File - FolajimiClass
Contents
1. Knowleage processor which is modelea atter the expert reasoning power Processor in an expert system that matches the facts contamned m the working memon withthe domain KnNowedage contamned n tne Knowedge pase to araw Conclusion about the problems m ittaps tne knowedge pase and workng memory to derve new Mnitormation and solve problems The Structure of ES THE USER INTERFACE ine user communicates witn ine expert system through the user interface t allows the user to query the system supply Information and receive advice Whe aims are to provide the same form of cCommuUnNIcaton facilities provided Dy tine experi But normally has less capability of understanding natural language and general knowledge The Structure of ES THE EXPLANATION FACILITY A trademark of expert systems ability to explain thelr reasoning An additional component of expert system l ES can provide explanation on WHY itis asking the question HOW it reached some conclusion The Structure of ES Explaining WEN Example MM Will the car not start Person WHY MM HS Know imat tne car wont start then usually assume the problem Is the electrical system l Expert responds wiin what iney mgoni conclude fromthe answer ES respond to a WHY query by displaying the rule It IS UK pursuing The Structure of ES Explaining mi Oday Besides giving final results expert system can explain how IL arriveg2. Table of Content What is an ES Conventional and ES The Characteristic of ES The Structure of ES The Characteristic of ES ES Development Lifte Cycles Participants in ES Development Selected Business Expert Systems and Functions Advantages ot ES Limitations of ES When to Use Expert Systems J stitying the Problem Domain MYCIN A medical expert system WEBCANDI A Web based Expert System for Cancer Diagnosis Building TOolS What is an ES Expert System ES is a branch Mo Artifielal Intelligence iat attempt to mimic human experts Expert systems caneliner Support decision makers or completely replace them Expert systems are the most widely applied amp commercially successiul Al technology What is an ES Prot Edward Feigenbaum oil Stanford University leading researchers in ES has producega ihne iollowing deiinition so wiitalllejgnteorniott2ar dgdejseislitsein Ua no WMIsele 2sulel Innaranog orooscltura3 to solve probiems thatare eliieultanotejnuto require significant human expertise for their solution Whatis an ES gt Expertise Is the extensive task specitic Knowledge acquired from training reading and experience gt The transter of expertise from an expert toa computer and then to the user involves four activities v Knowledge acquisition trom experts or other SOUrCes Knowleage representation In the computer Knowledge iniereneing resutmg ina recomm
3. Although more restrictive and require more memory they are effective the way they present rules and control their processing hey process symbols instead of numbers to reach concd sion on allogicalTevelof Knowledge representation 65 Programming Language Example of L ISP programming language rule 43 eoncerns car auto monilehome application It mn vehicle frame antecedent 1s if entity has wheels and owner and roof production _Is then entity isa auto aliernate productions hen entity isa mobilehome iurtner information top _ aown see transportation status sympol bottom up see wheels driver owner confidence use function auto_confidence update rule apply learning function number 32 66 Programming Language Example o PROLOG programming language gt StUdent Cavid gt siudent mary S3jilfe nie fe ala gt siugent ann Fa gt siuaent Jerry gt siudent Sue Likes Someone accounting concentrating Someone accounting RULE gt Concentrating david mis gt Concentrating mary marketing gt Concentrating mary mis gt ConNcCentrating Jonn accounting gt Concentrating ann management FACTS gt concentrating ann mis gt concentrating Jerry finance gt concentrating sue accounting Support Aid and Tools SUpport aids automate the time consuming phase of acquisition Improve ettectiveness ofi re
