Determining the critical risk factors for predicting the severity of ship collision accidents using a data-driven approach

Ship collision accidents often result in serious casualties and property losses. Predicting the severity of ship collisions is beneficial to improve maritime transport safety. Therefore, this study proposes a data-driven approach integrating association rule mining (ARM), complex network (CN), and random forest (RF) to explore the correlation among risk factors and determine the critical risk factors for predicting the severity of ship collision accidents. Specifically, ARM is integrated with CN to develop the risk interaction network of ship collisions and to identify the criticality of risk factors. Then, RF is employed to predict the severity of ship collisions, and determine the risk factors that have a critical effect on severity prediction. The results show that poor team communication is the most critical risk factor for predicting the severity of ship collisions. Moreover, the criticality of risk factors is different in the risk networks and prediction model. Results from this study would help relevant stakeholders to assess current risks and tailor safety strategies to reduce the severity of ship collisions.

Automated summary

This study proposes a data-driven approach integrating association rule mining, complex network, and random forest to determine the critical risk factors for predicting the severity of ship collision accidents. The results show that poor team communication is the most critical risk factor.