UBC Theses and Dissertations
Automating knowledge acquisition and representation with fuzzy logic Jordon, Lester Arlington
An interactive computer program has been developed to automate the construction of knowledge bases in expert system development. The program incorporates many user friendly features which allow users who may not understand the concepts of fuzzy logic or the complexity of the knowledge acquisition process to develop useful and complete Expert Systems. Expert systems formalism are being successfully used in the industry. A major bottleneck in development of these systems lies in the knowledge acquisition phase. This study presents an induction system, based on fuzzy logic, which successfully automates the knowledge acquisitions process when the domain knowledge is embodied within discrete numerical data. A tabloid database representing some desired input/output relationship is input to the system. Fuzzy expressions such as "high", "low" or "large" are either created automatically by the system or described explicitly by the user. These fuzzy expressions are automatically associated into an accurate and minimum rule set which models the desired input/output relationship. The system was tested on the development of a knowledge base for a secondary crusher controller and a second more complex non-monotonic input/output relationship. In the controller example the knowledge bases developed were simpler, more accurate and built much faster than using a traditional interview approach. The non-monotonic relationship was also accurately described by a simple knowledge base.
Item Citations and Data