Abstract
Accurate forecasting of gold prices is crucial for financial decision-making in various sectors, including investment and mining. This study introduces a multi-objective optimization framework that utilizes the Pareto alpha-cut technique to evaluate and enhance forecasting models for gold prices. We employed three distinct models: the Autoregressive Distributed Lag (ARDL) model, a stochastic model, and the Autoregressive Integrated Moving Average (ARIMA) model, to capture the underlying dynamics of gold price fluctuations influenced by macroeconomic factors. The methodology incorporates the Pareto optimality approach combined with fuzzy logic to manage trade-offs among multiple performance metrics, specifically Root Mean Squared Error (RMSE), volatility, and R-squared. By applying the alpha-cut technique, we filtered out less optimal models, retaining only those that met a predefined level of acceptability across all criteria. Results indicate that the ARDL model consistently outperformed the others, achieving superior accuracy and fit, while the stochastic model exhibited robust stability. This framework not only facilitates the identification of Pareto optimal models but also provides valuable insights into the balance between accuracy and stability in gold price forecasting. The findings contribute to a deeper understanding of forecasting methodologies and highlight the practical implications for stakeholders in the financial and commodity sectors.•This study introduces a multi-objective optimization framework leveraging the Pareto alpha-cut technique.•Compared with ARDL, ARIMA and Stochastic mode•The validity analysis confirms the accuracy and stability of gold price forecasting.