chap12. exp syst
Post on 07-Apr-2018
253 Views
Preview:
TRANSCRIPT
-
8/6/2019 Chap12. Exp Syst
1/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
1
Chapter 12: Fundamentals of
Expert Systems
12.1 Opening Vignette:
CATS-1 at General Electric
The Problem
General Electric's (GE)Top Locomotive Field Service
Engineer was Nearing Retirement
-
8/6/2019 Chap12. Exp Syst
2/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
2
Traditional Solution:
Apprenticeship
Good Short-term Solution
BUT GE Wanted A more effective and dependable wayto disseminate expertiseTo prevent valuable knowledge fromretiringTo minimize extensive travel ormoving the locomotives
-
8/6/2019 Chap12. Exp Syst
3/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
3
The Expert System
SolutionTo MODEL the way a humantroubleshooter works
Months of knowledge acquisition 3 years of prototyping
A novice engineer or technician canperform at an experts level
On a personal computer Installed at every railroad repair shop
served by GE
-
8/6/2019 Chap12. Exp Syst
4/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
4
12.2 (ES) Introduction
Expert System : from the termknowl edge-based expert system
An Expert System is a system thatemploys human knowledgecaptured in a computer to solveproblems that ordinarily requirehuman expertiseES i m i tate the experts reasoning
processes to solve specific
-
8/6/2019 Chap12. Exp Syst
5/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
5
12.3 History of
Expert Systems1. Ear l y t o M i d- 1960 s
One attempt: the General-purpose Problem
Solver (GPS)General-purpose Problem Solver (GPS)
A procedure developed by Newell andSimon [1973] from their Logic TheoryMachine -
Attempted to create an "intelligent"computer
Predecessor to ES Not successful, but a good start
-
8/6/2019 Chap12. Exp Syst
6/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
6
2 . M i d- 1960 s: Special-purpose ES programs
DENDRAL MYCIN
Researchers recognized that the problem-solving mechanism is only a small part of acomplete, intelligent computer system
General problem solvers c a nno t be used to buildhigh performance ES
H uman problem solvers are good only if theyoperate in a very n arr ow d oma i n
Expert systems must be constantly u pdated withnew information
The complexity of problems requires aconsiderable amount of knowl edge ab o u t the
pr ob l em area
-
8/6/2019 Chap12. Exp Syst
7/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
7
3. M i d 1970 s Several Real Expert Systems Emerge Recognition of the Central Role of
Knowledge AI Scientists Develop
Comprehensive knowledge representationtheoriesGeneral-purpose, decision-making proceduresand inferences
Limited Success Because Knowledge is Too Broad and Diverse Efforts to Solve Fairly General Knowledge-
Based Problems were Premature
-
8/6/2019 Chap12. Exp Syst
8/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
8
BUTSeveral knowl edge represe n tat i on sworked
Key InsightT he p ow er of a n ES i s der iv ed f r om
the spe c i f i c knowl edge i t p ossesses,no t f r om the part i cu l ar fo rma l i smsa n d i nf ere n c e s c hemes i t emp lo ys
-
8/6/2019 Chap12. Exp Syst
9/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
9
4 . Ear l y 1980 s
ES Technology Starts to go Commercial X CON X SEL
CATS-1Programming Tools and Shells Appear
EMYCIN E X PERT META-DENDRAL EURISKO
About 1/3 of These Systems Are VerySuccessful and Are Still in Use
-
8/6/2019 Chap12. Exp Syst
10/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
10
Latest ES Developments
Many tools to expedite theconstruction of ES at a reduced cost
Dissemination of ES in thousands of organizationsExtensive integration of ES with otherCBISIncreased use of expert systems inmany tasksUse of ES technology to expedite IS
construction
-
8/6/2019 Chap12. Exp Syst
11/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
11
The object-oriented programming
approach in knowledge representationComplex systems with multipleknowledge sources, multiple lines of reasoning, and fuzzy informationUse of multiple knowledge basesImprovements in knowledgeacquisition
Larger storage and faster processingcomputersThe Internet to disseminate softwareand expertise.
