**4. Conclusions**

Multivariate linear regression (MLRM) and Adaptive Neuro-Fuzzy Inference System (ANFIS) could be used for predicting sensory data in mayonnaises. Both linear (MLRM) and non-linear (ANFIS) models seems to fit well for this task (good predictability), additionally they are moderately easy to implement, and present good versatility against the differences of the mayonnaises' formulation. Regarding the computational skills required for ANFIs model, it present a higher complexity, but due to its good performance, this model could be proposed for industrial applications or scientific approaches (research and development). In contrast, MLRM is easy to understand, easy for data interpretation, and rapid implementation; these features

could be exploited for academic purposes or to explore simple reformulations by new food manufacturers (i.e: start-ups). For further studies, it is considered as priority to expand the dataset for validating both models with more observations.
