FUZZY INFERENCE SYSTEM AND FUZZY RULE ORGANIZATION SYSTEM

摘要

PROBLEM TO BE SOLVED: To provide a fuzzy inference system recognizing a pattern with high flexibility even when the teacher data are not sufficient. SOLUTION: The teacher data 107 are divided into clvsters by a data division part 104, and a rule extraction part 105 extracts a fuzzy rule in every divided cluster. A rule adjustment part 106 adjusts a fuzzy rule based on the inference result of a fuzzy inference part 102, and a threshold optimization part 110 optimizes the threshold. Then the part 102 decides that the input data does not belong to any clvster (an abnormal state) when the maximum membership function value exceeds the threshold. The fuzzy rule defines an elliptic area as an input data existence space of every clvster in the existence space of the two-dimensional input data consisting of two dimensions among those dimensions of the input data.