<![CDATA[Gold price prediction plays a vital role in financial decision-making, particularly during periods of heightened market volatility when gold functions as a strategic hedge against inflation and economic uncertainty. This study examines the effectiveness of a Bidirectional Gated Recurrent Unit (BI-GRU) model enhanced with Monte Carlo Dropout for forecasting XAU/USD prices using key macroeconomic indicators, namely CPI, DXY, S&P 500, and crude oil prices, covering the period from May 6, 2015, to May 1, 2025. The research addresses the need for forecasting approaches capable of capturing nonlinear dependencies while simultaneously quantifying predictive uncertainty. The methodological workflow includes constructing a multivariate time-series dataset, performing comprehensive preprocessing, and evaluating multiple temporal window lengths and model configurations. Performance is assessed using MAE, RMSE, and R², with uncertainty estimation derived from repeated stochastic forward passes. Empirical analysis reveals strong correlations between gold prices and the S&P 500 (r ≈ 0.93) and CPI (r ≈ 0.89), affirming the substantial influence of macroeconomic conditions on gold dynamics. The optimal configuration, consisting of a 70:30 data split, a 60-day window, 128 BI-GRU units, and a 0.3 dropout rate, achieved an MAE of 0.0354, an RMSE of 0.044, and an R² of 0.9349, outperforming the baseline BI-GRU without dropout. Multi-step forecasting further shows that the model maintains stable predictive behavior during the initial horizon, with uncertainty increasing gradually in extended forecasts. These findings demonstrate that integrating BI-GRU with Monte Carlo Dropout provides a reliable uncertainty-aware framework that offers both accurate predictions and practical value for financial practitioners requiring risk-sensitive forecasting tools.]]>
