How unsafe acts occur: an automatic text mining study
Shi Xin, Dong Xu, Hui Zhuang, Chen Liu
Abstract
Unsafe acts influence occurrence of maritime accidents; they thus constitute an important topic in maritime safety management. This study investigated the mechanism through which unsafe acts occur through automatic text mining of 344 accident reports. The causal factors and their interrelationships in this mechanism were analyzed using Kruskal–Wallis tests and in terms of their Pearson correlations. The text mining findings strongly support the classical Swiss cheese model, with the exception of findings of the fifth group of causal factors and findings of connections between factors within the same level and in nonadjacent levels. The distribution of causal factors differs with respect to ship type and accident type. Practical measures based on interactions between different factors are proposed, and the methodological implications of automatic text mining are also discussed.