Wholesale price forecasts via neural networks for soybeans and white wheat

Commodity price forecasts have always been a vital issue to a wide spectrum of market participants. These problems for soybeans and wheat are of no exception, particularly considering their significant roles as strategic food resources. In this work, we conduct forecast exercises for weekly wholesale price indices of soybeans and white wheat in the Chinese market during January 1, 2010 – January 3, 2020 through explorations of non-linear auto-regressive neural networks as forecast models. We study forecast performance associated with a variety of settings over algorithms for model estimations, numbers of hidden neurons and delays, and ratios for splitting the data. We arrive at models that are relatively simple and generate forecasts of high accuracy and stabilities. Specifically, we obtain relative root mean square errors of 0.670%, 0.648%, and 0.684% for training, validation, and testing, respectively, for the case of white wheat and those of 1.930%, 1.605%, and 1.704% for the case of soybeans. We show empirically that neural networks are useful tools for price forecasts of these two important agricultural commodities. Our results could, on one hand, serve as standalone technical forecasts. They could, on the other hand, be combined with other (fundamental) forecast results for forming perspectives of price trends and carrying out policy analysis.