FORECASTS OF WHOLESALE SOYBEAN OIL PRICE INDICES VIA GAUSSIAN PROCESS REGRESSIONS

Forecasting the prices of agricultural commodities has always been an important task for regulators and investors. The weekly price prediction problem for wholesale soybean oil in the Chinese market from January 1, 2010 to January 3, 2020 is examined in this study, with data sourced from China’s National Wholesale Price Information System, since price forecasting for this important commodity price measure has received insufficient attention in the literature. Our study is facilitated by Gaussian process regressions, and cross-validation and Bayesian optimizations are used in the model training processes. The developed models accurately forecasted the price index with an out-of-sample relative root-mean-square error of 0.4253% between January 5, 2018 and January 3, 2020 as the testing period and led to higher accuracy as compared to several benchmark models, including the support vector regression, random forest, and autoregressive models. The created models might be applied to the policy analysis and decision-making processes of investors and policymakers. The forecasting findings might be helpful in creating comparable commodity price indices as they offer reference data on the price trends suggested by the models.