Problem
Research questions and friction points this paper is trying to address.
Ship Fuel Consumption Prediction
Temporal Leakage
Model Validation
Time Series Cross-Validation
Time-Aware Validation of Machine Learning Fuel Consumption Models: Evidence from 1\,Hz Operational Data, CCGS \textit{Sir Wilfrid Laurier}
🤖 AI Summary
This study addresses the issues of temporal leakage and inflated evaluation metrics caused by random data splitting in ship fuel consumption prediction. We propose a time-aware, blocked cross-validation strategy to mitigate these biases. Leveraging a dataset of 3.88 million steady-state records sampled at 1 Hz, we systematically evaluated multiple regression models alongside physics-based baselines. The proposed approach effectively eliminates temporal dependency bias and overcomes validation challenges inherent to high-frequency data. Consequently, this method yields reliable performance indicators that accurately reflect real-world deployment conditions. Ultimately, this work provides a rigorous methodological foundation for the fair assessment and practical engineering application of ship energy efficiency models, ensuring that predictive performance is validated without artifacts from improper temporal partitioning.
📝 Abstract
Ship fuel consumption (SFC) prediction supports vessel operation optimisation, emissions estimation, and decision support systems (DSS) for sustainable maritime transportation. Numerous data-driven fuel models have been developed over the past two decades, but a critical and often overlooked limitation lies in their validation practices: most studies evaluate performance using random train--test splits, which, applied to high-frequency records, admit temporal leakage and yield optimistic results that do not reflect deployment conditions. This paper examines that gap using time-aware evaluation, specifically Time Series Cross-Validation (TSCV) and Blocked TSCV (BTSCV). Using the Canadian Coast Guard Ship (CCGS) \textit{Sir Wilfrid Laurier} as a case study, six regression models and a physics baseline are tuned under three time-aware schemes and three feature configurations, then evaluated on a common chronological hold-out set drawn from approximately 3.88 million steady-state 1\,Hz records.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Time-Aware Validation
Time Series Cross-Validation
Temporal Leakage
Ship Fuel Consumption
High-Frequency Operational Data
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
