PURPOSE: To provide a fuzzy inference rule generator capable of generating judging rules containing fuzziness by classifying data using attributes and category classes.

CONSTITUTION: An initial example set preparation part 21 prepares an example set from a supplied data group, a membership function generation part 24 generates a membership function from the example set and an example set dividing part 25 divides the example set based on the membership function and defines the divided pieces as the new example sets. An example set evaluation part 22 judges the propriety of division based on the new example sets. In the case of performing the division, a dividing example set selection part 23 selects the example set to be divided and repeats processing after membership function generation to the selected example set. At the point of time when the division is completed, a rule generation part 26 generates the discrimination rules provided with fuzziness based on the membership function and the corresponding category classes.