Developing neural networks to forecast agricultural commodity prices
Jason Snyder, J. Sweat, Matthew P. Richardson, Douglas Craig Pattie · 1992
The paper evaluates neural networks as a univariate forecasting tool for two agricultural price series: weekly closing prices for live cattle and daily settlement prices for corn. Performance was evaluated using root mean squared error and mean absolute percentage error. Neural networks outperformed the best traditional method for cattle price forecasts made four, eight, and twelve weeks into the future, and for corn made ten trading days into the future.>