Data-driven Bayes approach on marine accidents occurring in Istanbul strait

Abstract

Analysing of marine accidents is crucial for vessels passing through narrow and busy waterways. The Istanbul Strait is one of the narrowest channels in the world and is exposed to intense maritime traffic. Taking accidents that occurred in the Istanbul Strait into account, this study proposes a quantitative assessment. 418 vessel accidents, which are taken place in the four sectors (Turkeli, Kandilli, Kadiko center dot y, Marmara) that constitute the Istanbul Strait area under Istanbul Vessel Traffic Services (VTS) scope, are investigated. Considering accident type as a target variable, this study concentrates on the probabilistic relationships among the factors (i.e., vessel age, flag, wind speed, visibility, current) which are thought to influence the occurrence of accidents. Therefore, Tree Augmented Naive Bayes (TAN) which is one of the most utilized data-driven Bayesian Network approaches is employed. The outcomes of the research indicate that small vessels especially under 300 GRT are more prone to experience adrift accident which is also found as the most frequent accident type in the Istanbul Strait. Domestic maritime authorities can utilize the findings of this study to prevent the reoccurrence of accidents and develop more effective measures.