-
8/6/2019 Chap12. Exp Syst
12/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
12
12.4 Basic Concepts of
Expert SystemsExpertiseExpertsTransferring ExpertiseInferencing Rules
Explanation Capability
-
8/6/2019 Chap12. Exp Syst
13/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
13
ExpertiseExpertise is the extensive, task-specific knowledge acquired fromtraining, reading and experience
Theories about the problem area H ard-and-fast rules and procedures Rules (heuristics) Global strategies Meta-knowledge (knowledge about
knowledge) Facts
Enables experts to be better and fasterthan nonexperts
-
8/6/2019 Chap12. Exp Syst
14/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
14
IS In Focus 12.2: Some
Facts about ExpertiseExpertise is usually associated with ahigh degree of intelligence, but notalways with the smartest personExpertise is usually associated with avast quantity of knowledgeExperts learn from past successes andmistakesExpert knowledge is well-stored,organized and retrievable quicklyfrom an expert
Experts have excellent recall
-
8/6/2019 Chap12. Exp Syst
15/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
15
Experts
Degrees or levels of expertiseNonexperts outnumber expertsoften by 100 to 1
-
8/6/2019 Chap12. Exp Syst
16/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
16
H uman Expert Behaviors
Recognizing and formulating theproblem
Solving the problem quickly andproperlyExplaining the solution
Learning from experienceRestructuring knowledgeBreaking rulesDetermining relevanceDe radin racefull awareness of
-
8/6/2019 Chap12. Exp Syst
17/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
17
Transferring Expertise
Objective of an expert system To transfer expertise from an expert to a computer
system and Then on to other humans (nonexperts)
Activities Knowledge acquisition Knowledge representation Knowledge inferencing Knowledge transfer to the user
Knowledge is stored in a knowl edge base
-
8/6/2019 Chap12. Exp Syst
18/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
18
Two Knowledge Types
FactsProcedures (Usually Rules)
Regarding the Problem Domain
-
8/6/2019 Chap12. Exp Syst
19/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
19
Inferencing
Reasoning (Thinking)The computer is programmed sothat it can make inferencesPerformed by the I nf ere n c e
E n g i n e
-
8/6/2019 Chap12. Exp Syst
20/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
20
Rules
IF-T H EN-ELSE
Explanation Capability By the justifier, or explanation
subsystem
ES versus Conventional Systems(Table 12.1)
-
8/6/2019 Chap12. Exp Syst
21/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
21
TABLE 12.1 Comparison of Conventional Systems and Expert Systems
Conventional Systems Expert Systems
Information and its processing are usually combined in
one sequential program
Knowledge base is clearly separated from the processing
(inference) mechanism (i.e., knowledge rules separated fromthe control)
Program does not make mistakes (programmers do) Program may make mistakes
Do not (usually) explain why input data are needed orhow conclusions are drawn
Explanation is a part of most ES
Require a ll input data. May not function properly withmissing data unless planned for
Do no t require all initial facts. Typically can arrive atreasonable conclusions with missing facts
Changes in the program are tedious Changes in the rules are easy to accomplish
The system operates only when it is completed The system can operate with only a few rules (as the firstprototype)
Execution is done on a step-by-step (algorithmic) basis Execution is done by using heuristics and logic
Effective manipulation of large databases Effective manipulation of large knowledge bases
Representation and use of data Representation and use of knowledge
Efficiency is a major goal Effectiveness is the major goal
Easily deal with quantitative data Easily deal with qualitative data
Use numerical data representations Use symbolic knowledge representations
Capture, magnify and distribute access to numeric dataor to information
Capture, magnify and distribute access to judgment andknowledge
-
8/6/2019 Chap12. Exp Syst
22/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
22
12.5 Structure of
Expert SystemsDevelopment EnvironmentConsultation (Runtime)Environment
(Figure 12.2)
-
8/6/2019 Chap12. Exp Syst
23/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
23
Three Major ES
ComponentsKnowledge BaseInference EngineUser Interface
-
8/6/2019 Chap12. Exp Syst
24/71
-
8/6/2019 Chap12. Exp Syst
25/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
25
Knowledge Acquisition
SubsystemKnowledge acquisition is theaccumulation, transfer andtransformation of problem-solving expertise from expertsand/or documented knowledge
sources to a computer programfor constructing or expanding theknowledge baseRequires a knowledge engineer
-
8/6/2019 Chap12. Exp Syst
26/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
26
Knowledge Base
The knowledge base contains the knowledgenecessary for understanding, formulating,and solving problems
Two Basic Knowledge Base Elements Facts Special heuristics, or rules that direct the use of
knowledge
Knowledge is the primary raw material of ES Incorporated knowledge representation
-
8/6/2019 Chap12. Exp Syst
27/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
27
Inference Engine
The bra i n of the ESThe control structure or the ruleinterpreterProvides a methodology forreasoning
-
8/6/2019 Chap12. Exp Syst
28/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
28
Inference Engine
Major ElementsInterpreterSchedulerConsistency Enforcer
-
8/6/2019 Chap12. Exp Syst
29/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
29
User Interface
Language processor for friendly,problem-oriented communicationNLP, or menus and graphics
-
8/6/2019 Chap12. Exp Syst
30/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
30
Blackboard (Workplace)
Area of working memory to Describe the current problem Record Intermediate results
Records Intermediate H ypothesesand Decisions1. Plan2. Agenda3. Solution
-
8/6/2019 Chap12. Exp Syst
31/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
31
Explanation Subsystem
(Justifier)Traces responsibility and explains theES behavior by interactively
answering questions Why?H ow?
