Hybrid Metaheuristic Optimization of Support Vector Regression and Prophet Models for Short-Term Gold Price Forecasting

Abstract

Accurate forecasting of gold prices is of critical importance for investors, financial institutions, and policy makers due to gold’s role as a hedge against inflation, market uncertainty, and financial instability. However, gold price dynamics are characterized by nonlinear behavior, regime shifts, and time-varying volatility, which limit the effectiveness of traditional linear forecasting models. This study proposes a hybrid forecasting framework that integrates Support Vector Regression (SVR) and the Prophet model with three metaheuristic optimization algorithms, namely the Bat Algorithm (BAT), Grey Wolf Optimizer (GWO), and Whale Optimization Algorithm (WOA), to enhance predictive performance through automated hyperparametertuning. Daily gold price data spanning from 1 January 2025 to 25 January 2026 are analyzed using a strict time-ordered train–validation–test protocol to ensure robust out-of-sample evaluation. Model performance is assessed using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and information-theoretic criteria (AIC and BIC). The empirical results indicate that the SVR model optimized by GWO achieves the lowest forecasting error and the most favorable accuracy–complexity trade-off among all configurations. Based on this selection, a 30-day ahead forecast is generated with 95% confidence intervals, revealing a moderate short-term correction in gold prices following a strong upward trend. The findings demonstrate that metaheuristic optimization substantially improves model calibration and that the SVR–GWO framework provides a robust and reliable tool for short-term gold price forecasting with uncertainty quantification.