Steel price index forecasts through machine learning for northwest China

Commodity price projections have long been relied upon by investors and governments in general. This study examines the complex issue of predicting steel price indices reported daily for the regional market of northwest China, utilizing data from 01/01/2010 to 04/15/2021. Predictions of this significant commodity price indication have attracted scant attention in the literature. Forecasts are generated via Gaussian process regression models that are estimated through cross-validation and Bayesian optimizations. The built models successfully forecast price indices for the out-of-sample period from 01/08/2019 to 04/15/2021, with a relative root mean square error of 0.3899%. Investors and governments may use models constructed to investigate pricing and make informed judgments. When reference data on trends of prices offered by the models are used, predicting results may contribute to establishment of comparable commodity price indices.