What?
(Where? When? Who?)Knowledge Refining System
Learning for improving performance
-
8/6/2019 Chap12. Exp Syst
32/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
32
12.6 The Human Element in
Expert SystemsBuilder and UserExpert and Knowledge engineer.
The Expert H as the special knowledge,
judgment, experience and methodsto give advice and solve problems
Provides knowledge about task
performance
-
8/6/2019 Chap12. Exp Syst
33/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
33
The Knowledge Engineer
H elps the expert(s) structure theproblem area by interpreting andintegrating human answers toquestions, drawing analogies,posing counterexamples, and
bringing to light conceptualdifficulties
Usually also the System Builder
-
8/6/2019 Chap12. Exp Syst
34/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
34
The UserPossible Classes of Users
A non-expert client seeking direct advice -the ES acts as a C on s u l ta n t or Ad vi s or
A student who wants to learn - an I n str uc t o r
An ES builder improving or increasing theknowledge base - a P art n er
An expert - a C oll eag u e or Ass i sta n t
The Expert and the KnowledgeEngineer Should Anticipate Users'Needs and Limitations When
Designing ES
-
8/6/2019 Chap12. Exp Syst
35/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
35
Other Participants
System BuilderTool Builder
VendorsSupport Staff
Network Expert(Figure 12.3)
-
8/6/2019 Chap12. Exp Syst
36/71
-
8/6/2019 Chap12. Exp Syst
37/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
37
ES Development
Construction of the knowledge baseKnowledge separated into
D e c l arat iv e (factual) knowledge and P r o c ed u ra l knowledge
Construction (or acquisition) of aninference engine, a blackboard, anexplanation facility, and any othersoftwareDetermine appropriate knowledge
representations
-
8/6/2019 Chap12. Exp Syst
38/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
38
Domain ExpertKnowledge Engineer and(Possibly) Information System
Analysts and Programmers
Participants
-
8/6/2019 Chap12. Exp Syst
39/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
39
ES Shell
Includes All Generic Componentsof an ESN o Knowledge
EMYCIN from MYCIN RESOLVER (was E X SYS)
-
8/6/2019 Chap12. Exp Syst
40/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
40
Consultation
Deploy ES to Users (TypicallyNovices)ES Must be V ery Easy to Use
ES Improvement By Rapid Prototyping
-
8/6/2019 Chap12. Exp Syst
41/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
41
12.8 An Expert System
at WorkSee text or do a demo in Resolver
(Exsys)
-
8/6/2019 Chap12. Exp Syst
42/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
42
12.9 Problem Areas Addressedby Expert Systems
Generic Categories of Expert Systems (Table12.2)
Interpretation systems Prediction systems Diagnostic systems Design systems Planning systems Monitoring systems
Debugging systems Repair systems Instruction systems Control systems
Example in H uman Resource Management(Table 12.3)
-
8/6/2019 Chap12. Exp Syst
43/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
43
TABLE 12.2 Generic Categories of Expert SystemsCategory Problem Addressed
Interpretation Inferring situation descriptions from observations
Prediction Inferring likely consequences of given situations
