Expert System

Designing an Expert System - Choosing a

Problem
GUIDELINES TO DETERMINE A PROBLEM IS SUITABLE FOR AN EXPERT SYSTEM
SOLUTION:
❖ Need for a solution must justify the cost involved in the development.
❖ Problem may be solved using symbolic reasoning techniques.
❖ Human expertise is not available in all situations where it is needed.
❖ Problem is well structured and does not require common sense knowledge.
❖ Problem cannot be easily solved using more traditional computing methods.
❖ Cooperative and articulate experts exist.
❖ Problem is of proper size and scope.
Knowledge Engineering

 Based on the problem knowledge need to be extracted from the expert and
represent it in expert system shell.
 This is the job of KNOWLEDGE ENGINEER(KE) in collaboration with
Experts and End Users.
 KE → AI Language and representation expert.
 KE must be first familiar with problem domain.
 KE will abstract general rules and checks with the expert.
 To develop initial prototype, the knowledge engineer must make provisional
decisions about appropriate knowledge representation and inference
methods.
Rule Based Systems

 A very common inference engine is based on representing knowledge in

a rule base.
 Each rule is of the form:
if X then Y (X → condition & Y → consequents)
 Rule base → deduction system (consequent → new fact)
 Rule base → reaction system (consequent → action)
Rule Based Systems
R1: If ?x has wings
then ?x is a plane
R2: If ?x flies
then ?x is a plane
R3: If ?x runs on tracks
then ?x is a train-or-tram
R4: If ?x is a plane
?x can take off vertically
?x has rotors
then ?x is a helicopter
R5: If ?x is a train-or-tram
?x stays underground
then ?x is a subway car
R6: If ?x is a helicopter
?x made in South Africa
then ?x is a Rooivalk
Deductions in Rule Bases – Forward Chaining

 While(no new assertion made) and (unresolved)

a. For each rule (and for each possible binding)

i. Try to support rule’s conditions from known facts

ii. If all supported then assert consequent

Deductions in Rule Bases – Backward Chaining
 While(no untired hypothesis) and (unresolved)

a. For each hypothesis

i. For each rule with hypothesis as consequent

1) Try to support rule’s conditions from known facts or via

recursion (trying all possible bindings)

2) If all supported the assert consequent

Characteristics and Advantages of
Expert System
CHARACTERISTICS:
 High Performance
 Adequate Response Time
 Reliability
 Understandable

ADVANTAGES:
 Availability, Cheaper, Reduce danger,
 Permanence, Multiple Expertise, Explanation, Fast Response
 Unemotional and response at all times
Roles and applications of Expert Systems

ROLES
 Expert
 Knowledge Engineer
 User

APPLICATIONS
 Domains – Design, Medical, Knowledge
 Systems – Monitoring, Process Control
 Finance & Commerece