Quality changes and predictive models of radial basis function neural networks for brined common carp (Cyprinus carpio) fillets during frozen storage

Abstract

To investigate and predict quality of 2% brined common carp (Cyprinus carpio) fillets during frozen storage, free fatty acids (FFA), salt extractable protein (SEP), total sulfhydryl (SH) content, and Ca2+-ATPase activity were determined at 261K, 253K, and 245K, respectively. There was a dramatic increase (P<0.05) in FFA and a sharp decrease (P<0.05) in SH at 261K during the first 3weeks. SEP decreased to 67.31% after 17weeks at 245K, whereas it took about 7weeks and 13weeks to decrease to the same extent at 261K and 253K, respectively. Ca2+-ATPase activity kept decreasing to 18.28% after 7weeks at 261K. Furthermore, radial basis function neural networks (RBFNNs) were developed to predict quality (FFA, SEP, SH, and Ca2+-ATPase activity) of brined carp fillets during frozen storage with relative errors all within ±5%. Thus, RBFNN is a promising method to predict quality of carp fillets during storage at 245–261K.