Use of artificial neural networks to predict the stock prices of meat and derivatives companies listed on B3 after the COVID-19 pandemic/Uso de redes neurais artificiais para prever os valores das acoes das empresas de carnes e derivados listadas na B3 apos a pandemia de COVID-19/Uso de redes neuronales artificiales para predecir los precios de las acciones de las empresas de carne y derivados cotizadas en B3 despues de la pandemia de COVID- 19

Artificial Neural Networks (ANNs) are useful for prediction with a relevant degree of accuracy, utilizing artificial intelligence for both linear and nonlinear models. This study focused on predicting the stock values of meat and derivatives companies listed on B3 using historical data series of these stocks and economic variables that may influence their valuation. Examining data from four companies from January 2022 to December 2023 and employing a Multilayer Perceptron ANN, it was trained to forecast stock values using the SELIC rate, Dow Jones Industrial Average (DJIA), Consumer Price Index (CPI), and Bovespa Index as independent variables. During the ANN training, the Mean Percentage Error was 12.7%, with the lowest error found in Mafrig Global Foods S/A (6.6%) and the highest in Minerva S/A. The SELIC rate, at 38.2%, contributed the most to the model construction, while the Bovespa Index, at 12.5%, had the least contribution to the RNA estimation. Once trained and tested, the ANN was used to predict stock values for the four companies on January 7, 14, 21, and 28, 2024. Compared with observed values on these days, it showed an excellent average error of 2.1%. Further studies are suggested, utilizing traditional prediction models to assess the predictive capability of ANNs alongside such models.