Detecting Fraudulent Registrations of Ships in China’s Inland Waterway Transport Based on Automatic Identification System Data

Abstract

In China’s inland waterway transport, some private cargo ships are illegally and nominally registered under the names of large shipping companies for benefit, as a result of the hidden agreements between these companies and private shipowners. These are known as “guakao” ships. To reduce costs, private shipowners often employ insufficient and underqualified crew members, making the fraudulently registered guakao ships substandard in safety management and leading to numerous accidents. To address the challenge of identifying guakao ships, this paper presents a data-driven identification approach based on background knowledge of ship operations. Firstly, the automatic identification system (AIS) data are collected and analyzed to derive ship behavior characteristics by data mining and community detection algorithms. Subsequently, machine learning models are formulated and trained to predict ships’ authentic owner types from the input behavior features, thereby identifying the potential guakao ships with fraudulent owner information. Finally, the effectiveness of our identification approach is validated via a case study on a typical bulk shipping company in China. Furthermore, the optimal trade-off between identification accuracy and timeliness is identified by examining the impacts of the length of the AIS dataset on identification results. The proposed approach can help authorities better detect fraudulent registrations of ships and improve the allocation of inspection resources.