Machine learning in occupational accident analysis: A review using science mapping approach with citation network analysis

Abstract

The present study reviews the publications that examine the application of machine learning (ML) approaches in occupational accident analysis. The review process includes four phases of analysis, namely bibliometric search, descriptive analysis, scientometric analysis, and citation network analysis (CNA). In the bibliometric search, a total of 232 articles are systematically screened out from 1995 to 2019 (up to May). Then, descriptive analysis and scientometric analysis are carried out to find the influences of journals, authors, authors’ keywords, articles/documents, and countries/regions in developing the domain. Thereafter, CNA is carried out to classify the publications according to the research themes and methods used. From this extensive review, several key findings are obtained in the application of ML approaches in occupational accident analysis. USA, China, and Taiwan are the leading countries/regions in publishing articles. The four major research domains are (i) prediction of incident outcomes, (ii) extraction of rule based patterns, (iii) prediction of injury risk, and (iv) prediction of injury severity. Then, a taxonomy of the ML algorithms used is developed. Finally, research gaps and safety issues are highlighted and the scope for future is discussed.