To acquire general inference knowledge corresponding to an object by performing the selection of an effective input, fuzzy division, and the setting of a membership function and a rule synchronously under the same evaluation.

A knowledge structuring part 2 reads in output data from a data file 6 and a fuzzy division part 3 selects an input or generates an arbitrary point on the dimension of the selected input, generates a new point by division between adjacent points, deletes a selection point, and generates a fuzzy set and a membership function from those results. A rule generation part 4 generates the rule from the fuzzy set, membership function, and input/output data and an inference evaluation part 5 performs inference, evaluates results, and selects an excellent result. The knowledge structuring part 2 repeats those processes, and finally selects an effective input for an output, performs fuzzy division, and generates a fuzzy set, a membership function, and a rule to acquire inference knowledge corresponding to the object in a knowledge file 7.