Previsão do preço futuro de commodities agrícolas: um estudo para enriquecer séries temporais
Keywords: agribusiness commodities, prices forecasting, machine learning, time series
AbstractProducts derived from corn and soy are consumed on a large scale in the world. Market price fluctuations have far-reaching effects on grain consumers, criteria and indices. Thus, the forecast of future grain and grain prices has attracted the attention of investments and agribusiness companies. Estimation, forecasting models use time series to predict future values. However, external factors can originate the data in time series, such as political events, improvement patterns and the foreign exchange market. This information is not explicit in time series data and can make it difficult to predict variable values. Textual data extracted from news, forums and social media can be a source of knowledge about external factors and potentially useful for weather forecasting models. Some studies present text mining techniques to combine textual data with time series. However, existing representations have limitations, such as the curse of dimensionality and ineffective attributes. In this sense, this work proposes representations of time series enriched with textual information. The results indicate that the methods used can be an alternative to improve the prediction performance in regression tasks.