<p>Accurate prediction of copper futures prices is crucial due to copper’s significant role in global markets, influenced by various economic, geopolitical, and supply-demand factors. Traditional forecasting models often fail to capture the complex, non-linear, and non-stationary nature of copper price time series. This paper proposes a novel intelligent copper futures prediction model based on multi-feature fusion and multi-stage optimization. The proposed model demonstrates high prediction accuracy and stability, effectively capturing the nonlinear dynamics of copper futures prices. The model first uses Advanced Variational Mode Decomposition (AVMD) to decompose the copper price time series into intrinsic modes, uncovering underlying patterns. Feature Extraction (FE) is then performed on these decomposed sequences to extract critical information. A Long Short-Term Memory (LSTM) network, optimized by Grey Wolf Optimization (GWO), is employed to predict future values, capturing temporal dependencies while minimizing overfitting risks. Additionally, an ensemble of Random Forest (RF) models, also optimized by GWO, combines the LSTM forecasts to enhance prediction precision. Empirical results demonstrate that this model outperforms traditional methods in both accuracy and stability, offering valuable insights for copper futures forecasting and market analysis.</p>