PURPOSE: To provide a fuzzy inference rule extracting device which can divides the input/output relation into a fuzzy inference rule that shows the essential relation of the input/output relation and a fuzzy inference rule for fine adjustment and can identify these rules by means of a CMAC learning function.

CONSTITUTION: An input part 21 is provided to input the information given from a sensor and the user information together with a membership function storage part 22, plural cerebellum model storage parts 23, a weight calculation part 24 which considers the weight of the parts 23, a falling method learning part 25 which learns the consequent part of a fuzzy inference rule by a falling method, a fuzzy inference part 26 which performs the fuzzy inference based on the antecedent part of the fuzzy inference rule decided by the weight of the part 24 and the consequent part obtained at the part 25, a deciding part 27 which decides the error between the result of the fuzzy inference and the output value received from the part 21, and an output part 28 which selects a part 23 having the least inference error based on the decision of the part 27 and also displays the results obtained through other parts 23.