Diagnosis Inferring system malfunctions from observations
Design Configuring objects under constraints
Planning Developing plans to achieve goal(s)
Monitoring Comparing observations to plans, flagging
exceptions
Debugging Prescribing remedies for malfunctions
Repair Executing a plan to administer a prescribed remedy
Instruction Diagnosing, debugging and correcting student
performance
Control Interpreting, predicting, repairing and monitoring
system behaviors
-
8/6/2019 Chap12. Exp Syst
44/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
44
TABLE
. Exper t st ems i H man es rce anagemen t ategory
Planning y Emp loymentr equ ireme nty Existing emp loymentinv entoryy Strat egic planning
pPlann ingES
Eliminat eany ga ps that may existb etweensupp ly andd emand p
y Recruitmenty Layoff y Overti mey Retireme nty Dismissal
Int erpretation,
Design
y Organizationand process charty Charact eristics of th e joby Requ iredbehaviorsy Emp loyee charact eristics
p J obAna lysis ESFiteachjob into th e total fabric of th eorganization
py Job descri ptiony Job specification
Design, planningprediction
y Emp loyers r equ ireme ntsy Candidat es perfor manc ey Recruiting goalsy Stat eof labor mark ety Stat us of possibl e insid e recruits
pRecru it men t ES
Disting uishb etween qualifications of potential a pplicants andd esirabl eones p
y Methods of r ecruitingy Sources of r ecruitsy Recruiting policies
Planning, d esign,int erpretation,diagnosis
y A pplicationblanky Personality and perfor manc e testy Physical examinationy Selectioncrit eria
pSelec ti onES
Mostlik elymee torganizations standards of perfor manc eandincr ease the proportionof successf ul emp loyees selected
py Selectionratioy A cceptanc e/Re jection
Design, planningmonitoring, control
y Individ ual diff erencey Profita mountof organizationy Contrib utions to mee ting organizational
goals
pCompen sati onES
Meetall th ecrit eria; equ itabl e to emp loyer and emp loyee; inc entiv e providing; andacc eptabl eto th e emp loyee
p
y Methods of paymenty Incentiv e formy Pay l evely Pay str ucturey Rate rang es
Monitoring,int erpretation,instr uction, control
y Quality and quantity or worky Job knowl edgey Personal qualiti esy Perfor manc estandardsy Eval uation policies
pPer f ormanceEva lua ti onES
Eval uate emp loyee perfor manc e to mee twithstandards p
y Salary adj ustmenty Pro motiony Improveme ntof perfor manc ey Layoff y Transf erence
Debugging,
instr uction, control
y Organizations n eedsy K nowl edge, skill, andability n eeds to
perfor mthe joby Emp loyees needs, perfor manc e, andcharact eristics p
Tra iningES
Matchth eorganizations obj ectiv es withth e firms humantal ent, str ucture, climate, andefficiencyp
y Determining training n eedsy Developing training crit eria
y Choosing train ers andtrain eesy Determining training ty pey A ssign ment, plac eme nt, and
orientationfollow- up
Monitoring, d esign,planning
y Stat eof th e econo myy Goals of th ebargaining parti esy Public s enti menty Issues being disc ussedy Labor laws andr egulationy Pr ecedents inbargaining
pLabor - anagemen t Relati on s ES
Negotiat ewithth e uniononth econtract p
y Contracty Locko uty A rbitration
(Source: Bas ed onBy um, D. H. and E. H. Soh, H uman Resource Manag eme nt Expe rt Syst emsTechnology, Expert Systems, May 1994 .
