Time-statistical laws of workers' unsafe behavior in the construction industry: A case study

The construction industry is extremely high risk, and the unsafe behavior of workers is thought to be a critical factor in that risk level. Considering the limitation that existing studies merely analyze the relationship between unsafe behavior and the time factor, this paper uses a case study to explore two aspects of the time-statistical laws of workers' unsafe behavior: (i) the characteristics of interevent time distributions and (ii) association rules for different worker types. First, workers' unsafe acts at one metro construction site are collected and classified. Second, interevent time distributions of workers' unsafe behavior from different types are analyzed via a human dynamics approach. Finally, a rule mining database is built, from which association rules concerning unsafe behavior, worker type and construction phase are determined using the Apriori algorithm. The results indicate that the interevent time distributions are fat tailed and show that workers' unsafe behavior has the characteristics of burstiness and memory. Furthermore, the strong rules 'construction phase → unsafe behavior' exist for different worker types. The research presented in this paper can facilitate a better understanding of workers' unsafe behavioral patterns and can accordingly be used to control frequent and widespread unsafe acts among different types of workers in different construction phases.