A novel hybrid interval prediction framework integrating multiobjective optimization and quantile deep learning for...

Area of Science:

• Financial forecasting
• Econometrics
• Computational finance

Background:

• Accurate copper price forecasting is vital but challenging due to market volatility.
• Traditional single-factor methods lack robustness and fail to account for multiple influencing factors.

Purpose of the Study:

• To develop a novel hybrid interval prediction framework for copper price forecasting.
• To enhance prediction accuracy and robustness by incorporating multiple variables and advanced optimization techniques.

Main Methods:

• A hybrid framework combining quantile deep learning and multi-objective optimization was developed.
• Four probabilistic forecasting algorithms and four multi-objective optimization algorithms were employed.
• Feature selection methods were used to identify crucial variables for prediction.

Main Results:

• The Quantile Regression Long Short-Term Memory (QRLSTM) model, optimized with the Multi-Objective Salp Swarm Algorithm (MOSSA), demonstrated superior performance.
• Achieved a Prediction Interval Coverage Probability of 94.5205% and a Prediction Interval Normalized Average Width of 0.0066 at 95% confidence levels.
• The model yielded an Average Interval Score of -373.9687, indicating high prediction precision.

Conclusions:

• The proposed probabilistic forecasting framework is reliable and comprehensive for copper price prediction.
• The hybrid approach effectively addresses the limitations of single-factor prediction methods.
• This research offers a robust solution for navigating the complexities of financial market fluctuations.