Previsão do preço futuro do óleo de girassol: uma abordagem baseada no modelo CNN-Attention

Abstract

In recent years, the Attention Mechanism has been widely used in several areas of deep learning application. In this context, this work aims to propose the use of the CNN-Attention network to predict the price of sunflower oil. The database presents a monthly series of the price of sunflower oil in the period between January/1960 and June/2023, totaling 762 observations. Prediction models, based on CNN (Convolutional Neural Network), CNN-Attention and LSTM (Long Short Term Memory) Neural Networks were implemented in the Python language. Results obtained from the three models were compared using the metrics R2 (Coefficient of Determination), MAE (Mean Absolute Error), RSME (Root Mean Squared Error) and MAPE (Mean Absolute Percent Error). It was found, for a 13-month horizon, that the CNN-Attention model performed better.