Assessment of occupational risks In Turkish manufacturing systems with data-driven models

Mutlu N. G., Altuntaş S.

JOURNAL OF MANUFACTURING SYSTEMS, vol.53, pp.169-182, 2019 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 53
  • Publication Date: 2019
  • Doi Number: 10.1016/j.jmsy.2019.09.008
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.169-182
  • Keywords: Turkish manufacturing industry, Occupational accident, Data mining, Association rules, Decision tree analysis, DATA MINING TECHNIQUES, ACCIDENTS, INDUSTRY, INJURY, CLASSIFICATION, PREDICTION, HEALTH, SAFETY
  • Yıldız Technical University Affiliated: Yes


Occupational accidents often occur due to a combination of multiple factors depending on the machine equipment, working conditions and organizational factors during work. According to the occupational accident statistics of the International Labour Organization, Turkey has ranked 1-7 in fatal occupational accidents in the last 9 years. Occupational accidents in Turkey are a critical problem to solve. The vast majority of occupational accidents in Turkey are experienced in the manufacturing industry. Moreover, the number of studies analyzing occupational accidents in the manufacturing industry is limited in the literature. For this reason, the record of 242,537 occupational accidents in the manufacturing industry between 2013 and 2016 was taken from the Turkey Social Security Institution for this study. As methods of data analysis, feature selection, decision tree analysis and association rule mining were chosen as data mining methods and were applied. As a result of the analysis, it was found that the following 5 factors had an effect on the type of accident in manufacturing systems at a level of 92.40%: material (the equipment used during the accident), special activity, general activity, place (department where the accident occurred) and profession. For these factors, tree models were formed in terms of the accident type, and association rules were obtained for each type of accident. The obtained results are believed to be useful and are practical for use in the development and update of the occupational accident prevention politics of the Turkish Republic National Council for Occupational Health and Security, in the risk analysis and assessment processes of occupational health and security professionals, and in trainings on occupational health and security.