ニューラルネットワークと遺伝的アルゴリズムを用いた吟醸酒の品質モデリング

This paper deals with the quality modeling of Ginjo sake using a neural network (NN) and genetic algorithm (GA). A NN model was constructed to estimate 7 sensory evaluations concerning the quality of Ginjo sake from 18 chemical component analytical values. The performance index, J, of the NN model was significantly small compared with that obtained using multiple regression analysis (MRA). Using the model, analytical data on the chemical components was estimated from the 7 given sensory evaluation values by means of a genetic algorithm, which was employed as an optimizing method. It was found that almost all the estimated values coincided with the actual values within an error range of less than 0.3.