-
8/6/2019 Chap12. Exp Syst
45/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
45
12.10 Benefits of
Expert SystemsMajor P o te n t i a l ES Benefits
Increased Output and Productivity Decreased Decision Making Time Increased Process(es) and Product Quality Reduced Downtime Capture of Scarce Expertise Flexibility Easier Equipment Operation Elimination of the Need for Expensive
Equipment
-
8/6/2019 Chap12. Exp Syst
46/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
46
Operation in H azardous Environments
Accessibility to Knowledge and H elp Desks Increased Capabilities of Other Computerized
Systems Integration of Several Experts' Opinions Ability to Work with Incomplete or Uncertain
Information Provide Training Enhancement of Problem Solving and Decision
Making Improved Decision Making Processes Improved Decision Quality Ability to Solve Complex Problems Knowledge Transfer to Remote Locations Enhancement of Other CBIS
(provide intelligent capabilities to large CBIS)
-
8/6/2019 Chap12. Exp Syst
47/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
47
Lead to
Improved decision makingImproved products and customerservice
A sustainable strategic advantage
Some may even enhance theorganizations image
-
8/6/2019 Chap12. Exp Syst
48/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
48
12.11 Problems and
Limitations of Expert SystemsKnowledge is not always readily availableExpertise can be hard to extract from
humansEach experts approach may be different, yetcorrectH ard, even for a highly skilled expert, towork under time pressureUsers of expert systems have naturalcognitive limitsES work well only in a narrow domain of knowledge
-
8/6/2019 Chap12. Exp Syst
49/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
49
Most experts have no independent means tovalidate their conclusionsThe vocabulary of experts is often limitedand highly technicalKnowledge engineers are rare and expensiveLack of trust by end-usersKnowledge transfer is subject to a host of perceptual and judgmental biasesES may not be able to arrive at conclusionsES sometimes produce incorrectrecommendations
-
8/6/2019 Chap12. Exp Syst
50/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
50
Longevity of Commercial
Expert Systems(Gill [1995])
Only about one-third survived five years
Generally ES Failed Due to ManagerialIssues Lack of system acceptance by users Inability to retain developers Problems in transitioning from development to
maintenance Shifts in organizational priorities
Proper management of ES development anddeployment could resolve most
-
8/6/2019 Chap12. Exp Syst
51/71
-
8/6/2019 Chap12. Exp Syst
52/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
52
A friendly user interface The problem must be important and
difficult enough Need knowledgeable and high quality
system developers with good people skills The impact of ES as a source of end-users
job improvement must be favorable. Enduser attitudes and expectations must beconsidered
Management support must be cultivated.
Need end-user training programsThe organizational environmentshould favor new technology adoption
-
8/6/2019 Chap12. Exp Syst
53/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
53
For Success
1. Business applications justified bystrategic impact (competitiveadvantage)
2. Well-defined and structuredapplications
-
8/6/2019 Chap12. Exp Syst
54/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
54
12.13 Types of
Expert SystemsExpert Systems Versus Knowledge-based Systems
Rule-based Expert SystemsFrame-based SystemsH ybrid Systems
Model-based SystemsReady-made (Off-the-Shelf) SystemsReal-time Expert Systems
-
8/6/2019 Chap12. Exp Syst
55/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
55
12.14 Expert Systems and
the Internet/Intranets/Web1. Use of ES on the Net2. Support ES (and other AI
methods)
-
8/6/2019 Chap12. Exp Syst
56/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
56
Using ES on the Net
To provide knowledge and advice to largenumbers of usersH elp desks
Knowledge acquisitionSpread of multimedia-based expert systems(Intelimedia systems)
Support ES and other AI technologiesprovide to the Internet/Intranet(Table 12.4)
TABLE 12 4 Artificial Intelligence Contributions to the
-
8/6/2019 Chap12. Exp Syst
57/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
57
TABLE 12.4 Artificial Intelligence Contributions to theInternet
Technology Application
Intelligent Agents Assist Web browsing
Intelligent Agents Assist in finding information
Intelligent Agents Assist in matching items
Intelligent Agents Filter E-mail
Intelligent Agents
Expert Systems
Access databases, summarize information
Intelligent Agents Improve Internet security
Expert systems Match queries to users with canned answers tofrequently asked questions (FAQ)
W W W Robots (spiders),Intelligent Agents
Conduct information retrieval and discovery, smartsearch engines (metasearch)
Expert Systems Intelligent browsing of qualitative databases
Expert Systems Browse large documents (knowledge decomposition)
Intelligent Agents Monitor data and alert for actions (e.g., looking forwebsite changes), monitor users and usage
-
8/6/2019 Chap12. Exp Syst
58/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
58
Summary
Expert systems imitate the reasoning processof expertsES predecessor: the General-purposeProblem Solver (GPS).
