FORECASTING GOLD PRICE DIRECTION: MACHINE LEARNING VERSUS TECHNICAL INDICATORS

Аннотация

Accurate forecasting of gold price movements remains a challenging task because of the complex and nonlinear nature of financial markets. This study evaluates the effectiveness of traditional technical indicators and ensemble machine learning algorithms for predicting the daily direction of XAU/USD price movements. Historical daily gold price data covering the period from July 2023 to March 2026 were collected and transformed into predictor variables, including the Relative Strength Index (RSI), Moving Average Convergence Divergence (MACD), MACD Signal, Bollinger Bands, and calendar-based features. Traditional technical indicator models were first established as baseline benchmarks and subsequently compared with three ensemble learning algorithms: Random Forest, XGBoost, and Gradient Boosting. Model performance was evaluated using a five-fold TimeSeriesSplit cross-validation procedure together with an independent test dataset. The empirical results indicate that ensemble machine learning algorithms generally outperform traditional technical indicator models. Among the baseline models, Bollinger Bands achieved the highest classification accuracy (57.53%), whereas Random Forest produced the strongest overall performance among the machine learning models with an accuracy of 59.47%, followed by Gradient Boosting (55.94%) and XGBoost (54.55%). The cross-validation results further demonstrated the stability of the ensemble learning algorithms across different chronological validation periods. Overall, the findings suggest that machine learning provides a more effective approach than standalone technical indicators for forecasting the direction of gold price movements. The proposed framework contributes to the growing application of artificial intelligence in financial forecasting and offers practical decision-support insights for investors, traders, and financial analysts.