Predicting the outcome of construction incidents

Ayhan B. U., TOKDEMİR O. B.

SAFETY SCIENCE, vol.113, pp.91-104, 2019 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 113
  • Publication Date: 2019
  • Doi Number: 10.1016/j.ssci.2018.11.001
  • Journal Name: SAFETY SCIENCE
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI), Scopus
  • Page Numbers: pp.91-104
  • Keywords: Construction occupational health and safety, Multiple regression analysis, Artificial Neural Network (ANN), Fuzzy set theory, OCCUPATIONAL INJURIES, SAFETY PERFORMANCE, NEURAL-NETWORK, HEALTH, TOOL, STRATEGIES, PROJECTS, MODELS, AHP
  • Istanbul Technical University Affiliated: No


Ensuring the Occupational Health and Safety (OHS) is the utmost essential issue for all construction projects. The problem with the construction industry is that even though there is a significant emphasis on safety requirements, for most of the contractors unless requested by the client or enforced by law, there is a tendency to undermine safety management to save money especially in the case of competitive bidding including incident collection systems. The objective of this study is to develop a model to predict incidents at construction sites and propose an effective mechanism to prevent incidents utilizing data through incident collection systems. This research consists of three steps as collection and categorization of data, development of prediction models, and selection of the most appropriate method for prediction of construction incidents. In the first step, real data about construction incidents were categorized with the help of experts using the Delphi method. In the second step, both conventional and artificial intelligence techniques were used to predict the outcome of construction incidents. The prediction step can predict the 84% of the incident outcome within 90% confidence. In the final step, fuzzy sets were introduced to tackle the vagueness of prediction output, and the process was finalized with improvements in the prediction results. Ultimately, it is argued that if a construction site has a mechanism in place to record the incidents as proposed in this study, problematic areas can be detected, and preventive actions can be taken.