Investigation of accident severity in sea lanes from an emergency response perspective based on data mining technology

Due to the unobserved heterogeneity inherent in the severity data of maritime accidents, traditional methods used to investigate the severity of maritime accidents always result in hiding some underlying relationships. This paper proposes a methodology for analysing the factors affecting the severity of maritime accidents from an emergency response perspective. A two-step clustering method is first used to classify maritime accidents into two homogeneous clusters. Subsequently, multinomial ordered logit models are developed for each cluster to analyse the unobserved heterogeneity that affects the severity of the three accident consequences: ship damage, casualties, and environmental damage. Analyses are performed based on data collected from China's Maritime Safety Administration reports on maritime accidents that occurred in sea lanes of the country's maritime transportation system. The model estimation results indicate that a wide spectrum of factors affect the severity of accident consequences, including natural environmental factors, ship characteristics, accident characteristics and rescue conditions. However, the factors have different effects on the severity of accident consequences. Furthermore, variations in estimated coefficients across clusters reveal unobserved heterogeneity. The proposed methodology can effectively explore the factors affecting the severity of maritime accidents, and the results provide a decision-making strategy in maritime emergency rescue measures.