4. a Nospitalkacgquiredimntectioniz 9 ein clinical history consistent with ERIGEOTHI IS hie Sally clinical history consistent with OlMIMNS MEDIA 7 favre ine patient ever been to an area tnat is endemic Tor coccidiomycoses UnKnown 12 IS Saly a burn patient Yas WA DoncIuSIon The type of intection may be BAC TERIAL 93 A consultation With MYCIN cont iNe final phase prescription Ol antibIoties 13 Does SALLY have clinically significant allergic reaction to any antimicrobial agent No 14 Is Sally pregnant of breast feeding NO 15 Do you have reason to suspect that SALLY may DaveJympaired renal TUnctions No 16 SALLY weight in kilograms 51 4 04 A consultation With MYGIN My preferred therapy recommendation s as follows Give the TolownNg combination 1 ETHAMBUTAL Dose 1 28g 13 0 100mg tablets q24h PO for 60 days then 770 mg 7 5 100 mg tablets a24h PO Comments periodic vision screening tests are recommended Tor optie nNneuntis INH Dose 513 mg 8 0 100mg tablets q24h PO RIFAMPIN Dose 600 mg PO q24h Comments Administer dose on empty stomach OD HOW GOOD AN EXPERT IS MYCIN In one complex evaluation 8 independent evaluators evaluate NY GIN with 9rotaers human diagnostican Tor 10 difficult cases of meningitis The task used was the selection of drugs for cases of meningitis betore causative agents haa Been lelevalitiicvep WO phases of the eva
5. solution time E SyYMDOIC knowledge and reasoning A stable domain Size Ofthe Knowledge base 100 rules for firsttime domain Available test cases Complexity ofthe domain Degree of uncertainty or fuzziness Demonstration of worth scarce expertise Appropriate depth ofrequired knowledge 44 MY GIN A medical expert system Developed at Stanford University in the mid 1970 s he first large expert system that perform at the level ofi human expert and use as benchmark by expert system developers Provide consultative advise about bacteremia and meningltis Bacteremia is an infections that involve bacteria in the blood 45 MYCIN A medical expert system Meningitis is an inieciion which is an inilammalt on ey tale memnsnoranes thatenvelop ine prain and spinal Gora Can be fatal thus need quick response but positive identification normally takes 24 48 hours Normally doctors must pegin treatmentin tne absence of lab results Very complex and doctor need advice of an expert of pacteremia and meningitis 46 How MYCIN works MY GIN begin dby Initiating a dialogue Physician response to the questions MYCIN provides a diagnosis D prescription 4 How MYGIN reasons Laboratory results of body fluid analyses Symptoms that patientis displaying Characteristics of the patient such as age sex etc 48 MYCIN MYCIN consultative proceeds in 2 phases Diagnosis is made to identit
6. Expert Systems Needs complete Information to Can operate with incomplete or operate uncertain Information Effective manipulation of large Effective manipulation of large databases knowledge bases Representation and use of data Representation and use of Knowledge Efficiency Is a major goal Effectiveness Is a major goal Easily deals with quantitative data Easily deals with qualitative data The Structure of ES Human Expert Problem solving The Structure of ES Expert System problem solving The Structure of ES THE KNOWLEDGE BASE LONG TERM MEMORY ine Key poitleneckin developing an expert system 3 Gonialin everyining necessary tor understanding formulating ana soving a problem 2 It contains facts and heuristics ine most popular approach to representing domain KNOwWedge S3 using prod ction rules Rule 1 lE car wont start THEN problem in electrical system Rule 2 IF problem in electrical system AND Battery voltage 1s below 10 volts THEN pad battery The Structure of ES WORKING MEMORY SHORT TERM MEMORY Contains facts about a problem that are discovered auring consultation with ine expert system 3 System matches thisdntormatuon with Knowledge contained 1n tne Knowledge base to inter new tacts ine conclusion reach will enter the working memory Thus working memory contain Iniormation eiiner supplied py user or inter by ine system The Structure of ES INFERENCE ENGINE
7. all phases of the DIEVI production process 39 selected Business Expert Systems and FUNC ONS System Developer Business Activity Function Inspector Banking Monitor Worlawide foreign exchange trading to identity irregular activities Prohibited Law Help attorney evaluate Transaction transactions subject to Exemption Employee Retirement TPE Analyst Income security Act Personnel Personnel Help devise employee Policy Expert policies amp write employee handbooks 40 When to Use Expert Systems 7 Provide a ign potential eyehVeyit or sighiticantiy reduced downside nisi D Capture and preserve irreplaceable human expertise Provide expertise needed at a number of locations at tne same time orin a nostile environment tnatis dangerous to human neath 41 When to Use Expert Systems Provide expertise inatisexpensive or rare Bevelop a solution tastertnan human experts can Provide expertise needed for training ang development to share the wisgdom ot human experts wiin alarge number of people 42 Justiiying tne Problem Domain Fhe first step toward successful system is to pick ine ngnti propleem and n US UH IL selection A Selecting the nont orobem shold Pe tne Tirst consideration mM ES development 3 NIS step entalis Tdentifying the domain expert tne user and the payott trom the system 43 Justitying the Problem Domain Key Domain Characteristics A narrow well defined focus E Moderate
