2026-05-10T07:56:17.830Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:2:9789 2024-12-03T20:10:47Z 2

Long short-term memory tuning by enhanced Harris hawks optimization algorithm for crude oil price forecasting 2024 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/2/9789 https://www.sciencedirect.com/science/article/pii/S0065245824000123 L. Jovanovic M. Antonijevic M. Zivkovic M. Dobrojevic M. Salb I. Strumberger N. Bacanin The effects of crude oil price variation on economic stability are significant. The need for a robust method for mitigating volatility in crude oil prices is apparent. By regarding price data as a time-series, artificial intelligence (AI) methods can be applied to this problem. A method that has shown great potential is the adaptation and application of long short-term memory (LSTM) networks. However, the complexity intrinsic to crude oil prices makes it difficult to use a simple network. Therefore, signal processing methods are applied to decompose complex data into simpler subcontinents that can be addressed individually with greater success. The method used in this research is the variational mode decomposition (VMD). AI technique depends on selected hyperparameters. This work applied an enhanced Harris hawks optimization (HHO) algorithm with enhancing overall prediction accuracy. The performance of the introduced method is validated on real-world financial data concerning daily spot prices of Brent crude oil from May 20, 1987 to October 3, 2022. The novel proposed approaches outperformed tested contemporary metaheuristics with respective overall R2 scores of 0.989431 without decomposition and 0.992177 with VMD when casting predictions five steps ahead, solidifying this methods potential for addressing this category of problem. article 10.1016/bs.adcom.2024.01.002 1 40 0065-2458 ADVANCES IN COMPUTERS