(An) effective inference strategy for fuzzy knowledge-based systems = 퍼지 지식기반시스템을 위한 효과적인 추론방법

Since knowledge-based systems generate solutions to problems by using knowledge, it is important for them to have effective knowledge representation and manipulation methods. In real problems, some pieces of knowledge may be given fuzzily. Such knowledge can be easily represented in knowledge-based systems by employing fuzzy techniques. For manipulating such represented knowledge to produce solutions, several processing techniques are needed. Among them, inferencing for fuzzy rule-based systems, ranking of fuzzy values and tuning of fuzzy production rule-bases are fundamental processing to fuzzy knowledge-based systems. This thesis presents effective methods for such fuzzy knowledge processing. First, for the purpose of inferencing of fuzzy rule-based systems, several measures are introduced which evaluate matching degrees resulting from fuzzy matching, fuzzy comparison and interval inclusion tests occurring in the course of performing inference. Then an inference method is presented for fuzzy rule-based systems which enables the flexible use of both conventional rules and fuzzy production rules. Second, in order to rank fuzzy values, a way to make a fuzzy preference relation for fuzzy values is considered. Then a ranking algorithm for fuzzy values is developed which is based on that preference relation. Third, in order to tune fuzzy production rulebases so as to improve their behavior, two new fuzzy neural network models are introduced which can embody fuzzy production rulebases and carry out fuzzy inference. Two tuning methods based on these fuzzy neural network models are proposed. Finally, a general purpose fuzzy expert system shell called FOPS5 is presented which has been designed in consideration of the proposed fuzzy knowledge processing methods. Our inference strategy is as follows: For fuzzy production rulebases to be used in knowledge base, we refine them with the proposed tuning methods before their use. When we need to perform inference for a fuzz...
Lee, Kwang-Hyungresearcher이광형researcher
Issue Date
101791/325007 / 000925238

학위논문(박사) - 한국과학기술원 : 전산학과, 1995.8, [ vi, 156 p. ]


Fuzzy Neural Network; Tuning; Ranking; Inference; Fuzzy Knowledge-based Systems; Expert System Shell; 전문가시스템쉘; 퍼지신경회로망; 조정; 랭킹; 추론; 퍼지지식베이스시스템

Appears in Collection
Files in This Item
There are no files associated with this item.
  • Hit : 148
  • Download : 0
  • Cited 0 times in thomson ci


  • mendeley


rss_1.0 rss_2.0 atom_1.0