- Collect data
Soybean historical data is from Kaggle https://www.kaggle.com/motorcity/soybean-price-factor-data-19622018. This dataset includes not only the soybean future daily price (1962-2018) but also the associated commodities price as well, like gold, oil, soyoil, meat, etc. The size of dataset contains 20K rows and 47 columns.
-
Feature engineering Based on the original features, I use the lag features and rolling average on past prices for new features to help soybean_settle price prediction.
-
Modeling With two different approaches, the first part is using regression model and the other one is adopting time-series model.
-
Performance