Failed - ignored the importance of specificknowledge
The power of an ES is derived from itsspecific knowledge
Not from its particular knowledge representationor inference scheme
Expertise is a task-specific knowledgeacquired from training, reading, andexperienceExperts can make fast and good decisions
-
8/6/2019 Chap12. Exp Syst
59/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
59
Most of the knowledge in organizations ispossessed by a few expertsExpert system technology attempts totransfer knowledge from experts anddocumented sources to the computer andmake it available to nonexpertsExpert systems involve knowledgeprocessing, not data processingInference engine provides ES reasoningcapabilityThe knowledge in ES is separated from the
inferencingExpert systems provide limited explanationcapabilities
A distinction is made between a development
environment (building an ES) and a
-
8/6/2019 Chap12. Exp Syst
60/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
60
The major components of an ES are theknowledge acquisition subsystem, knowledgebase, inference engine, blackboard, userinterface and explanation subsystemThe knowledge engineer captures theknowledge from the expert and programs itinto the computer
Although the major user of the ES is a
nonexpert, there may be other users (such asstudents, ES builders, experts)Knowledge can be declarative (facts) orprocedural
Expert systems are improved in an iterative
-
8/6/2019 Chap12. Exp Syst
61/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
61
The ten generic categories of ES:
interpretation, prediction, diagnosis, design,planning, monitoring, debugging, repair,instruction and controlExpert systems can provide many benefits
Most ES failures are due to non-technicalproblems (managerial support and end usertraining)
Although there are several limitations tousing expert systems, some will disappearwith improved technology
-
8/6/2019 Chap12. Exp Syst
62/71
-
8/6/2019 Chap12. Exp Syst
63/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
63
CASE APPLICATION 12.1: GateAssignment Display System
(GADS)Case Questions
1. Why is the gate assignment task so
complex?2. Why is GADS considered a real-time ES?3. What are the major benefits of the ES over
the manual system? (Prepare a detailedlist.)
4. What measures were taken to increase thereliability of the system and why werethey needed?
-
8/6/2019 Chap12. Exp Syst
64/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
64
CASE APPLICATION 12.2:Expert System in Building
Construction (EXSOFS)
-
8/6/2019 Chap12. Exp Syst
65/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
65
CASE APPLICATION W12.1:DustPro--Environmental
Control in Mines
APPENDIX 12 A: Expe r t Sys t ems C i ted in Chap te r
-
8/6/2019 Chap12. Exp Syst
66/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
66
APPENDIX 12 -A: Expe r t Sys t ems C i ted i n Chap te r
System Vendor (Developer) Description
DENDRA L S t for
U i e rsit
(S t for
, C A )It i f ers t
e mol ecul r structur e of u
o compou
sfrom m ss sp ectr l
ucl ea r m agne tic r espo n se
a ta . T
es st em us es a sp eci a l a lg orit
m to s st em a tic a ll en um e r a tea ll possi le mol ecul a r structur es; it us es c
emic a l e ! p e rtis eto pru ne t
is list of possi iliti es to a m anageab le siz e ."
n o ledge in DENDRAL is r epr esen ted a s a proc ed ur a lco de .
E U R ISKO S tan for d U n i e rsit
(S tan for d , C A )It l ea r n s new
euristics and new d om a in -sp ecific de fi n itio n sof co n cepts i n a pro b lem d om a in . T
e s st em c an lea r n byd isco e r y in a n um be r of d iff e r en t pro b lem d om a in s,in clu d ing V LS I de si gn . E U R ISKO op e r a tes by gene ra ti ng adev ic e co n fi g ur a tio n , computi ng its i n put /output behav ior,a ss essi ng its fu n ctio na lit y and then eva lu a ti ng it aga in stot he r comp a rab le dev ic es.
# E T A -DENDRA L S tan for d U n ive rsit y
(S tan for d , C A )
It he lps c he mists de te rmi ne the de p enden ce of m a ss
sp ectrom e tric fr ag m en ta tio n on su bstructur a l f ea tur es. I td oes t h is by d isco ve ri ng fr ag m en ta tio n rul es for g ivencl a ss es of mol ecul es. M E T A -DENDRAL first gene r a tes a se tof h igh ly sp ecific rul es t ha t a ccou n t for a si ng lefr ag m en ta tio n proc ess i n a p a rticul a r mol ecul e . T hen it us esthe tr a in ing e ! a mpl es to gene r a liz e the se rul es. Fi na ll y, t hesyst em r ee ! a mi ne s t he rul es to r emo ve r ed u ndan t or Syst em
V end or ( D eve lop e r ) D escriptio n in corr ect rul es.
