Applied Computational Intelligence and Soft Computing, Год журнала: 2025, Номер 2025(1)

Опубликована: Янв. 1, 2025

The ongoing conflict between Russia and Ukraine has introduced significant volatility into the global oil markets, highlighting need for robust forecasting models to understand anticipate price fluctuations during such geopolitical events. This study presents a comprehensive hybrid modeling approach predict prices in context of across three distinct periods: before war, immediate aftermath conflict. Using advanced machine learning techniques, we developed system combining Random Forest, ElasticNet, K‐Nearest Neighbors, Gradient Boosting, Support Vector Regressor models. These were integrated through Voting enhance prediction accuracy. Our analysis revealed varying levels predictive performance different periods. Prior showed strong capabilities, evidenced by low mean‐squared error (MSE) values high coefficients determination ( R 2 ). However, struggled capture extreme volatility, resulting significantly increased MSE negative values. Predictions demonstrated improvements, with reduction positive values, indicating return relatively more stable market conditions. Notably, data from both war periods could further improve accuracy, as it would reduce impact conflict’s on model performance. results emphasize challenges instability underscore importance approaches adapt rapidly changing dynamics.

Язык: Английский