Soybean and Soybean Oil Price Forecasting through the Nonlinear Autoregressive Neural Network (NARNN) and NARNN with Exogenous Inputs (NARNN–X)

Abstract

• We develop nonlinear autoregressive neural network (with exogenous inputs) models for daily agricultural commodity price forecasting. • We explore the forecasting issue for soybeans and soybean oil over a period of greater than fifty years. • We construct simple, accurate, and stable models for the two agricultural commodities of economic significance. • The models are useful as technical forecasting tools and for policy analysis. Price forecasting is a key concern for market participants in the agriculture sector. This study explores usefulness of the nonlinear autoregressive neural network (NARNN) and NARNN with exogenous inputs (NARNN–X) for forecasting issues in data sets of daily prices over periods of greater than fifty years for soybeans and soybean oil. The exogenous input in the NARNN–X for prices of soybeans (soybean oil) are prices of soybean oil (soybeans). Through investigating various model settings across the algorithm, delay, hidden neuron, and data splitting ratio, models resulting in accurate and stable performance for these two commodities are arrived at. The overall relative root mean square errors based on the chosen NARNNs (NARNN–Xs) are 1.701% (1.695%) and 1.777% (1.775%) for soybeans and soybean oil, respectively. Usefulness of the machine learning approach for price forecasting issues of the two commodities is demonstrated, as well as potential usefulness of prices of closely related commodities for providing predictive content. Results here could be used on a standalone basis as technical forecasts or combined with fundamental forecasts for forming perspectives of price trends and conducting policy analysis. The empirical framework here should not be difficult to implement, which is an important consideration to many decision makers, and has potential to be generalized for forecasting prices of other commodities.