A Research On Reasoning Model Based On Fuzzy Number-Valued Fuzzy Integral

Approximate reasoning is a fundamental issue in the field of artificial intelligence. In order to make the computer have intelligence; it is necessary to make it have the ability of reasoning, so as to realize the simulation of man's thought. The uncertainty of knowledge and information is aroused from the fuzziness mostly, it makes the research of fuzzy reasoning much more important. Using the fuzzy set as an interior representation can make the reasoning accord with the man's thought better. The paper uses fuzzy numbers to represent the weights in the fuzzy reasoning and presents the reasoning model based on fuzzy number-valued fuzzy integral. The local weights of a rule are represented by fuzzy numbers in this model. This model presents the extended Choquet integral to calculate the reasoning result and also proposes the fuzzy reasoning algorithm based on Petri Nets. As for the fuzzy classification, the global weight of a rule is represented by a fuzzy number. It also presents the extended Choquet integral to calculate the reasoning result and acquire the weights by Genetic Algorithm. The paper tests this method in the Pima, Mango leaf, Rice taste, Wine and Glass data sets. The experiments show that the classification accuracy of the model based on fuzzy number-valued fuzzy integral is higher than the accuracy of the weighted average model.