Forecasting gold price using machine learning methodologies

Abstract

This study investigates the potential of advanced Machine Learning (ML) methodologies to predict fluctuations in the price of gold. The study employs data from leading global stock indices, the S&P500 VIX volatility index, major commodity futures, and 10-year bond yields from the US, Germany, France, and Japan. Lagged values of these features up to 10 previous days are also used. Four machine learning models are used: Random Forest, Gradient Boosted Regression Trees (GBRT), and Extreme Gradient Boosting (XGBoost), to forecast future gold prices. The study finds that the most influential stocks indices for prediction are one-day lagged data of ASX, S&P500, TA35, IBEX, and AEX, as well as U.S. and Japan bonds yields and delayed data of gas and silver. Furthermore, the study's models identify that one-day lagged VIX score and our VIX dummy variable have a significant impact on gold price, indicating that economic uncertainty affects gold prices. The results suggest that incorporating various financial indicators and moving averages can be a powerful tool for predicting future gold prices. GBRT and XGBoost can be valuable models for making informed decisions about gold investments.