An application of fuzzy linear modeling: prediction of uncertainty for betaglucan content

Abstract

Beta-glucan (BG) has positive health effects for the mamalians. However, the BGsources have limited content of it. Besides, the production of the BG has stringentprocedures with low productivity. Economical production of the BG needs theimprovement of the BG production steps. In this study, it is aimed to improve theBG content during the first step of the BG production, microorganism growth step,by obtaining the optimal values of additive materials (EDTA, CaCl2 and Sorbitol).For this purpose, the experimental data sets with replicated response measures(RRM) are obtained at spesific levels of EDTA, CaCl2 and Sorbitol. Fuzzymodeling, a flexible modeling approach, is applied on the experimental data setbecause of the small sized data set and diffulty of satisfying probabilistic modelingassumptions. The predicted fuzzy function is obtained according to the fuzzy leastsquares approach. In order to get the optimal values of EDTA, CaCl2 and Sorbitol,the predicted fuzzy function is maximized based on multi-objective optimization(MOO) approach. By using the optimal values of EDTA, CaCl2 and Sorbitol, theuncertainty for predicted BG content is evaluated from the economic perspective