Вы находитесь на странице: 1из 34

Unit 6:

Design of Expert
Systems
Knowledge Based Expert System
Figure 6.1 Project Management Tasks
Selecting the Appropriate
Problem

We need to answer the questions, Why are we


buiding this expert system!"#
$
%ntee&tua 'roperty (greement must be &onsidered
$
)eary identify the probem
$
)eary identify the expert
$
)eary identify the users

We need to *now the payoff $ money, effi&ien&y,


et&#
Selecting the Appropriate
Problem

What toos wi be a+aiabe to buid the expert system!


$
,3
rd
law o !isdom" if you-re ost, as* someone#. )he&* the Web
for appi&ations in existen&e
$
,#
nd
law o !isdom" if you-re going to as* someone for hep, it-s
good idea to *now the anguage#. /now the anguage
ne&essary to &reate a semanti& net of reationships on whi&h the
system wi be based
Selecting the Appropriate
Problem

0ow mu&h wi the expert system &ost!


$
1
st
law o !isdom" %f you-+e bought a &ar and are not worried
about the &ost of gas, it-s is a good idea to own your own oi we#
$
$ero law o !isdom" %t-s better to use someone ese-s &ar and gas#
$
( fun&tion of peope, resour&es, time, et&#
$
0ow a+aiabe is the *nowedge!
Stages in the %e&elopment

0ow wi the system be de+eoped!


$
1his wi depend on the resour&es pro+ided
Stages:
2# 3easibiity Study $ see if the pro4e&t is feasibe
5# 6apid 'rototype $ demonstrate ideas 7 impress
8# 6efined System $ +erifi&ation by *nowedge engineers
9# 3ied 1estabe $ system tested by see&ted users
:# )ommer&ia quaity system $ +aidation 7 testing
6# ;aintenan&e and e+oution $ repair bugs, enhan&e
&apabiities
'ther (onsiderations

0ow wi the system be dei+ered!


$
Shoud be &onsidered in eariest stages of de+eopment
$
%ntegration with existing programs

0ow wi the system be maintained and e+o+e!


$
,)1 law o !isdom" %f you don-t ha+e a spe&ifi&
destination, go e+erywhere be&ause you ne+er *now
what you are missing#.
$
'erforman&e is dependent on *nowedge 7 expertise
$
'erforman&e must be maintained
$
<ew *nowedge wi be a&quired
$
=d *nowedge wi be modified
Figure 6.# *eneral Stages in the
%e&elopment o an +,pert S-stem
+rrors in %e&elopment Stages

Expert-s *nowedge may be erroneous,


propagating errors throughout the entire
de+eopment pro&ess#

> ,)# law o !isdom" %f you are going to set out on


a 4ourney, %t-s good if you o+e to 4ourney#.
$
3orma pro&edures may be ne&essary to &ertify expert
$
1e&hnique panes &an s&rutini?e expert-s *nowedge
$
3o&us groups &an aso be used ,attempting to find out
what wi reay se#.
+rrors in the %e&elopment Stages

/nowedge may not be propery &ommuni&ated to


*nowedge engineer, or *nowedge may be misinterpreted#

/nowedge base may be &orrupted by entering in&orre&t


form of a rue or fa&t#

%nferen&e engine errors may resut from errors in pattern


mat&hing, &onfi&t resoution, and exe&ution of a&tions#

%nferen&e &hain errors may be &aused by erroneous


*nowedge, semanti& errors, inferen&e engine bugs,
in&orre&t spe&ifi&ations of rue priorities, and strange
intera&tion among rues#

@imits of ignoran&e $ a probem &ommon to a pre+ious


stages
Figure 6.3 Major +rrors in +,pert
S-stems and Some (auses
Sotware +ngineering
and +,pert S-stems

Expert systems are produ&ts i*e any other


software produ&t and require good standards for
de+eopment#

Expert systems may ha+e serious responsibiities


$ ife and death#

0igh standards are a ne&essity and &an be


measured by mean time between faiures"#
Figure 6.. Sotware +ngineering
Methodolog-
+,pert S-stem /ie (-cle

Aegins with the initia &on&ept of the software and


ends with its retirement from use#

,)3 law o !isdom" %f you don-t o+e to 4ourney,


then o+e what you ha+e got.

Expert systems require more maintenan&e be&ause


they are based on *nowedge that is:
$
0euristi&
$
Experientia

( number of ife &y&e modes ha+e been


de+eoped#
!aterall Model

Ea&h stage ends with a +erifi&ation and


+aidation a&ti+ity to minimi?e any probems in
that stage#

(rrows go ba&* and forth ony one stage at a


time#

%t is assumed that a information ne&essary for


a stage is *nown#
Figure 6.0 !aterall Model o the
Sotware /ie (-cle
(ode)and)Fi, Model

Some &ode is written and then fixed when it does


not wor* &orre&ty#

Usuay the method of &hoi&e for new


programming students in &on+entiona and expert
systems

1his e+entuay ed to the do>it>twi&e &on&ept


where a prototype then a fina system was buit#
1ncremental Model

1his is a refinement of the waterfa and top>


down>approa&h#

1he idea is to de+eop software in in&rements of


fun&tiona &apabiity#
$
;a4or in&rement $ assistant &oeague expert
$
;inor in&rement $ expertise within ea&h e+e
$
;i&roin&rement $ add7refining indi+idua rues
Spiral Model
Ea&h &ir&uit of the spira adds some fun&tiona
&apabiity to the system#
%etailed /ie (-cle Model
/inear Model
2# 'anning Stage
1he purpose of this stage is to produ&e a forma wor* pan for
the expert system de+eopment $ do&uments to guide and
e+auate the de+eopment#
/inear Model
5# /nowedge Definition
1he ob4e&ti+e of this stage is to define the
*nowedge requirements of the expert system,
whi&h &onsists of two main tas*s:

/nowedge sour&e identifi&ation and see&tion

/nowedge a&quisition, anaysis, and extra&tion


Table 6.3 2nowledge Source 3
1dentiication
Table 6.. 2nowledge Ac4uisition5
Anal-sis5 and +,traction Tasks
/inear Model
8# /nowedge Design
1he ob4e&ti+e is to produ&e the detaied design
for an expert system and in+o+es:

/nowedge definition

Detaied design
Table 6.0 2nowledge %einition
Tasks
Table 6.6 %etailed %esign o
2nowledge Tasks
/inear Model
9# )ode and )he&*out
1his begins the a&tua &ode impementation
/inear Model
:# /nowedge Berifi&ation
1he ob4e&ti+e here is to determine the
&orre&tness, &ompeteness, and &onsisten&y of
the system#

3orma tests

1est (naysis
1abe 6#C 3orma 1est 1as*s of /nowedge Berifi&ation Stage
1est (naysis 1as*s
/inear Model
6# System E+auation
1his stage is for summari?ing what has been
earned with re&ommendations for impro+ements
and &orre&tions#
Figure 6.6 /inear Model o +,pert
S-stem %e&elopment /ie (-cle
Summar-

%n this &hapter, we ha+e dis&ussed a software


engineering approa&h to the &onstru&tion of an
expert system#

'rin&ipes about good inter+iewing te&hniques


were mentioned#

<ow that expert systems are widey used, they


must be quaity produ&ts due to the sensiti+e
nature of their appi&ations#
Summar-

3a&tors to be &onsidered in the design of expert


systems in&ude probem see&tion, &ost, and
payoff#

Aoth manageria and te&hni&a aspe&ts must be


&onsidered#

Ay &onsistenty foowing the outined ife


&y&e" it shoud be possibe to &onstru&t quaity
software#

Вам также может понравиться