A systematic approach for discovering causal dependencies between observations and incidents in the health and safety domain. (October 2019)

• Record Type: Journal Article
• Authors: Polyvyanyy, Artem; Pika, Anastasiia; Wynn, Moe T.; ter Hofstede, Arthur H.M.
• Is Part Of: Safety science. Volume 118(2019)
• Page Start: 345
• Page End: 354
• Publication Date: 2019-10
• Subjects: Big data -- Data mining -- Process mining -- Proximity of events -- Causality -- Health and safety -- Cause of incidents; Industrial accidents -- Periodicals; Accident Prevention -- Periodicals; Safety -- Periodicals; Travail -- Accidents -- Périodiques; 363.11
• DOI: 10.1016/j.ssci.2019.04.045
• Languages: English
• ISSNs: 0925-7535

Abstract

Highlights: An approach for mining causalities between events encoded in Big data is proposed. The approach is based on the notion of proximity of events. The approach was coined in the discussions with an Australian public energy company. The approach was evaluated in a case study with the project partner. The case study aimed at improving the culture of health and safety practices. Abstract: The paper at hand motivates, proposes, demonstrates, and evaluates a novel systematic approach to discovering causal dependencies between events encoded in large arrays of data, called causality mining . The approach has emerged in the discussions with our project partner, an Australian public energy company. It was successfully evaluated in a case study with the project partner to extract valuable, and otherwise unknown, information on the causal dependencies between observations reported by the company's employees as part of the organizational health and safety management practices and incidents that had occurred at the organization's sites. The dependencies were derived based on the notion of proximity of the observations and incidents. The setup and results of the evaluation are reported in this paper. The new approach and the delivered insights aim at improving the overall health and safety culture of the project partner practices, as they can be applied to caution and, thus, prevent future incidents.