Which uncertainty measure better predicts gold prices? : new evidence from a CNN-LSTM approach

Quantifying the influence of uncertainty on gold prices is significant for improving related financial decision making. This study proposes a novel CNN-LSTM neural network that can extract potential features from sample data to effectively predict gold prices. Specifically, we demonstrate various uncertainty measures containing market volatility information, such as the economic policy uncertainty index (EPU), epidemic stock market volatility index (IDEMV), and volatility index (VIX), which can contribute to the prediction of gold prices rather than relying solely on the history of tickers, which is conventionally used for prediction. In addition, the proposed model is evaluated against SVR and two different LSTM models. The empirical findings reveal that incorporating additional features, such as uncertainty measures, contributes to improving the predictive accuracy of the model. The CNN-LSTM model, with the inclusion of EPU, IDEMV, and both, achieves a high prediction accuracy. Additionally, the overall prediction accuracy of the CNN-LSTM model outperforms the other proposed methods. The findings provide profound insight into portfolio diversification and risk management practices for governments and businesses.