Work

Revealing the correlation between economic indicators and gold prices for forecasting: Medium term forecast framework with data patch

Year: 2025

Type: article

Abstract: Predicting gold prices through the analysis of key economic indicators such as inflation rates, Government Bond Yields, and the U.S. Dollar Index, alongside historical Gold Prices, is crucial for enabling investors to better understand market dynamics and make vital decisions to maximize returns. However, previous studies have faced challenges in extracting hidden factors related to gold price prediction from diverse economic indicators, and the comprehensive exploration of gold price data is yet to be fully achieved. To address this, the present study introduces a mid to long-term gold price prediction model named DPformer. This model utilizes a patching strategy to investigate the relationships between different economic indicators and Gold Prices. It also employs a decomposition approach to discover the mid to long-term trend characteristics and yearly seasonal patterns of Gold Prices. The core of the model integrates a Transformer module, which is solely based on an Encoder structure, and enhances it with multiple attention mechanisms and convolutions. This enhancement allows the improved Transformer model to more effectively capture the long-term dependencies of Gold Prices. The empirical results demonstrate that DPformer consistently outperforms a suite of advanced models widely adopted in terms of mid to long-term forecasting accuracy, including LSTM, GRU, Transformer, DLinear, and PatchTST. Notably, for the 30-step gold price prediction task, DPformer achieves a 21.78 % reduction in Mean Squared Error compared to PatchTST. Moreover, by quantitatively analyzing how various economic indicators influence gold price forecasts, this study provides substantial support for investors in making informed decisions at critical moments.

Source: Expert Systems with Applications

DOI: 10.1016/j.eswa.2025.129594

Authors: Guanhao Bao, Yunbo Niu, Baisheng Cui, Wanying Ji

Institutions: George Washington University, Chongqing University, Shanghai Normal University, Southwestern University of Finance and Economics

Language: English

Topic: Stock Market Forecasting Methods

Subfield: Management Science and Operations Research

Field: Decision Sciences

Domain: Social Sciences

Open Access status: hybrid

Estimated APC paid: $3,500

Funder: National Social Science Fund of China