Chapter 5 Expert Systems

1
Chapter 5Expert Systems

2
Expert System

• p. 547
• MYCIN (1976)
• see section 8.2
• backward chaining certainty factor and
  rule-based systems p.233
• Bayesian network p. 239
• Fuzzy logic p. 246
• Probability and Bayes theorem p. 231
• PROSPECTOR (1976), DENDRAL (1978)
• expert systems shells EMYCIN

3
Expert System

• using domain knowledge
• knowledge representation p. 297
• reasoning with the knowledge, explanation
• Knowledge acquisition (p. 553)
• 1) entering knowledge
• 2) maintaining knowledge base consistency
• 3) ensuring knowledge base completeness
• MOLE (1988) is a knowledge acquisition system for
  heuristic classification problems such as
  diagnosing diseases.

4
Expert System

• problem the number of rules may be large
• control structure depend on the specific
  characteristic of the problem
• 1) Brittleness (????????) no general knowledge
  that can be used, the data is out of date
• 2) Lack of meta-knowledge the limitation of the
  control operation for reasoning
• 3) Knowledge acquisition difficult to transform
  the knowledge from human to machine
• 4) Validation the correctness of the knowledge
  in the system, no formal proof that machine is
  better than human or human better than machine.

5
Introduction to Expert Expert System

• ?????? 12.1 ??????????????? MYCIN p.387
• ?????? 12.2 ??????????????? EMYCIN p.388
• ???????????????????????????????? p.388
• ?????? 12.3 ??????????????????????????????????????
  ? p.395
• ?????? 12.4 ??????????????????????????????????????
  ????????? p.396
• ?????? 12.5 ??????????????????????????????????????
  ???????? p.398
• ?????????????????????????????? p.404

6
Knowledge Acquisition and Validation

• ?????? 13.1 ?????????????????????????????? p.416
• ?????? 13.2 ????????????????????? p.417
• ?????? 13.3 ?????????????????? ?????????????
  p.418
• ?????? 13.4 ????????????????????? p.423

7

8
(No Transcript)
9
DEMO
EXPERT SYSTEMS

10
?????????????

• ?????????????????????????? ??? ????????????????
• Decision Support Systems Expert Systems
• ????? ?????????????
• ????? 12 16 p.383-561