A data-driven bayes approach for investigating international safety management code-sourced detention of ships in port state controls

Abstract

For port authorities and shipping firms to enhance vessel quality and ensure safety of maritime, Port State Control (PSC) inspections are crucial. Notwithstanding the tremendous efforts made in recent years to improve PSC, one issue that persists in PSC inspection practices today is the absence of pertinent schemes or scholarly studies that concentrate on the particular deficiency type-centric perspective of detention of vessel, which is crucial to the inspection mechanism. Considering the International Safety Management (ISM) Code type-sourced deficiencies, which is one of the most prevalent deficiency types, this paper reveals and evaluates the correlation between various influencing factors and types of deficiencies, and their effect on detention caused by ISM Code deficiency. In this regard, it is aimed to develop a data-driven machine learning-based model using the detention records collected within the Tokyo MoU region from 2017 to 2023 in this paper. Tree Augmented Naive Bayes (TAN), one of the most popular data-driven Bayesian Network techniques, is therefore exploited. Findings of this study point out that detention period appears as the most important predictor to determine the occurrence of detention caused by ISM Code deficiency followed by detention place and ship type, respectively. The findings of this research may provide significant insights to port authorities and ship operating companies for developing policy formulation and setting priorities to mitigate the detention risk.