Analyzing subway operation accidents causations: Apriori algorithm and network approaches

Abstract

Subway operation safety management has become increasingly important due to the severe consequences of accidents and interruptions. As the causative factors and accidents exhibit a complex and dynamic interrelationship, the proposed subway operation accident causation network (SOACN) could represent the actual scenario in a better way. This study used the SOACN to explore subway operation safety risks and provide suggestions for promoting safety management. The SOACN model was built under 13 accident types, 29 causations and their 84 relationships based on the literature review, grounded theory and association rule analysis, respectively. Based on the network theory, topological features were obtained to showcase different roles of an accident or causation in the SOACN, including degree distribution, betweenness centrality, clustering coefficient, network diameter, and average path length. The SOACN exhibits both small-world network and scale-free features, implying that propagation in the SOACN is fast. Vulnerability evaluation was conducted under network efficiency, and its results indicated that safety management should focus more on fire accident and passenger falling off the rail. This study is beneficial for capturing the complex accident safety-risk–causation relationship in subway operations. It offers suggestions regarding safety-related decision optimization and measures for causation reduction and accident control with high efficiency.