|What kinds of problems make good candidates for expert systems? Problems involving diagnosis like our "my car won't start" scenario are frequently approached with this technology. In fact, many of the early expert systems research efforts focused on medical diagnosis. Expert systems might support Planning activities like developing a marketing strategy for a new product. Instructional expert systems can take advantage of the adaptive nature of expert system consultations to diagnosis student weaknesses and customize instruction. Systems have also been developed that support configuration activities like making sure all the necessary components for a customized equipment installation are shipped together.|
|How are expert systems developed?
Knowledge engineering is the process of codifying a human
expert's expertise and representing that expertise in the knowledge
base. If the expert is not capturing his or her own knowledge, this
process is usually based on interviewing techniques.
Attributes that define problem structure must be identified and named. The subset of the attributes that will represent the goals for a consultation must also be identified.
Rules can then be constructed. Rules combine attributes into logical expressions in the premise and assign values to attributes in the consequent.
Control strategies such as choosing whether to use forward or backward chaining, and the user interface must then be specified. The objective is to make sure the interaction with the expert system user will be similar to interacting with a human expert.
|General advice on picking appropriate initial applications is to pick a problem that an expert could solve in about an hour and that is solved repetitively. It is also important to pick a problem for which access to a human expert is available. Expert systems cannot solve problems that human experts cannot solve.|
This concludes our Expert System Introduction. If you'd like to learn more about how expert system software works, the Inference Methods and Uncertainty tutorial is recommended. If you are interested in developing knowledge bases take a look at the Knowledge Engineering presentation. If you have never run the expert systems provided on this Web site, Using eXpertise2Go's Knowledge Bases is a good starting point.