Occupational accident analysis according to professionals of different construction phases using association rules

Despite continuous efforts to improve safety, worker safety awareness on construction sites is a major concern as it remains one of the most dangerous industries. The large number of factors involved in accidents and the complexity of the relationships between them make management difficult for managers. Therefore, potential hazards cannot be identified in order to develop effective safety procedures. This study addresses this problem by using the association rule method of data mining to extract knowledge from historical data of construction accidents. It can help managers to identify and provide frequent conditions that can be prevented in future by controlling risks on site. Occupational accidents that have been notified through an official electronic system on Spanish construction sites between 2003 and 2015 are analysed. Data have been divided according to professionals to explore the accidents in each construction phase. The results show patterns and recurrent factors with multiple relationships in all phases of the construction works. This is the case for the outsourcing variable which is a critical factor in occupational accidents in all construction phases. Similarly, the results have also shown that risk assessment is not an insurmountable barrier to accidents. Consideration of the different stages of the work provides flexibility in order to improve risk reduction and control actions. The results of the study provide a framework for improving safety practices, providing a valuable reference for all agents involved in the construction industry to improve risk management, preventive measures and action plans.