Knowledge based systems

1
Knowledge based systems

• There are alternatives to the rule-based approach
• For example
• Case Based Reasoning decision making based on
  prior experience
• see course readings p 205 for an example of a
  rule base/case hybrid
• 'this case?' -----------gt prior cases
• Structure of a case based reasoning system
• A library of historical cases (knowledge base').
  
• A means of using key elements of the present
  problem to find and retreive the most similar
  case(s) from the library ('reasoning').
• A means of modifying the proposed solution when
  the case on which the solution is based in not
  identical with the current problem ('reasoning').
  

2
Knowledge based systems

• Building a case based system involves the
  development of a robust and well
• indexed set of cases).
• The system retrieves cases, matches and compares
  significant features with
• current problem
• selection may involve
• closeness' or similarity measures
• the searching of a shared feature' network
• the order of importance or weightings of
  attributes

3
Knowledge based systems

• Advantages
• a case library is a more explicit representation
  -- it represents the domain directly rather than
• through an interpretation of expert/human
  decision making.
• knowledge acquisition is simplified.
• the system can reason by analogy.
• Disadvantages
• Computational cost of searching the library
• issues concerned with proper indexing, quality'
  of case library, number of cases to use etc
• are also important
• . An evolving area of KBS development

4
Knowledge based systems

• Future trends
• Hydrid Approaches ?
• combining different approaches within the one
  system. For example
• a rule based approach with a case based approach
• a rule based and a frame based approach
• symbolic and connectionist (numeric) systems
• Adaptive systems?
• systems that change' or adapt during use. For
  example
• adapting the knowledge base
• learning from examples

5
Knowledge based systems

• Legal Issues
• Liability How liable are the developers of
  (KBS/ES) software for errors that cause injury to
  a person?
• strict liability applies to products
• negligence applies to services
• some ambiguity is computer software a product or
  a service?
• ES or knowledge based systems may provide
  decisions on which people might act. Incorrect
  or erroneous
• decision may cause harm.
• issues
• can an expert/KBS system be considered to
  like a text or reference book?
• the expert system/KBS may make a mistake due
  to poor verification/validation
• (ie a programming' error) or ...
• the mistake may be due to human expert
  (which the system replicates)
• use of the ES/KBS ...
• ie as an advisor with a person making
  the final decision
• or initiating action automatically
  based on the decision

6
Knowledge based systems

• Evaluation of Candidate Expert Systems Shells (or
  development platforms)
• General Criteria
• ltmost likely relate to evaluation of any
  softwaregt
• hardware requirements
• testing/de-bugging support
• quality of documentation
• ltetcgt
• Expert System specific criteria
• knowledge representation schemes (rule, frames
  etc)
• type of reasoning strategies (eg backward
  chaining, forward chaining)
• ability to access an external database
• uncertainty management options (certainty
  factors, fuzzy approaches etc)
• ability to partition or modularise the Knowledge
  base
• explanation ability
• limitations eg on the Knowledge Base size

7
Next time

• Review the subject as a whole and
• An exam briefing