Machine learning predictions of regional steel price indices for east China

Abstract

From 1 January 2010 to 15 April 2021, this study examines the challenging task of daily regional steel price index forecasting in the east Chinese market. We train our models using cross-validation and Bayesian optimisations implemented through the expected improvement per second plus algorithm, and utilise Gaussian process regressions to validate our findings. Investigated parameters as part of model training involve predictor standardisation status, basis functions, kernels and standard deviation of noises. The models that were built accurately predicted the price indices between 8 January 2019 and 15 April 2021, with an out-of-sample relative root mean square error of 0.57%, root mean square error of 0.84, mean absolute error of 0.48, and correlation coefficient of 99.81%.