Predicting the Trading Volume of the Thermal Coal Futures Through Gaussian Process Regressions
Bingzi Jin, Xiaojie Xu
Abstract
For many market players, including policy makers, estimating the trading volumes of commodity futures is a critical task. Using daily data from January 2016 to December 2020, we investigate the trading volume forecast problem for the thermal coal futures traded on the Zhengzhou Commodity Exchange in China, focusing on the energy sector in this study. We use the Gaussian process regression for this purpose, and we build the models using cross-validation and Bayesian optimizations. The trading volume from the out-of-sample period of January 2, 2020–December 31, 2020 is successfully predicted by the created models with a low prediction error and high prediction correlation. Here, the thermal coal trade volume forecast issue is demonstrated to benefit from the application of Gaussian process regressions. The projection’s results might be applied as stand-alone technical forecasts or in combination with other forecasts for policy research that involves formulating opinions about trade trends.