LSTM Networks Outperform Traditional Models in Gold Price Forecasting

A new study proposes a multivariate deep learning framework based on stacked Long Short-Term Memory (LSTM) networks for short-term gold price forecasting. The model leverages multiple input features, including Open, High, Low, and Close (OHLC) prices, and employs a sliding window technique to capture temporal dependencies in time series data.

Experimental results on real-world gold price data demonstrate that the proposed model achieves high predictive performance, with a Mean Absolute Percentage Error (MAPE) of 2.34% and a coefficient of determination (R²) of 0.9588. The LSTM model significantly outperforms traditional Recurrent Neural Networks (RNNs) in terms of stability, accuracy, and robustness.

The study highlights the effectiveness of LSTM-based architectures for financial time series forecasting and their potential for practical applications in investment analysis and decision support systems.