Chua, P. C., Chieng, R., Wang, R., Lee, K., Xu, H., Fu, X., & Yan, R. (2025). Development of a knowledge-based noise test for robustness of vessel fuel consumption prediction models. Maritime Policy & Management, 52(8), 1178–1207. https://doi.org/10.1080/03088839.2025.2548786
Abstract:
Machine learning techniques have been applied to vessel fuel consumption prediction with increasing availability of noon reports and sea/weather sensor data. However, such data often contain noise, raising concerns about model robustness. This study proposes a maritime knowledge–based noise addition approach, focusing on identified features of noon reports and sea/weather data. White, multiplicative, and systematic noise were introduced into features and tested on Domain Knowledge–based Artificial Neural Networks (DK-ANN) and Bi-directional Long Short-Term Memory (Bi-LSTM) models. Two key findings emerged: (1) vessel speed is the most sensitive feature to all noise types, exerting the greatest influence on model accuracy; (2) while noise degrades performance in both models, noise on wave period produces greater accuracy changes in Bi-LSTM compared to DK-ANN. This systematic evaluation not only helps identify the most robust models for real-world application but also pinpoints key parameters that significantly affect prediction stability. The proposed approach offers a practical guideline for testing robustness in future maritime prediction models, ensuring more reliable fuel consumption forecasts and supporting improved decision-making in maritime operations.
License type:
Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
Funding Info:
This research / project is supported by the Singapore Maritime Institute (SMI) - Maritime Artificial Intelligence (AI) Research Programme
Grant Reference no. : SMI-2022-MTP-06