Realising advanced risk-based port state control inspection using data-driven Bayesian networks

In the past decades, maritime transportation not only contributes to economic prosperity, but also renders many threats to the industry, causing huge casualties and losses. As a result, various maritime safety measures have been developed, including Port State Control (PSC) inspections. In this paper, we propose a data-driven Bayesian Network (BN) based approach to analyse risk factors influencing PSC inspections, and predict the probability of vessel detention. To do so, inspection data of bulk carriers in seven major European countries from 2005 to 20081 in Paris MoU is collected to identify the relevant risk factors. Meanwhile, the network structure is constructed via TAN learning and subsequently validated by sensitivity analysis. The results reveal two conclusions: first, the key risk factors influencing PSC inspections include number of deficiencies, type of inspection, Recognised Organisation (RO) and vessel age. Second, the model exploits a novel way to predict the detention probabilities under different situations, which effectively help port authorities to rationalise their inspection regulations as well as allocation of the resources. Further effort will be made to conduct contrastive analysis between 'Pre-NIR' period and 'Post-NIR' period to test the impact of NIR started in 2008.