Knowledge discovery of correlations between unsafe behaviors within construction accidents

Purpose - Knowledge discovery related to unsafe behaviors promotes the performance of accident prevention in construction. Although numerous studies on accident causation models have discussed the correlations of unsafe behaviors with various factors (e.g., unsafe conditions), limited research explores correlations between unsafe behaviors within accidents. The purpose of this paper is mining strong association rules of unsafe behaviors from historical accidents to clarify this kind of tacit knowledge. 
Design/methodology/approach - A case study was adopted as the research approach, in which accident records from building and urban railway construction in China were selected as data resources. The groups of unsafe behaviors extracted from accident records were expressed by the definitions of unsafe behaviors from safety regulations and operating procedures. Frequent Pattern (FP)-Growth algorithm was used for association rule mining, and the critical correlations between unsafe behaviors were represented by the effective strong rules. 
Findings - The findings identify and distinguish correlations between unsafe behaviors within construction accidents. In building construction, workers and managers should pay attention to preventing unsafe behaviors related to personal protective equipment and machines and equipment. In urban railway construction, workers should especially avoid unsafe behaviors of inadequately dealing with environmental factors. 
Practical implications - Tacit knowledge is transferred to explicit knowledge as the critical correlations between unsafe behaviors within accidents are determined by the effective strong rules. Additionally, the findings provide practice guidance for safety management, to collaboratively control unsafe behaviors with strong correlations. 
Originality/value - This study contributes to the body of safety knowledge in construction and provides a further understanding of how construction accidents are caused by multiple unsafe behaviors.