Causal Factors Coded from eMARS Dataset

Published: 21 May 2024| Version 2 | DOI: 10.17632/6m82rtk89v.2
shuo yang,


This dataset was originally gathered from the eMARS dataset. The eMARS database is a lesson-learned database in accordance with the purpose stated in the Seveso Directive (2012/18/EU). To preserve its nature as an objective source of lessons learned information and maintain the willingness of authorities and operators to provide an honest and detailed account of what really happened, to prevent future similar incidents, the identifying information of the name of the sites and the location of the incidents in the database is not publicly disclosed.


Steps to reproduce

Five columns: The accident ID, Industry Type, Plant/Equipment Causative Factor Type, Human Causative Type, and Organizational Causative Factor Type. The raw data was filtered by Industry Type, and only the process industry data was kept. Then, based on the factor label also given in the dataset, the causative factors of the process industry data were transferred to binary data, which will be suitable for utilization as training and testing datasets. The cases with less than 10 evidences to be happened were excluded. Based on the CI(conditional independent) test, the dependent factors were highlighted.


Politecnico di Torino


Accident Case, Accident Investigation, Process Industry