ARTIFICIAL INTELLIGENCE IN ASSESSING THE INFLUENCE OF EMOTIONS ON BEER ACCEPTANCE IN IMMERSIVE VIRTUAL ENVIRONMENTS

In the field of artificial intelligence, machine learning helps to identify patterns in data and understand how different variables relate to each other. Thus, the objective was to apply machine learning to analyze the influence of emotions on beer acceptance, considering how immersive virtual environments impact consumers' sensory perception. 100 consumers rated an industrial beer (American Lager) in four scenarios: three immersive contexts (abstract, village, and sea) and a traditional sensory booth as a control. Consumers rated the acceptance of the beverage using a 9-point hedonic scale and, in the virtual environments, consumers indicated the emotions felt (nostalgic, disappointed, calm, bored, friendly, satisfied, relaxed, refreshed/refreshed, and contented) when drinking the beer, using the Check-all-the-apply (CATA) methodology, these were carried out in the sensory analysis laboratory of UFV under ethical approval (6.915.346/2024). To evaluate the data, machine learning was applied, using a decision tree (Decision Tree Regressor), which was optimized through the gp_minimize algorithm, from the scikit-optimize library. The 5-fold cross-validation was performed, using the mean square error (MSE) as a metric. After analysis, the most important emotional variables were selected, as they had the most significant impact on the acceptance score. A decision tree was then generated to illustrate how each emotion impacted the acceptance score in the virtual contexts. In the four scenarios evaluated, the acceptance scores varied significantly. In the sensory booth and the abstract context, most of the scores were between 7 and 8, representing around 60% and 70% of the evaluations, respectively. In the immersive contexts Mar and Vila, the predominant scores were between 8 and 9, with 80% of the evaluations in Mar and 71% in Vila, indicating the positive influence of virtual environments on beer acceptance. The decision tree showed that, in the Abstract environment, the emotions happy, invigorated/renewed and satisfied were decisive for beer acceptance. In the Village environment, the most relevant emotions were satisfied, refreshed and relaxed. For the Sea environment, emotions such as satisfied, relaxed, calm, refreshed/renewed, happy, friendly and nostalgic influenced acceptance. The decision tree allowed us to see in an illustrative way how the environment influences consumers' emotions, directly impacting on acceptance, and is a promising methodology for evaluating sensory data, allowing the identification of complex relationships that traditional statistical methods cannot capture.