LAUSR

Abstract Obtaining information about honey from Mozambique is the first step towards the economic and nutritional exploitation of this natural resource. The aim of this study was to evaluate physicochemical (moisture, hydroxymethylfurfural “HMF”, electrical conductivity, Pfund colour, CIE L*a*b* colour and sugars) and rheological parameters elastic modulus G′, loss modulus G″ and complex viscosity η*) obtained at different temperatures (from 10 to 40 °C). All the physicochemical parameters were in agreement with the international regulations. Most of the honey samples were classed as honeydew honey since they were dark and had conductivity values above 0.800 mS/cm. The moduli G′, G″ and η* decreased with increasing temperature. G ′ and G″ were strongly influenced by the applied frequency, whereas η* did not depend on this parameter, demonstrating Newtonian behaviour. An artificial neural network (ANN) was applied to predict the rheological parameters as a function of temperature, frequency and chemical composition. A multilayer perceptron (MLP) was found to be the best model for G″ and η*(r 2  > 0.950), while probabilistic neural network (PNN) was the best for G′(r 2  = 0.758). Sensitivity testing showed that in the case of G″ and G′ frequency and moisture were the most important factors whereas for η* they were moisture and temperature.