The main aim of this work was to compare response surface methodology and artificial neural network in order to describe the behaviour of the crystallization temperature (Tc) in the production of oleogels of avocado oil. The data of the Tc, obtained by dynamic-mechanical thermal analysis, were adjusted according to response surface methodology and compared with a neural network with varying numbers of neurons in the hidden layer (10-15-20). The network was then trained with Levenberg-Marquardt, conjugate gradient, Polak-Ribiére conjugate gradient and Bayesian algorithms. The results show that response surface methodology explained the behavior of the Tc with a coefficient of determination of R2=0.93. The Bayesian algorithms were more accurate in the adjustment of the data with a mean absolute deviation of 1.35.