Citation

Gong X, Jiang H, Yang D. Mar Pol 2023; 150: e105547.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.marpol.2023.105547

PMID

unavailable

Abstract

Maritime piracy is a serious threat to the safety of cargo, ships, and crews, which can cause enormous losses to stakeholders, highlighting the significance of piracy risk assessment and prevention to the maritime industry. In this study, we propose a two-step analytical framework based on a Random Forest (RF) model, Generative Adversarial Nets (GANs), and Matrix Completion (MC) algorithm to assess the risks of successful piracy attacks. We consider different decision-makers in each step, namely, pirates first select which ships to attack and then ship operators determine the probability of a (un)successful attack. We propose different influencing factors for each step and, in the meantime, solving the problems of incomplete and imbalanced data. A case study in Southeast Asia is then conducted based on the proposed approach. The results show that, in mild wind weather, the likelihood of a piracy attack on a ship with a DWT not exceeding 50,000 tons is up to 90 %. Further, if the attack occurs between 0 and 6 a.m., the probability of success is over 90 %. These results provide more specific information for ship operators and local authorities to develop efficient anti-piracy strategies and policies.

Language: en

Keywords

Assessment of piracy risk; Impact factors analysis; Maritime safety; Piracy prediction; Random Forest