**Abstract**

Sensory evaluations are usually expensive and time consuming but are indispensable to ensure the success of a food product in the market. Artificial intelligence, such as machine learning models, is emerging as a powerful tool in food science to predict or classify sensory attributes. This study aimed to evaluate two machine learning models, a Multivariate Linear Regression Model (MLRM), and an Adaptive Neuro-Fuzzy Inference System (ANFIS) to predict sensory acceptability of mayonnaise. A dataset of 64 data points was used, the analysis of the data indicates that oil content (%), viscosity (Pas), taste score, and texture score were the most pertinent data for the models (correlation index of 0.25–0.75). MLRM showed an adjustment of r<sup>2</sup> = 0.864 in the prediction with a positive bias. ANFIS model demonstrated a slightly higher adjustment (r<sup>2</sup> = 0.910) than the MLRM; additionally, ANFIS fit better to the data as it does not show any bias in the results. The obtained results suggest that tested algorithms can predict sensory acceptability, being a useful approach that can be used for reducing time and number of food sensory evaluations.

**Keywords:** machine learning, mayonnaise, ANFIS, multivariate regression, sensory evaluation
