Shelf life determination of honey using an accelerated method

This study predicted the shelf life of honey using an accelerated method, setting the hydroxymethylfurfural (HMF) content at 60 mg/kg as the perishability limit. HMF content, water activity, free acidity, pH, and color were evaluated over time at storage temperatures of 45, 55, and 65 °C. A kinetic study was conducted to determine zero- and first-order reaction rate constants for HMF formation, as well as the activation energy (Ea) and temperature acceleration factor (Q10). Artificial Neural Networks (ANNs) were employed to estimate honey shelf life (output layer) from temperature data (input layer), using the K-fold and Holdback validation methods. A significant correlation was observed between honey color change and HMF formation during heating (P < 0.05). HMF formation followed a zero-order kinetic model (0.9856 ≤ R² ≤ 0.9949), with an Ea of 29.349 cal/mol and a Q10 of 3.92. The ANN model using the Holdback method and 10 neurons in the hidden layer showed excellent predictive performance (R² = 1; RMSE = 1.07 × 10⁻⁵), estimating a shelf life of 344 days at 25 °C. Colorimetry results also demonstrated potential for the indirect evaluation of HMF content, supporting its applicability in honey quality and preservation assessments. These findings highlight the effectiveness of combining kinetic modeling with ANN for honey shelf life prediction.