Lakhasly

Conclusions: This study delves into the interdependence and combination of fatal causes within the construction industry.The proposed method aids project managers in identifying critical fatal cause groups promptly and reliably In terms of author contributions, conceptualization, methodology, software development, validation, formal analysis, investigation, data curation, original draft preparation, visualization, supervision, project administration, and funding acquisition were carried out by Q.S. and Z.Z. All authors have read and approved the manuscript The research received funding from several sources, including the Fundamental Research Funds for the Central Universities (grant number 2019JBW007), the National Natural Science Foundation of China (grant number 71501008), the Beijing Municipal Social Sciences Foundation (grant number 18GLC070), and the Ministry of Education of Humanities and Social Science Foundation (grant number 20YJC630121) The data used in the study is available for download, and models or code generated can be obtained from the corresponding author upon request.Utilizing an optimized machine learning (ML) model and permutation importance method, the research identifies 21 causes influencing fatalities.It develops a predictive model based on 34 fatal attributes to uncover these relationships.

Conclusions: This study delves into the interdependence and combination of fatal causes within the construction industry. It develops a predictive model based on 34 fatal attributes to uncover these relationships. Utilizing an optimized machine learning (ML) model and permutation importance method, the research identifies 21 causes influencing fatalities. Through an iterative analysis algorithm, the study further quantifies the combination relationships, revealing three cause combinations and one critical cause with the highest risk, totaling 12 causes. With the inclusion of combination factors, the predictive F1 score reaches 92.93%, showing an 8.89% improvement over the ML model without combinations. This demonstrates the efficacy of the developed model in extracting various accident causes with fatal outcomes. The proposed method aids project managers in identifying critical fatal cause groups promptly and reliably In terms of author contributions, conceptualization, methodology, software development, validation, formal analysis, investigation, data curation, original draft preparation, visualization, supervision, project administration, and funding acquisition were carried out by Q.S. and Z.Z. All authors have read and approved the manuscript The research received funding from several sources, including the Fundamental Research Funds for the Central Universities (grant number 2019JBW007), the National Natural Science Foundation of China (grant number 71501008), the Beijing Municipal Social Sciences Foundation (grant number 18GLC070), and the Ministry of Education of Humanities and Social Science Foundation (grant number 20YJC630121) The data used in the study is available for download, and models or code generated can be obtained from the corresponding author upon request. The authors declare no conflicts of interest.