Enhancing Commodity Futures Price Prediction With Geopolitical Risk Embedding: A Comparative Study of Deep Learning Models

ABSTRACT This study uses daily closing prices of nine Chinese commodity futures from 2015 to 2023 to analyze price fluctuations and improve prediction reliability. It compares traditional time series model (ARIMAX), benchmark deep learning models (LSTM, GRU), and generative adversarial networks (GAN, WGAN), while also exploring the impact of geopolitical risk (GPR). The results show that deep learning models outperform traditional methods. LSTM and GRU excel at capturing temporal features, while WGAN offers superior versatility and stability, addressing GAN prediction flaws. Including GPR enhances forecasting accuracy for most commodities, revealing a dynamic correlation between GPR and commodity prices, with significant variation across different commodities. This study provides empirical evidence for the use of deep learning in financial time series forecasting and highlights the role of geopolitical risks in futures markets.