Analysis and forecasting of daily global gold price: an SARIMA-LSTM approach with Random Forest technique

Abstract

Forecasting gold prices remains vital in financial markets, given gold’s dual role as both a hedge against inflation and a safe-haven asset during economic uncertainty. This study proposes a hybrid model integrating SARIMA, LSTM, and RF to improve predictive accuracy by capturing both linear and nonlinear dependencies in historical gold price data. SARIMA models linear trends and seasonal components, LSTM captures nonlinear patterns from SARIMA residuals, and RF refines predictions using macroeconomic indicators such as the USD Index, Federal Interest Rate, US CPI, Oil Prices, S&P 500 Index, and Bond Yields. Utilizing real-world data, the model effectively tracks market trends with reduced forecasting errors, indicating continued price fluctuations and potential long-term growth. The findings provide valuable insights for investors and policymakers, with future research focusing on additional macroeconomic factors and advanced hybrid forecasting techniques.