Nickel price forecasting based onempirical mode decomposition and deep learning model with expansion mechanism

As a critical material for stainless steel production, electric vehicle (EV) batteries, and advanced technology alloys, nickel plays a pivotal role in the global energy transition, with its strategic value becoming increasingly evident. This study presents a novel hybrid forecasting framework that combines Ensemble Empirical Mode Decomposition (EEMD) with a Dilated Long Short-Term Memory (Dilated LSTM) network to address the high uncertainty and complexity of nickel price fluctuations. By leveraging EEMD for multi-scale decomposition and Dilated LSTM for advanced temporal feature extraction, the proposed EEMD-DilatedLSTM model is designed to enhance predictive precision across different time horizons.Empirical results demonstrate that the proposed model outperforms benchmark algorithms in both short-term and medium-to-long-term prediction of nickel futures prices. Ablation studies validate the effectiveness of the hybrid architecture, and interpretability analyses highlight the decisive influence of low-frequency components in medium-to-long-term forecasting. Additionally, the Shapley value of copper price fluctuations is identified as a key driver, emphasizing its transmission effect on nickel prices.This study provides a robust methodological framework for strategic metal price forecasting, offering valuable insights for risk management in resource-driven enterprises and informing evidence-based industrial policy design by governments.