Modeling Quality Changes in Brined Bream (Megalobrama amblycephala) Fillets During Storage: Comparison of the Arrhenius Model, BP, and RBF Neural Network

To evaluate and predict the freshness of brined bream (Megalobrama amblycephala) fillets stored at different temperatures, changes in quality [nucleotide degradation products (IMP, HxR, Hx), K value, sensory assessment (SA), total aerobic counts (TAC), thiobarbituric acid reactive substances (TBARS), and total volatile base nitrogen (TVB-N)] were investigated. The Arrhenius model, back-propagation neural network (BP-NN), and radial basis function neural network (RBF-NN) were established and compared. The RBF-NN predicted changes of SA, TAC, K value, TVB-N, TBARS, and HxR of brined fillets during storage with relative errors all within ±5 %, while the BP-NN values were all within ±10 % (except for the values at day 2 for K value, day 2 and day 4 for HxR). For the Arrhenius model, the relative errors of TVB-N were all within ±10 %, and those of SA, TAC, K value, TBARS, and HxR ranged from 0.58 to 44.37 %. Thus, RBF-NN is a promising method for predicting the changes in the quality of bream during storage at 270–282 K.