Probabilistic Modeling of Ship Collision in The Lombok Strait, Indonesia: A Bayesian Network Approach

Abstract Ship collisions are a type of maritime accident that can result in significant losses due to their direct impact on the structural integrity of the vessels involved. Collisions can occur between two or more ships crashing into each other. Such accidents can cause severe damage, threatening the safety of passengers and crew on all vessels involved, and generally result in greater losses than single-ship accidents. According to KNKT (National Transportation Safety Committee of the Republic of Indonesia) investigations, ship collisions occurred in the Lombok Strait from 2007 to 2019. The Lombok Strait is part of ALKI (Indonesian Archipelagic Sea Lanes) II, which connects Australia and East Asia. Therefore, assessing the frequency of collisions in these waters is necessary to determine the prevalence of such incidents. If the rate is high, efforts can be made to minimize risks and ensure safety in the shipping lanes, particularly in ALKI. A Bayesian Network can provide information about the likelihood of ship collisions, facilitating decision-making to prevent or mitigate the risks posed by such incidents. The model’s formation involves calculating the formed structure’s prior, conditional, and joint probability of the formed structure. Thus, conclusions were drawn based on the model’s predictions of ship collision chances in Indonesia, yielding an accuracy rate of 96.97%, a specificity of 90.00%, and a sensitivity of 100%. Based on Bayesian Network analysis and modeling, the causal probability values for the Lombok Strait are as follows: Pc Head-on, Pc Overtaking, and Pc Crossing are 2.85 x 10^-4, 1.03 x 10^-5, and 6.24 x 10^-5, respectively. Based on the Bayesian Network modeling results, the frequency of ship collisions in the Lombok Strait for each type of collision are 0.000026 collisions/year, 0.0000031 collisions/year, and 0.0000015 collisions/year, respectively.