8. bility Because of the large amount ot knowledge that an expert system may have it is important to have an efficient mechanism for modifying the knowledge pase m Characteristics of Expert Systems 7 Symbolic Reasoning a Exper systems represent Knowledge symbolically as sets of symbols that stand for problems concepts hese symbols can be combined to express relationship between them When these relationship are represented In a program they are called symbol SIrUCLUTES For example Assert Ahmad has a fever Rule IF person has fever THEN take panadol Conclusion Ahmad takes panadol 21 Characteristics of Expert Systems o Reasons Heurisiically Experts are adapt at drawing on their experiences to nelp them efficiently solved some current problem Typical heuristics used Dy experts always check the electrical first People rarely get a cold auring the summer if l suspect cancer then I always check tne family DIStOTy eism llsiteu s Expert systems can make mistakes since ine Knowledge of expert have to be captured as close as possible in expert system like its human counterpart it can make mistakes 22 Characteristics of Expert Systems 10 Thrives on Reasonable Complexity ine problem should be reasonably complex 16 too easy Or too difficult 11 Focuses Expertise Most experts are Skillful at Solving problems within their narrow area of expertise but have limited a
9. bility outside this area 23 ES Development Life Cycles ESDLC a FESBLC contains the following phases j omoi Assessment Knowledge Acquisition Dyets lela Testing DISE OA MreUaltetarctalers 24 ES Development Life Cycles Reformulation Requirements Exploration Knowledge an Refinement Structur e Dhae Evaluatio n gt Product ocumente ES Development Life Cycles Assessment 3 Determine Teasipility amp Justiicauon eye tne problem J Defne overall goal and scope 0 fot a3 project 2 Hesources requirement Sources ot knowledge 26 ES Development Life Cycles 2 Knowledge Acquisition E Acquire the Knowledge ofithe problem E nvolves meetings with expert E Bottleneck in ES development 2 ES Development Life Cycles 3 Design Selecting KNowedge representations approach and problem solving strategies Detined overall str cture ana organization of system Knowledge Selection of software tools B ilt initial prototype iterative process 28 ES Development Life Cycles 4 Vesting KIE elceleresssssuialcelule 16 U M L6 928 O Testing and modiiying system Knowledge Study the acceptability of the system by end user l e d Closely with domam expert tnat guide tne growin ot tne Knowledge and end user tnat guide Tn user interface design 29 ES Development Life Cycles 5 Documentation E Compile al tne projects miormation nto a document for t
10. d at a result s Example MM The battery is bad Person HOW MM Since your car won t start assumed there was a problem with the electrical system ITounad the Daltery voltage was below 10 knew the battery was pad ES respond by tracing back through the rules that fire the conclusion h is tracing is a map of the system line of reasoning Characteristics of Expert Systems 1 High level expertise ine most useiul characteristic of an expert system nis expertise can represent ine best thinking evi top experts IN the field leading to problem solutions tnat are imaginative accurate and eficient 2 Adequate response time line system must also perform in a reasonable amount ottime comparable to or Detter than tne time required Dy an expert to solve a proplem Characteristics of Expert Systems 3 Permits dnexact Reasoning hese types of applications are characterized by information tnatis uncertain ampDgUOUS Or unavailable angd ASE TONE SE inherently mexact 4 000 Reliability ine system must be relable and not prone to crashes because it will not be used Characteristics of Expert Systems 9 Gomprenensipility he system should pe able to explain ine steps ofits reasoning while executing so that it Ts nderstandable s iNe systems should nave an explanation capability 1m tne same way tnat human experts are suppose to pe able to explain tnei reasoning 6 Flexi
