eal ime. Addi ionally, he ESs
may simul aneously un hei own ule-based knowledge base, inc easing he eli-
abili y o he di e en sys ems.
The usage o he PSO algo i hm ins ead using he mos simpli ied logic exp es-
sion shows an inc ease o e iciency o he dis ibu ed ope a ions, educing he
ope a ions and message exchanging in 20% ela ed o ini ial uzzy and ype-2 uzzy
logic exp essions and in 5% ela ed o ini ial logic exp essions. This ac is due o
how he da a is dissemina ed by he di e en nodes; di e en dis ibu ions o da a
be ween nodes p o ide di e en op imiza ions. Thus, i he numbe o nodes o
Figu e 5.
P oposed a chi ec u e o applica ion o DREE o he ESs in [17].
Applica ion o Expe Sys ems - Theo e ical and P ac ical Aspec s
8
Au ho de ails
Juan IgnacioGue e o Alonso*, En iquePe sonal, An onioPa ejo, S.Ga cía,
An onioMa ín and Ca losLeón
Depa men o Elec onic Technology, Uni e si y o Se ille, Se ille, Spain
*Add ess all co espondence o: [email protected]
da a se s om he nodes is upda ed o modi ied, i is necessa y o ecalcula e he
decomposi ion o logic exp ession.
In o he cases, like [4, 18], he in eg a ion o he p oposed DREE dec eased he
ha dwa e equi emen s ela ed o he s o ing sys ems.
Al hough he sys em inc eases he message exchange, he sys em a oids in eg a -
ing all in o ma ion in a cen e ed da a base, wi hou using big da a in as uc u e,
aking ad an age om edge compu ing in as uc u e and dis ibu ed nodes.
Addi ionally, he sys em can quickly eac o any upda ing o modi ica ion in any
da a se om he nodes in ol es in he DREE.
6. Conclusions
The p oposed DREE p o ides he oppo uni y o in eg a e a g ea quan i y
o in o ma ion in he in e ence engine, wi hou he equi emen o a big da a
in as uc u e and he ex ac , ans o m, and load o physically in eg a e all he
da a se s. DREE makes a me ada a-le el in eg a ion. A his le el he in eg a ion is
quicke and smalle , and i does no need a g ea quan i y o ha d disk o memo y
space. Al hough he message exchanging inc eases he olume o he exchanged
in o ma ion, he load o edge compu ing nodes is op imized in he LDE, be o e he
applica ion o ules.
The deploymen o DREE is simpli ied by adding a REST ul web se ice
in e ace o access and eplace he adi ional se ices consumed by he in e ence
engine. Addi ionally, he deploymen on he edge nodes is summa ized o ins all a
daemon, named ECD, which simpli ies he access o in o ma ion in he local nodes,
no i ying any modi ica ion o upda ing in he di e en da a se s, by upda ing
he me ada a model in he LDE.Thus, DREE eac s o any change in he da a se s
loca ed on edge nodes.
Finally, he p oposed DREE is independen om he ES, p o iding he pos-
sibili y o un simul aneously se e al ESs. Thus, he p ocessing load balancing is
au oma ically p o ided by he in o ma ion dissemina ion a ound he nodes.
© 2019 The Au ho (s). Licensee In echOpen. This chap e is dis ibu ed unde he e ms
o he C ea i e Commons A ibu ion License (h p://c ea i ecommons.o g/licenses/
by/3.0), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium,
p o ided he o iginal wo k is p ope ly ci ed.
9
Inc easing he E iciency o Rule-Based Expe Sys ems Applied on He e ogeneous Da a Sou ces
DOI: h p://dx.doi.o g/10.5772/in echopen.90743
Re e ences
[1] Jabba HK, Khan RZ.Su ey on
de elopmen o expe sys em in
he a eas o medical, educa ion,
au omobile and ag icul u e. In: 2015
2nd In e na ional Con e ence on
Compu ing o Sus ainable Global
De elopmen (INDIACom); 2015.
pp.776-780
[2] D’Aquila RO, C espo C, Ma e JL,
Pazos J.An in e ence engine based on
uzzy logic o unce ain and imp ecise
expe easoning. Fuzzy Se s and
Sys ems. 2002;129(2):187-202
[3] Chojnacki E, Plumecocq W,
Audouin L. An expe sys em based on a
Bayesian ne wo k o i e sa e y analysis
in nuclea a ea. Fi e Sa e y Jou nal.
2019;105:28-40
[4] Gue e o JI, León C, Monede o I,
Bisca i F, Bisca i J.Imp o ing knowledge-
based sys ems wi h s a is ical
echniques, ex mining, and neu al
ne wo ks o non- echnical loss
de ec ion. Knowledge-Based Sys ems.
2014;71:376-388
[5] Jimenez ML, San ama ía JM,
Ba chino R, Lai a L, Lai a LM,
González LA, e al. Knowledge
ep esen a ion o diagnosis o ca e
p oblems h ough an expe sys em:
Model o he au o-ca e de ici
si ua ions. Expe Sys ems wi h
Applica ions. 2008;34(4):2847-2857
[6] Naga a T, Sasaki H.Pe sonal
compu e based expe sys em o
powe sys em ope a ion educa ion.
In e na ional Jou nal o Elec ical
Powe & Ene gy Sys ems.
1996;18(3):195-201
[7] Végh J.A simple “embedded”
easoning in e ence engine wi h
applica ion example in he X- ay
pho oelec on spec oscopy.
Compu e Physics Communica ions.
2004;160(1):8-22
[8] Qian Y, Li X, Jiang Y, Wen Y.An
expe sys em o eal- ime aul
diagnosis o complex chemical
p ocesses. Expe Sys ems wi h
Applica ions. 2003;24(4):425-432
[9] Magalhães SC, Bo ges RFO,
Calçada LA, Scheid CM, Fols a M,
Waldmann A, e al. De elopmen o
an expe sys em o emo ely build
and con ol d illing luids. Jou nal o
Pe oleum Science and Enginee ing.
2019;181:106033
[10] Youse pou A, Fung C, Nguyen T,
Kadiyala K, Jalali F, Niakanlahiji A,
e al. All one needs o know abou og
compu ing and ela ed edge compu ing
pa adigms: A comple e su ey. Jou nal o
Sys ems A chi ec u e. 2019;98:289-330
[11] Gue e o JI, Ga cía A, Pe sonal E,
Luque J, León C.He e ogeneous da a
sou ce in eg a ion o sma g id
ecosys ems based on me ada a mining.
Expe Sys ems wi h Applica ions.
2017;79:254-268
[12] Gue e o JI, Pe sonal E, Pa ejo A,
Monede o I, Bisca i F, Bisca i J, e al.
High pe o mance da a analysis o
non- echnical losses educ ion. En:
Lou J, edi o . Sma G ids: Eme ging
Technologies, Challenges and Fu u e
Di ec ions. NewYo k, USA: No a
Science Publishe s; 2017. p.1-45.
(Ene gy Science, Enginee ing and
Technology)
[13] Gue e o JI, Ga cía A, Pe sonal E,
Pa ejo A, Pé ez F, León C.A Rule-based
expe sys em o he e ogeneous da a
sou ce in eg a ion in sma g id sys ems.
En: Ryan D, edi o . Expe Sys ems:
Design, Applica ions and Technology.
NewYo k, USA: No a Science
Publishe s; 2017. p.59-104. (Compu e
Science, Technology and Applica ions)
[14] Ma hema ical Ma kup Language
(Ma hML) Ve sion 3.0 2nd Edi ion