| Criteria to consider for successful expert system implementation include: |
 |
Expertise necessary to solve the problem must exist.
Expert systems cannot solve problems that no human expert can solve. In most
cases this means that access to an acknowledged expert who provides advice in the problem
domain is required. For small scale expert systems some of the
expertise might be documented in published sources, but interaction with
a person who has experience in actually solving the problem while you are
developing the knowledge base is still recommended. This will probably
be someone who has actually read and used (perhaps even wrote!) the
published sources. |
|
The task must be clearly defined.
Rule-based expert systems are effective for dealing with advising tasks
for which a finite set of explicit recommendations ("refill the fuel tank" or "recharge
the battery") can be identified in advance
and that use heuristics or rules-of-thumb to produce
these recommendations ("if the starter cranks and there is no fuel
smell, the tank is empty"). Problem situations that are resolved
with vague generalizations are probably not suitable expert system
candidates |
|
Continued... |