Risk Measurement using Bayesian Networks: Applications to Ship Collision Data in Malaysia

Abstract

Identifying hazardous factors that threaten safety is important so that actions can be planned to address the consequences of the hazard. The identification process requires specific measurements and computations, for which the probability and severity of the event are required. In this study, the Bayesan network approach is implemented to evaluate, identify and then rank the factors that contribute to the occurrence of ship collisions. Through a combination of information from expert views and past data, a Bayesian network was constructed to interpret the probability of a collision given the observed variables. Three types of hazards are considered in this study namely technical, natural, and human error. The increase in the probability of a collision is then calculated by conditional to the level in the observed variable. The first three factors that contribute to the occurrence of shipwrecks are major failures in communication systems, high impaired abilities and the absence of a driver as a navigational advisor. These findings help in highlighting the major potentials offered by the Bayesian network for risk analysis and probability calculations. In fact, this study offers a deeper understanding to practitioners in this field to plan or build the necessary action strategies to avoid the occurrence of collisions. This information is also important in assessing the safety of a ship in an effort to reduce the potential for a collision to occur.