A New Multi‐Objective Ensemble System for Geoscience Resource Price Forecasting: Evidence from the Green Metal Market

ABSTRACT

Green metals are an important type of geoscience resource and also essential components for building a clean energy future. However, their substantial price fluctuations make accurate price forecasting a crucial decision‐making enabler for energy transitions. Previous studies have paid limited attention to the adaptive selection of effective subseries‐level predictors. Consequently, this paper develops a novel multi‐objective ensemble system with an adaptive selection mechanism for geoscience resource market management, called DPBPAPSME. Four modules are designed in DPBPAPSME. Specifically, the data preprocessing module reduces the negative impact of noise and modeling complexity of the original time series. The basic predictor module introduces four different artificial intelligence models to overcome the shortcomings of using a specific predictor. The adaptive predictor selection module overcomes the limitations of traditional subjective and objective indicator selection, adaptively identifies effective predictors, and improves the prediction precision and robustness. Furthermore, a multi‐objective ensemble module optimizes and adjusts the weight coefficients of the predictors of the selected subseries and fully integrates the complementary forecasting information provided by the selected predictors. Three green metals, platinum, zinc, and copper, were used in the main experiments, and an additional gold dataset was employed to examine the generalizability of the proposed system. The mean absolute percentage errors of DPBPAPSME on the three datasets were 0.161292%, 0.143030%, and 0.091245%, respectively. The results indicate that the proposed DPBPAPSME system is a promising tool for the management of the geoscience resource market.