Multi-objective optimization of gold price forecasting using the pareto alpha-cut technique

Area of Science:

• Quantitative Finance
• Econometrics
• Operations Research

Background:

• Accurate gold price forecasting is vital for investment, mining, and financial planning.
• Existing forecasting models face challenges in balancing multiple performance metrics.
• Macroeconomic factors significantly influence gold price dynamics.

Purpose of the Study:

• To introduce a novel multi-objective optimization framework for evaluating gold price forecasting models.
• To enhance model selection by managing trade-offs between accuracy, volatility, and fit.
• To identify Pareto optimal forecasting models using the Pareto alpha-cut technique.

Main Methods:

• Utilized Autoregressive Distributed Lag (ARDL), Stochastic, and Autoregressive Integrated Moving Average (ARIMA) models.
• Applied Pareto optimality principles combined with fuzzy logic for multi-criteria decision-making.
• Employed the Pareto alpha-cut technique to filter and select optimal models based on RMSE, volatility, and R-squared.

Main Results:

• The Autoregressive Distributed Lag (ARDL) model consistently showed superior accuracy and model fit.
• The stochastic model demonstrated robust stability in forecasting gold prices.
• The Pareto alpha-cut framework effectively identified models balancing accuracy and stability.

Conclusions:

• The proposed framework provides a robust method for selecting superior gold price forecasting models.
• Findings offer practical insights for financial stakeholders managing investment and commodity risks.
• This approach enhances understanding of forecasting model performance under multiple objectives.