Investigating the coupling mechanism for subway construction safety risk through data mining methods involving accident reports

Purpose Subway construction accidents pose significant safety risks. However, accident reports can provide valuable knowledge for safety management practices. Therefore, the aim of this research is to develop a systematic and data-driven approach for identifying accident causes and their interactions in subway construction. Design/methodology/approach About 204 subway construction accident reports from 2003 to 2022 in China are taken as the dataset, and 41 accident causes are identified by text mining. Considering the correlation strength among accident causes, four accident cause clusters are obtained by a spectral clustering algorithm. Finally, the Apriori algorithm is adopted to extract intracluster and intercluster accident cause combinations. Findings Preventive strategies are proposed to control key accident causes and prevent the occurrence of safety risks. Originality/value This research provides new insights into subway construction accident causes using spectral clustering and the Apriori algorithm. By focusing on both intracluster and intercluster association rule mining of accident causes, a complex coupling mechanism for subway construction safety risks is revealed and precise prevention and control of key accident causes are proposed.