Abstract—The crude oil market is distinguished by significant volatility, primarily due to its responsiveness to economic and geopolitical events and the intricate factors that drive price fluctuations. Accurate forecasting of crude oil prices is essential for mitigating adverse impacts on national economic stability and growth. This study examines the prediction of Brent crude oil prices utilizing a 20-year time series dataset encompassing 2004 to 2024. A hybrid modelling framework integrates Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models as feature extractors, followed by regression analyses utilizing SVR (Support Vector Regression), Random Forest, and Extreme Gradient Boosting algorithms. The proposed architecture results in the development of six distinct models, evaluated across two different window sizes, specifically 1 and 5, to assess the impact of temporal granularity on predictive accuracy, which may help identify optimal configurations for enhancing model performance in various forecasting scenarios. Model performance is quantified using Mean Absolute Error (MAE) and R² Score (Coefficient of Determination) metrics. The experimental findings indicate that the LSTM-SVR model operating with a window size of 2 exhibits superior performance, achieving an MAE of 1.050 and an R² Score of 0.997 on the training dataset and an MAE of 1.558 with an R² Score of 0.973 on the test dataset. The optimal configuration for the LSTM feature extractor comprises 10 units, with a dropout rate of 0.05, 10 epochs, and a batch size 32. The SVR regressor utilizes an RBF (Radial Basis Function) kernel with parameters C = 100, epsilon = 0.01, and auto gamma.
 
Keywords—brent crude oil prices, time series, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Support Vector Regression (SVR), Random Forest (RF), Extreme Gradient Boosting (XGB)


Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).