COMPARATIVE EVALUATION OF MACHINE LEARNING AND DEEP LEARNING FOR GOLD PRICE PREDICTION USING HISTORICAL DATA AND MACROECONOMIC INDICATORS

Abstract

Gold price forecasting is an important problem in financial market analysis because gold prices are influenced by historical price dynamics as well as various economic and financial market conditions. This study comparatively evaluates the performance of Random Forest, Extreme Gradient Boosting, Long Short-Term Memory, and Gated Recurrent Unit models for one-step-ahead gold price forecasting under two input scenarios: historical gold price data alone and historical data combined with macroeconomic indicators. The dataset covers the period 2016–2025, and model performance is evaluated using Mean Absolute Error, Root Mean Squared Error, Mean Absolute Percentage Error, and the coefficient of determination. The test results show that the naïve baseline achieves the lowest prediction errors, with an MAE of 25.9345, an RMSE of 37.3274, and a MAPE of 0.8684%. Adding macroeconomic indicators did not improve model performance. The substantial difference in price distributions between the training and test datasets also affected model performance, particularly for tree-based models.