ADAPTIVE INFERENCE KNOWLEDGE EXTRACTING DEVICE AND POINT-OF-SALES DATA ANALYZING DEVICE USING SAME

摘要

PURPOSE: To provide the inference knowledge extracting device which can be adapted to a change input/output data. CONSTITUTION: Plural neuro fuzzy models are mounted on the neuro fuzzy model part 4 of a neuro fuzzy knowledge base 3. The respective output nodes of the neuro fuzzy models have arithmetic means which perform arithmetic by a Gaussian membership function. A fuzzy rule extraction part 7 actuates a model parameter learning part 5 and a neuro model learning part 6 to learn the neuro fuzzy models by using input/output data of an input/output data storage part 2. Then a fuzzy rule is represented by using the Gaussian membership function determined with obtained model parameters μklj and δklj and a model parameter ωkl . The represented regularity of the input/output data is adaptively extracted by performing a learning process for the neuro model part 4 at each time when the regularity that the input/output data is dynamically changed.