S T EA M ER U .S . N avy incoop e r a tio n w it h $ olt,$ e r anek , and N ew m an
In c.(C a m b ri dge , M A)
It is an in te lli gen t C A I tha t i n structs N avy w it h p e rso nne l i nthe op e r a tio n and m a in tenan ce of t he propulsio n pl an t for a1078 -c l a ss fri ga te . T he syst em c an mo n itor t he stu den t
e ! ecuti ng the boil e r li gh t-off proc ed ur e for t he pl an t,a ckn ow ledge a ppropri a te stu den t a ctio n s, and corr ectina ppropri a te one s. T he syst em w or k s by tying a simul a tio nof t he propulsio n pl an t to a sop h istic a ted g r a p h ic i n te rf a cepro g r a m t ha t d ispl ay s an im a ted color d iag r a ms of pl an tsu bsyst ems . T he stu den t c an m an ipul a te simul a tedcompo nen ts li ke va lve s, s w itc he s, and pumps, and obse r vethe e ff ects o n pl an t p a r a m e te rs, suc h a s c hange s i n pr essur e ,temp e r a tur e and flo w . S T EA M ER us es an ob ject-ori en tedrepr esen ta tio n sc he m e .
-
8/6/2019 Chap12. Exp Syst
67/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
67
APPENDIX 12-B:
Classic Expert Systems
MYCINX -CON
-
8/6/2019 Chap12. Exp Syst
68/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
68
1. MYCIN
To aid physicians in diagnosing meningitisand other bacterial blood infections and toprescribe treatmentTo aid physicians during a critical 24-48-hourperiod after the detection of symptoms, atime when much of the decision making isimprecise
Early diagnosis and treatment can save apatient from brain damage or even death
Stanford Medical School in the 1970s by
Dr. Edward H . Shortliffe
-
8/6/2019 Chap12. Exp Syst
69/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
69
MYCIN Features
Rule-based knowledgerepresentation
Probabilistic rulesBackward chaining methodExplanationUser-friendly system
2 XCON (Expert VAX
-
8/6/2019 Chap12. Exp Syst
70/71
Decision Support Systems and Intelligent Systems, Efraim Turban and Jay E. AronsonCopyright 1998, Prentice Hall, Upper Saddle River, NJ
70
2. XCON (Expert VAXSystem Configuration and
Mass Customization)Digital Equipment Corp. (DEC)
minicomputer systemconfigurationManually: Complex task, manyerrors, not cost effectiveCost savings estimated at about$15 million / yearLiterature: Over $40 million / year
l l
-
8/6/2019 Chap12. Exp Syst
71/71
71
APPE NDIX 12-C: Typical Expert Systems Applicat ions
Ma n u f a c tu r i n g/E n g i n eer i n g Product designDesign analysisProcess planning
Assembly mgt.
Process controlDiagnosis & repairSchedulingRosteringSimulationCost estimationConfiguration
Ut i l i t i es & T elec o mm u n i c at i on sConfigurationReal-time monitoring
A larm analysisDiagnosis
Network analysisMarketing analysisMarketing supportBack office operationsSchedulingBilling OperationsProvisioning
A cc o u n t i n g a n d F i n a n c eCredit analysisCustomer servicesLoan eligibilityBanking help desk
Insurance underwriting A uditingStock & commodity tradingFinancial planningTax advisingCredit control
Aer o spa c e/ D o DLogisticsManpower planningSituation assessmentDiagnosis & repair
Inventory managementSeismic analysisTactical schedulingTrainingMunitions requirements
T ra n sp o rtat i onSchedulingPricing
Y ield managementResource allocation
B u s i n ess Ser vi c esProfit forecastingProduct selectionData dictionaryCustom interfacesCustom trainingCustom software toolsSoftware requirements
Mar k et i n g & Sa les A llocation of advertisement mediaMarket assessmentsSalespersons assignment
A ccount marketingProduct selection
Ser vi c e I n d u stryFor a list of applications in the service industrysee Motiwalla and Gargaya [1992]
top related