2026-05-13T16:14:48.735Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:9798 2024-12-03T20:10:47Z 3

Marine Vessel Trajectory Forecasting Using Long Short-Term Memory Neural Networks Optimized via Modified Metaheuristic Algorithm, Chapter in AIS Algorithms for Intelligent Systems: ICTSM 2023: Trends in Sustainable Computing and Machine Intelligence, Springer 2024 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/9798 https://link.springer.com/chapter/10.1007/978-981-99-9436-6_5 A. Toskovic A. Petrovic L. Jovanovic N. Bacanin M. Zivkovic M. Dobrojevic Nautical vessels present a trove of data encompassing their whereabouts, trajectories, and speeds. However, beyond their meticulous monitoring for navigation upkeep, this data holds a wealth of additional contextual cues. This study delved into the realm of harnessing data-centric methods and deploying artificial intelligence (AI) to enhance the prediction of vessel trajectories, achieved through time series forecasting employing long short-term memory (LSTM) networks. Yet, considering the pivotal role of selected hyperparameters in the efficacy of AI models, an amplified adaptation of the well-established firefly algorithm (FA) was introduced. This custom-tailored version aims to precisely calibrate the hyperparameters of models employed in this research. This inventive approach was put into action by using authentic automatic identification system (AIS) data from real-world scenarios. In a comparative analysis against prevailing optimization techniques, the performance of this novel FA iteration leveraged promising outcomes. bookPart Springer, Singapore 51 66 10.1007/978-981-99-9436-6_5 AIS Algorithms for Intelligent Systems: ICTSM 2023: Trends in Sustainable Computing and Machine Intelligence