Performance Comparison of Gold Price Forecasting Models and Effectiveness of Hybrid Methods

Honghao Dan

doi:10.54097/0bef2168

Authors

Honghao Dan Shenzhen Audencia Financial Technology institute, Shenzhen University, Shenzhen, China

DOI:

https://doi.org/10.54097/0bef2168

Keywords:

Gold price prediction; Deep learning; Long Short-Term Memory; Generalized Autoregressive Conditional Heteroskedasticity; Hybrid model.

Abstract

Gold has long been regarded as an important risk-averse asset and an investment tool. Predicting its price trend has been a challenging topic for financial time series analysis. This study constructs six models, including artificial neural network (ANN), Long Short-Term memory (LSTM), bidirectional long Short-Term memory (Bi-LSTM) and their different hybrid strategies with generalized autoregressive conditional heteroscedastic (GARCH) model in gold price forecasting, and systematically compares their performance, based on the daily data between 2015/10/10 and 2025/10/10. Through multiple rounds of random seed experiments, the performance of each model is evaluated by the Root Mean Square error (RMSE), Mean Absolute error (MAE), Mean Absolute percentage error (MAPE), R²and trend prediction accuracy. The results indicate that the ANN model is the most robust in comprehensive performance. The further analysis shows that simply introducing GARCH volatility in a linear way or using bi-directional structure (Bi-LSTM) may reduce the stability of the model. However, adopting the non-linear fusion method by combining the point prediction from the LSTM and the volatility features from GARCH (LSTM-ANN) shows potential in fitting volatility and trend. The study reveals the characteristics of different models in gold price prediction, provides a basis for model selection, and points out that the introduction of external macroeconomic and news factors is a key direction for improving forecasting performance in the future.

Downloads

Download data is not yet available.

References

[1] Ramadhan T, Wahanani H E, Aditiawan F P. Gold Price Prediction Using Hybrid Deep Learning by Integrating LSTM-ANN Network with GARCH Model. Literasi Nusantara, 2025, 5(3): 315-328.

[2] Box G E P, Jenkins G M. Time Series Analysis: Forecasting and Control. Holden-Day, 1976.

[3] Engle R F. Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation. Econometrica, 1982, 50(4): 987-1008.

[4] Bollerslev T. Generalized Autoregressive Conditional Heteroskedasticity. Journal of Econometrics, 1986, 31(3): 307-327.

[5] Mombeini H, Yazdani-Chamzini A. Modeling Gold Price via Artificial Neural Network. Journal of Economics, Business and Management, 2015, 3(7): 699-703.

[6] Hochreiter S, Schmidhuber J. Long Short-Term Memory. Neural Computation, 1997, 9(8): 1735-1780.

[7] Yurtsever M. Gold Price Forecasting Using LSTM, Bi-LSTM and GRU. European Journal of Science and Technology, (31), 341-347. 2021.

[8] Santika I W K G, Saadah S, & Yunanto P E. Gold price prediction using convolutional neural network-long short-term memory (CNN-LSTM). Kinetik: Game Technology, Information System, Computer Network, Computing, Electronics, and Control, 6(3), 231-238. 2021.

[9] Risse M. Combining wavelet decomposition with machine learning to forecast gold returns. International Journal of Forecasting, 35(2), 601-615. 2019.

[10] Bao W, Yue J, & Rao Y. A deep learning framework for financial time series using stacked autoencoders and long-short term memory. PLOS ONE, 12(7), e0180944. 2017.

[11] Singh N, & Varshney J. Artificial Intelligence-Driven Model for Gold Price Prediction. International Journal of Scientific Research in Engineering and Management, 8(5). 2024.

[12] García-Medina A, Aguayo-Moreno E. LSTM-GARCH hybrid model for the prediction of volatility in cryptocurrency portfolios. Computational Economics, 2024, 63(1): 1511–1542.

[13] Kian Poor S., Fattahi S., Hajian M. Gold Price Forecasting: A Novel Approach Based on Text Mining and Big-Data-Driven Model. Studies in Business and Economics. 2024, 19(3): 157-171.

[14] Yao Yinhong, Wang Xiaoxu, Chen Wei, Chen Zhensong. Extreme Risk Spillover among Global Stock Markets Based on Transformer-LSTM Quantile Regression. Chinese Journal of Management Science. 2025, 33(8): 1-13.