11. endation or noViceSs v knowledge transfer to the user Whatis an ES CASE GE Models Human Troubleshooters Problem 2 GE wanted an effective amp dependable way of aisseminating expertise to its engineers amp preventing valuable knowledge from retiring from the company Solution Gi decided to build an expert system that modeled the way a numanirouplesnootier Works line sysiem builders spend several moninsintenviewing an employee amp transter tnei Knowedge to a computer m INe new diagnostic technology enables a novice engineer to ncover a talt oy spending only a tew min tes atthe computer terminal Results he system is currently installed at every railroad repair shop served by GE Conventional and ES Conventional Systems Expert Systems Knowledge and processing are Knowledge base Is clearly combined in one sequential separated from the processing inference mechanism knowledge rules are separated from the control Programs do not make mistakes Program may make mistakes CIS EE 0 Do not usually explain why input Explanation is a part of most program data are needed or NOW expert systems conclusions were drawn The system operates only when it The system can operate with only iS completed a few rules as a first prototype Execution is done on a step by Execution is done by using step algorithmic pasis neuristics and logic Conventional and ES Conventional Systems
12. he user and developers of the system such as User manual TETEME E Knowedge dictionary 30 ES Development Life Cycles ow Maintenance 2 Reiinea and upaate system Knowleage to meet current needs 31 Participants in ES Development The main participants in the process of p ilaing an expert system are a the domain expert o Me knowedge engineer c the user 32 Participants in ES Development THE DOMAIN EXPERT E XU person who Nas the special knowledge judgment experience SKills and methods to give advice and solve Problems n a manner superior to otners Although an expert system usualy models one or more experts it may also contain expertise from other sources such as books and journal articles 2 Qualifications needed py the Domain Expert Has experi Knowledge Has efficient problem solving skills Can communicate the knowledge E AVIEN Mu st pe cooperative 33 Participants in ES Development you call someone an expert for a project treat that person like one Even if the person doesn t know evenining apot tne domain tne person KNOWS more than you Patrick E Dessert 34 Particioants in ES Development THE KNOWLEDGE ENGINEER E A person who designs pullds and tests an expert systems TOES SSA TST ENSE Has Knowledge engineering skills an ol bullding expert system Has good communications skills Gan match problems to software nas expert system program
13. luation a MCYIN and 9 human experts evaluate 10 cases pD Each of them prescribe medications 96 HOW GOOD AN EXPERT IS MYCIN Ivo evaluative criteria was used to see whether the prescriptions a Would pe eiecive against tine aciualpactena atter it was tinaly dgentifiegd bo Adequately covered for other possible bacteria wnile avoiding over prescribing o HOW GOOD AN EXPERT IS MYCIN Results oniena 1 MYCIN and 3 other humans expert consistently prescribe therapy that would have been effective for all 10 cases Criteria 2 MYCIN received higher ratings 65 correct in all the cases whereas human expert 42 5 10 62 5 08 HOW GOOD AN EXPERT IS MYCIN MYCIN strengths Is based on 4 factors a MYCIN s knowledge base is extremely detail because acquired rom tne pestnuman practitioners D MYCIN do not overook anything or forget any details It considers every possibility c MYCIN never jumps to conclusions of fails to ask for Key pieces Of information d MYGCIN s maintained at a major medical center and consequently completely current MYCIN represents 50 man years of effort Back to Main nu Building Tools Since mid 1970s a wide choice ofitools and approaches fro developing ES have become available la range trom nionieveland Al programming language to shells to ready to use customized packages for indu stry and govermnment Wanich ojo to adopt aepena on he nature of the pr
14. ming skills 39 Participants in ES Development A KNOWLEDGE ENGINEER E nave been working as Knowledge engineeriora sollware house tor two years Each project 1s ditterent The Job IS challenging ana requires creative thinking and strong communication skills started as Junior knowledge engineer at a salany of 530 500 am now a lead engineer with a salary of 40 700 plus a nice annual bonus z GhrstineMelekran 96 Participants in ES Development THE USER S a person who uses the expert system oOnce ltis developed Gan alid n Knowledge acquisition giving oroad understanding of the problems Can aid in system development 3 selected Business Expert Systems and melnleiilojar System Developer Business Activity Function AS ASO Arthur Andersen Accounts Aid auditing Receivable procedures Authorizers American Consumer Evaluate credit Assistant Express Credit records to protect against credit card fraud Helpdesk Publix Retailing Handle problem advisor Supermarkets calls from store managers 38 selected Business Expert Systems and FUNC ONS System Developer Business Activity Function Intelligent Nippon T amp T Personnel Coordinate Secretary schedules of company personnel Mortgage Arthur Banking Help loan officer Ioan Analyzer Andersen make final decisions on home mortgage loan Direct Labor Ford Motor Manufacturing Improve efficiency Mgmt System Company in
