Maritime IoT-oriented ship speed prediction: Integrating adaptive wavelet analysis and attention-based uncertainty modelling

Liu, Y ORCID: 0009-0008-0729-5303, Liu, J, Zhang, Q ORCID: 0000-0002-0651-469X, Liu, Y, Wang, Y and Yang, L (2025) Maritime IoT-oriented ship speed prediction: Integrating adaptive wavelet analysis and attention-based uncertainty modelling. Regional Studies in Marine Science, 91. p. 104574. ISSN 2352-4855

Text Maritime IoT oriented ship speed prediction Integrating adaptive wavelet analysis and attention based uncertainty modelling.pdf - Accepted Version Access Restricted until 19 October 2026. Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (2MB)

Abstract

Accurate ship speed prediction with robust uncertainty quantification is a critical yet unresolved challenge in maritime IoT systems. The inherent volatility of speed signals, unlike the more stable spatiotemporal patterns of ship trajectories, complicates forecasting. This study addresses this gap by proposing an attention-enhanced bidirectional LSTM framework (ABi-LSTM) for probabilistic ship speed prediction. The methodology introduces two pivotal innovations. First, a novel method for navigation pattern analysis is presented, where speed is treated as a signal and its frequency-domain energy features are extracted using the Discrete Wavelet Transform (DWT). This technique captures distinct volatility signatures, enabling a more physically meaningful and homogeneous grouping of diverse operational patterns. Second, the proposed ABi-LSTM architecture integrates Lower Upper Bound Estimation (LUBE) for direct uncertainty output. Its attention mechanism dynamically prioritizes critical speed variations to enhance sensitivity to transient patterns, while LUBE constructs the prediction intervals. Comprehensive evaluations demonstrate the framework’s superiority over state-of-the-art benchmarks across reliability (PICP), sharpness (PINAW), and accuracy (MAE, RMSE) metrics. Furthermore, the model’s explainability is demonstrated by visualizing the attention weights, revealing a clear connection between ship manoeuvring and predictive uncertainty, thereby offering tangible operational insights for maritime decision support.

Item Type:	Article
Uncontrolled Keywords:	37 Earth Sciences; 3708 Oceanography; 31 Biological Sciences; 3103 Ecology; 7 Affordable and Clean Energy; 0405 Oceanography; 3103 Ecology; 3708 Oceanography
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering
Divisions:	Engineering
Publisher:	Elsevier BV
Date of acceptance:	16 October 2025
Date Deposited:	23 Jan 2026 13:16
Last Modified:	23 Jan 2026 13:16
DOI or ID number:	10.1016/j.rsma.2025.104574
URI:	https://researchonline.ljmu.ac.uk/id/eprint/27969

View Item