Machine Learning-Based Comparative Analysis of the Determinants of Gold Prices

In this study, various machine learning methods were compared to identify the economic factors influencing gold prices and to determine the most effective prediction model. The methods used include linear regression, multivariate adaptive regression splines (MARS), extreme gradient boosting (XGBoost), random forest, artificial neural networks (ANN), and the ensemble-based voting regressor. According to the test data results, the MARS model demonstrated the highest prediction accuracy, followed by the ANN model and the voting regressor model. The analysis of the three best-performing models revealed that the most influential factors on gold prices are silver prices, the BIST 100 Index, and the NASDAQ Index. Overall, machine learning approaches outperformed traditional models, with MARS providing the most reliable and accurate predictions for gold price forecasting.