15. oblem he skill of the builder m he function it Is expect to perform either diagnoses or 16 f 6 1 16 60 Building Tools Several levels or categories of tools are available for building ES such as follows LOAN EOE Support aids and tools Specific ES packages Shells ine figure pelowihustrates the 4 levels p944 61 Building Tools he tigure below illustrates the 4 levels Includes domain Mes Less programming Knowledge Base Less flexibility specific expert system application package te Ser custom made More focused More programming More flexibility Build from scratch m 98 29 m rt Em Am A Programming Canguage The language written is Important because Tt determines tne eticency and pernrormance of tne tool ES can be programmed in a variety of languages botn standard or Al onented Standard programming or algorithmic approach is a brute force method in which standard procedural languages are used to develop the IF THEN rules 63 Programming Canguage Ty Standard programming language are used for i3edollowing reasons Limitation ot hardware operating system Power and speed ofthe language E alere eye f tt o g ordigita 0 e U 46 0 6 0 n 6 S2 VIO Ss In the system Language execution time efficiency 64 Programming Canguage 1 AFprogramming are done tnrough symDOIIC language such as LISP and Prolog
16. presentation ensure erisp numan macnine nterace ana depugging aldas timat traces tme reasoning expanatomacilties sNOw NOW the system amves ata particular Solution diting facility ensures that the syntax Is correctly represented in the knowledge base 68 Specific Expert System packages The easiest product to use from users of view Specific ready to use program that advises a specific user in a specific industry to address a specie propleem adoman For example a consultation system that advises student on elective courses to take ine system essentially plugs In tne vital variable matches them against established criteria and determines the courses to take 69 Shells A shell provides a basic architecture for building ES Shells perform 3 different functions a Assists in building the knowledge base by allowing ine developer to mser KNnomedgeTmnto Knowledge representation structures o Provides methods otinterence ede deduction that reason on the basis of information in the Knowledge base and new tacts input by the user c Provides an interface that allows the user to set up reasoning task and query the system about Its reasoning strategy 740 Shells The main 3 Components a The knowledge base p Inference engine c User interface a Man components Rule Base et
17. y the mostlikely Infection organisms LSE AT S EOS 49 MYCIN Example of a rule IF the stain of the organism 1s gram positive AND tne morphology of ine organism IS COCCUS AND the growth contormation of the organism Is CIUImps THEN there Is suggestive evidence 0 7 that the Identity Of the organism Is staphyloccus 90 A consultation With MYCIN 1 Please enter information about the patient NEMIS Age Sex REW Sally 42 aars rermcle Melayu 2 Are there any positive cultures obtained from SALLY Nie 3 Are there any pending cultures of SALLY Yes 4 Site date collected stain examined method GSE REINE Yes A A 5 Stain of smear ivpe NONEiino organ sm was SSSA smear None Unable to reach conclusion trom ab test like physician continue asking questions 5 A consultation With MYCIN cont 6 Has Sally recently had symptoms ei persistent headache or ay abnormal neurologic symptoms dizzaness lethargy etc Yas 7 das SALEY recently had objective evidence ei abnormal nNeUrdlogIie signs documented Oy physician opsemvation Of examination YES Questions 6 and 7 traditional approach to diagnose 0 Symptoms Conclude from the questions by giving first conclusion S amp a The CSF cultures will be considered to be associated with MENINGITIS 52 A consultation With MYCIN cont I gt TONOwNg questions wiinadentity paricularorganismitnat cause the Infections 8 IS the meningitis
Download Pdf Manuals
Related Search