2026-06-10T09:50:02.950Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:9736 2024-01-09T13:04:19Z 3

Gold Price Forecast Using Variational Mode Decomposition-Aided Long Short-Term Model Tuned by Modified Whale Optimization Algorithm, Chapter in AIS Algorithms for Intelligent Systems: ICDICI 2023: International Conference on Data Intelligence and Cognitive Informatics, Springer 2024 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/9736 https://link.springer.com/chapter/10.1007/978-981-99-7962-2_6 S. Golubovic A. Petrovic A. Bozovic M. Antonijevic M. Zivkovic N. Bacanin The broad scope of gold’s importance makes it a strategic resource. Its reserves offer protection against systemic risk, currency depreciation, and inflation, highlighting its crucial role in the financial market throughout history. The price of gold is a paramount strategic information that can influence many aspects of the world economy, hence the need for accurate predictions of gold price exists. Predicting the gold price can be tackled as univariate time-series forecasting challenge. Therefore, research presented in this manuscript proposes long short-term memory (LSTM) and recurrent neural network (RNN) structure with respect to addressing this problem. However, the LSTM, as any other deep learning model, should be adjusted (tuned) for specific dataset due to its various hyperparameters, which is non-deterministic polynomial hard (NP-hard) problem. Since metaheuristics optimization methods are known as the robust NP-hard problem solvers, in the proposed research, a modified version of whale optimization algorithm (WOA) is devised and applied to LSTM tuning for gold price forecasting. Additionally, since the gold price time series are non-stationary, the variational mode decomposition (VMD) has been employed for decomposing gold price time series before the LSTM is applied. Moreover, evaluation was performed for one-step, two-step, and three-step ahead forecasts. The proposed method was compared to other metaheuristics-tuned LSTM models, and the obtained results were largely in the favor of approach devised in this study with respect to R2 score, mean average error (MAE), mean square error (MSE), and root mean square error (RMSE) standard regression metrics. bookPart Springer, Singapore 69 83 10.1007/978-981-99-7962-2_6 AIS Algorithms for Intelligent Systems: ICDICI 2023: International Conference on Data Intelligence and Cognitive Informatics