ランダム復元データに基づくファジィ関連係数の評価

抄録

Previously, the authors have applied the t-norm used in fuzzy set operations to define type-1 and type- 2 fuzzy partition tables. This approach enables the type-1/type-2 fuzzification of fuzzy correlation coefficients. This paper proposes a new probabilistic generative model for evaluating fuzzy correlation coefficients as relational indicators. The proposed model adopts a structure that probabilistically propagates the relational principle “true if true, false if false.” Using the vectors generated by this model, we quantitatively evaluate the directionality and strength of fuzzy correlation coefficients. Furthermore, we examine the validity of these coefficients as relational indicators using random data.