Previsão do preço do frango e do suíno vivos no estado do Paraná: uma comparação entre os modelos Prophet e NeuralProphet

Abstract

The poultry and swine production chains generate great wealth for the Brazilian economy. In this context, this work aims to apply models, based on the Prophet and NeuralProphet tools, to forecast the prices of live chickens and swine (R$/kg) and, thus, assist small and medium producers in the organization, production and negotiation of their products. The database, made available by the Institute of Applied Economic Research (IPEA), presents, in the state of Paraná, historical series of the average prices of chickens and live swine (R$/kg), in the period between January/2007 and June/2024. Forecasting models, based on the Prophet and NeuralProphet tools, were implemented in the Python language. Results obtained from the two models were compared through the Mean Absolute Percentage Error (MAPE), the Root Mean Square Error (RMSE) and the Mean Absolute Error (MAE). It was found, for a 12-month horizon, that the multivariate NeuralProphet model obtained the best forecasting performance in predicting the average prices of live chickens and pigs in the state of Paraná.