Quality Model of Ginjo Sake using Hierarchical Fuzzy Neural Networks

Abstract

FNN(fuzzy neural network)and HFNN(hierarchical fuzzy neural network)were applied to construct the models estimating the sensory evaluations of various Ginjo sake samples from the instrumental data of the samples. Errors between the sensory evaluations and estimation by FNN and HFNN models were about 10% and 7%, respectively. Input variables selected in FNN and HFNN models were in good agreement with experts' experience. By the analysis of fuzzy rules, the relationships between these input variables and output variables were found to be almost similar to those from experts' experience.