Price predictions of scrap steel for north China via machine learning

Historically, projections of the values of different commodities have been relied upon by governments and investors alike. This research examines the difficult task of estimating scrap steel prices, which are released daily for the north China market, utilizing data spanning the time period of 08/23/2013–04/15/2021. Predictions of this essential commodity price signal have not received adequate consideration in previous studies. Price predictions are produced here through Gaussian process regression approaches that are constructed via cross-validation procedures with Bayesian optimization techniques. With a 0.1325% relative root mean square error, the models produce rather accurate projections of prices throughout the out-of-sample testing timeframe spanning 09/17/2019–04/15/2021. Governments and investors might utilize price research models here to build informed judgments in the regional scrap steel market.