Predictive Modeling of Channel Catfish Under Varying Temperatures: Quality Dynamics and Warning Thresholds.

The objective of this work was to establish mathematical models and an artificial neural network to predict changes in channel catfish quality during storage. Secondary models of microorganisms, using the total viable count (TVC) as an indicator, were established based on the modified Gompertz equation combined with the Belehradek equation. The secondary kinetic models for total volatile basic nitrogen (TVB-N) were developed by combining the primary model with the Arrhenius equation, from which the early warning thresholds for quality change were determined based on the slopes of the kinetic curves. For most samples, the relative error between the measured and predicted values of the secondary kinetic model remained within ±20% across the tested storage temperatures, while during the practically relevant 2-6 days period, the error was tightly controlled within ±15% for the majority of samples. Moreover, the prediction models were established based on Back Propagation Neural Networks and Radial Basis Function Neural Networks, with determination coefficients (R2) exceeding 0.9. In conclusion, the developed predictive models provide a scientific basis and technical support for quality monitoring and cold-chain distribution of channel catfish under varying temperatures.