MSD-LLM: Predicting Ship Detention in Port State Control Inspections with Large Language Model

TL;DR

MSD-LLM tackles ship detention prediction in Port State Control data by fusing a Dual Subspace Recovery Auto-Encoder (DSRAE) with a multimodal Large Language Model (M-LLM). The DSRAE extracts robust PSC representations through a dual-branch reconstruction framework and a margin-based separation loss, while progressive learning mitigates extreme data imbalance. The LLM is trained via a grouping and ranking strategy, blending DSRAE features with prompts to yield detention probability scores and flexible thresholds for decision-making. Evaluated on extensive Tokyo MoU PSC records, MSD-LLM delivers substantial gains in AUC and related metrics over strong baselines, demonstrating practical robustness to real-world data shifts and paving the way for integrated maritime risk prediction using domain-aware LLMs.

Abstract

Maritime transportation is the backbone of global trade, making ship inspection essential for ensuring maritime safety and environmental protection. Port State Control (PSC), conducted by national ports, enforces compliance with safety regulations, with ship detention being the most severe consequence, impacting both ship schedules and company reputations. Traditional machine learning methods for ship detention prediction are limited by the capacity of representation learning and thus suffer from low accuracy. Meanwhile, autoencoder-based deep learning approaches face challenges due to the severe data imbalance in learning historical PSC detention records. To address these limitations, we propose Maritime Ship Detention with Large Language Models (MSD-LLM), integrating a dual robust subspace recovery (DSR) layer-based autoencoder with a progressive learning pipeline to handle imbalanced data and extract meaningful PSC representations. Then, a large language model groups and ranks features to identify likely detention cases, enabling dynamic thresholding for flexible detention predictions. Extensive evaluations on 31,707 PSC inspection records from the Asia-Pacific region show that MSD-LLM outperforms state-of-the-art methods more than 12\% on Area Under the Curve (AUC) for Singapore ports. Additionally, it demonstrates robustness to real-world challenges, making it adaptable to diverse maritime risk assessment scenarios.

Figures (3)

• Figure 1: Overview of MSD-LLM, including progressive learning strategy(top sub-figure), the DSR representation learning module(middle sub-figure) and LLM prediction module(bottom sub-figure).
• Figure 2: Training Loss(a) and Validation average precision(b) of different training phases in progressive training strategy. (c)demonstrates the impact of training data distribution on the final results. (c) also shows that aggregating different phases together can make the model compatible with different data distributions.
• Figure 3: Visualization of sample distance to the threshold. (a)shows the score of RSR-layer based representation learning module while (b) indicates the score of DSR-layer based module.