Organizing the storage of neuro-fuzzy model in relational databases

The paper considers the organization of the storage of ANFIS neuro-fuzzy models in relational databases. Two storage schemes were developed: the universal schema based on the ANFIS model representation as a function of several variables in a computation tree and the specific schema, which reflects layered model structure. Algorithms for computing ANFIS model output for two variants of the storage are described. The possibility of computing ANFIS model output by standard means of SQL is substantiated.