DEEP LEARNING-POWERED FORECASTING OF CRUDE OIL PRICES: CAPTURING NONLINEAR MARKET DYNAMICS

Abstract

Crude oil price forecasting is of paramount importance for various stakeholders, including policymakers, investors, and businesses. Understanding and predicting crude oil prices help in making informed decisions, managing risks, and planning economic strategies. The history of crude oil price forecasting dates back to the early days of oil exploration and production. Traditional methods for crude oil price forecasting often rely on linear models, which assume a constant relationship between variables. Linear regression, autoregressive integrated moving average, and exponential smoothing are commonly used techniques. While these methods work well for stable and linear relationships, they fail to capture the intricate patterns and nonlinearities present in crude oil price data, especially when influenced by a multitude of factors, including geopolitical tensions, natural disasters, economic indicators, technological advancements, and government policies. These factors interact in complex and nonlinear ways, making it challenging to capture their impact using linear models. Nonlinear dynamics modeling is essential because it allows researchers and analysts to explore the intricate relationships between these factors and crude oil prices, considering the nonlinear patterns and dependencies that exist in the data. Therefore, given historical data on crude oil prices and various influencing factors, this research aims to develop a predictive model that can capture the nonlinear relationships and patterns in the data. The goal is to accurately forecast future crude oil prices under different scenarios, considering the dynamic interplay of geopolitical events, market sentiment, economic indicators, and supply-demand dynamics. The proposed model is capable of adapting to changing market conditions and provide reliable forecasts, helping stakeholders make informed decisions and mitigate risks associated with crude oil price fluctuations