Forecasting future corn and soybean prices: an analysis of the use of textual information to enrich time-series.

The commodities corn and soybean are products consumed on a large scale in the world. Fluctuations in market prices have far-reaching effects on consumers, farmers, and grain processors. Thus, forecasting the prices of these grains has attracted significant attention from researchers. Forecasting models generally use quantitative time-series data. However, external qualitative factors can influence data in time-series, such as political events, economic crises, and the foreign exchange market. This information is not explicit in the time-series data, and these factors can influence the prediction of the variable values. Textual data extracted from news, forums, and social networks can be a source of knowledge about external factors and potentially useful for time-series forecasting models. Some studies present text mining techniques to combine textual data with time-series. However, the existing representations have some limitations, such as the curse of dimensionality and ineffective attributes. This work applies pre-processing methods in time-series and uses representations combined with textual data to predict the future price of corn and soybeans. The results indicate that the methods used can be an alternative to improve forecasting performance in regression tasks.