Multi-step metal prices forecasting based on a data preprocessing method and an optimized extreme learning machine by marine predators algorithm

The prediction of metal prices can provide references of future investment and decision-making for mining projects and affiliates. However, the improvement of metal prices prediction accuracy is still a highly challenging task owing to the complex influencing factors and nonlinear features. Most previous models usually overlook the significances of model optimization and data preconditioning, and often fail to obtain satisfactory results. Therefore, this study designed a hybrid model for multi-step copper and gold prices prediction. First, a robust data processing algorithm was introduced to decompose time series into several modes, which can effectively identify and mine the main characteristics of metal prices time series. Second, to enhance the prediction performance and conquer the limitations of individual models, nature-inspired metaheuristic algorithms are adopted to tune the parameters of these forecasting models. Third, two metal price data sets, hypothesis testing, seven model performance evaluation indices and several experiments are used to comprehensively compare and analyze the prediction ability of the proposed model. Finally, the results revealed that the developed hybrid model can not only significantly outperform the comparison models considered in this study, also offer effective and valuable suggestions for the related policymakers, investors and metal production and trading enterprises in metal market.