Ciência de dados e aprendizado de máquina aplicados à análise sensorial de cervejas

Abstract

Chemical Engineering plays a fundamental role in optimizing industrial processes, with a particular emphasis on the food industry. In this context, beer production stands out as an important field of research that combines tradition and technological innovation to compete in a highly competitive market. Sensory characteristics play a crucial role in beer, influencing consumer preferences. Sensory analysis in the food industry has evolved over time, adopting more formal and advanced methods, including computerized assessments and statistical analysis. Trained tasting panels play a central role in identifying sensory attributes, allowing adjustments in formulations to meet consumer preferences. Furthermore, sensory analysis plays an essential role in quality assurance by detecting undesirable variations and contributing to the development of new products. Data analysis and machine learning algorithms, in turn, play a crucial role in characterizing and predicting beer flavors and profiles. This undergraduate thesis aimed to present a brief literature review on the subject and also aimed to provide a brief case study on beer characterization. Data analysis techniques and regression using the k-nearest neighbors (KNN) model were applied. The results showed that despite its simplicity, KNN is capable of predicting the alcohol content of various types of beers, representing with MSE of 1.1 the test group of data.