Enterprise AI Analysis

Strategic Copper Price Forecasting

Accurate copper price forecasting is paramount due to its inherent volatility and critical economic impact. This study specifically addresses these challenges by applying a novel hybrid interval prediction framework. The framework integrates multi-objective optimization with quantile deep learning to enhance prediction accuracy and robustness.

The process begins with a feature selection module using four distinct methods (PCC, MIC, LASSO, and RF) to identify the most influential factors, such as crude oil, gold, iron ore, nickel prices, and key exchange rates (USD-CNY, USD-AUD, etc.). This ensures the model focuses on core economic drivers of copper prices.

Next, Quantile Regression (QR) deep learning models like QRLSTM are employed to generate initial prediction intervals, providing a probabilistic range rather than a single point estimate. This captures the inherent uncertainty of market fluctuations.

Finally, multi-objective optimization algorithms (MOMVO, MOALO, MODA, MOSSA) are applied to fine-tune these intervals, balancing reliability (coverage) and resolution (width). The MOSSA-QRLSTM model demonstrated superior performance, achieving a high Prediction Interval Coverage Probability (PICP) of 94.5205% and a minimal Prediction Interval Normalized Average Width (PINAW) of 0.0066 at 95% confidence levels. This robust framework provides reliable and comprehensive forecasts, crucial for strategic decision-making in industries affected by copper price dynamics.