The prediction of gold price through oil price and bitcoin price using machine learning method

Abstract

This research focuses on using a machine learning model to predict the price of gold, using the price of oil and Bitcoin as independent variables. Gold is often considered a safe-haven asset, but in recent years, its relationship with other volatile assets such as cryptocurrencies and commodities has become more complex. Historical daily data from January 10, 2012 to April 10, 2025 was used as input for the models. The research applies two different models such as: XGBoost, and LSTM to evaluate which approach delivers the most accurate predictions. After standard preprocessing and restructuring the data into a supervised learning format, each model was trained and assessed using Mean Squared Error (MSE). The results show that XGBoost performs best in terms of both accuracy and training efficiency. Although linear analysis suggests that Bitcoin has little effect on gold prices, nonlinear machine learning models reveal that it contributes much more to price prediction than expected. These findings highlight the importance of capturing hidden nonlinear patterns in cross-asset dynamics. The study contributes to the growing field of financial time series forecasting by showing how data-driven techniques can enhance short-term prediction in volatile markets. While the model focuses on oil and Bitcoin prices, future research could incorporate macroeconomic variables such as interest rates, inflation, or geopolitical risks to further improve predictive performance and interpretability.