Probabilistic Forecasting of Crude Oil Prices Using Conditional Generative Adversarial Network Model with Lévy Process

Abstract

Accurate crude oil price forecasting is essential, considering oil’s critical role in the global economy. However, the crude oil market is significantly influenced by external, transient events, posing challenges in capturing price fluctuations’ complex dynamics and uncertainties. Traditional time series forecasting models, such as ARIMA and LSTM, often rely on assumptions regarding data structure, limiting their flexibility to estimate volatility or account for external shocks effectively. Recent research highlights Generative Adversarial Networks (GANs) as a promising alternative approach for capturing intricate patterns in time series data, leveraging the adversarial learning framework. This paper introduces a Crude Oil-Driven Conditional GAN (CO-CGAN), a hybrid model for enhancing crude oil price forecasting by combining advanced AI frameworks (GANs), oil market sentiment analysis, and stochastic jump-diffusion models. By employing conditional supervised training, the inherent structure of the data distribution is preserved, thereby enabling more accurate and reliable probabilistic price forecasts. Additionally, the CO-CGAN integrates a Lévy process and sentiment features to better account for uncertainties and price shocks in the crude oil market. Experimental evaluations on two real-world oil price datasets demonstrate the superior performance of the proposed model, achieving a Mean Squared Error (MSE) of 0.000054 and outperforming benchmark models.

Keywords

• crude oil-driven conditional generative adversarial network
• Lévy–Merton jump-diffusion model
• oil price forecasting
• long short-term memory