Improving Workplace Hazard Identification Performance Using Data Mining

Abstract

Hazard identification, as the first major step of risk management, is a crucial activity for reducing accidents and other related losses. However, recent research has revealed that a large proportion of workplace hazards remain unidentified, and the identification process is also time consuming. To improve workplace hazard identification performance, an associated hazard prediction method is proposed which consists of an equivalence class transformation (Eclat) algorithm, a change mining algorithm, data visualization, and other data mining techniques. Through the data mining of historical hazard information, the method can extract association rules and changes related to an identified hazard and then predict other associated hazard information, including types, probabilities, and change trends, to assist with hazard identification and management. The function of the method is twofold. Firstly, associated hazard information can be predicted to help superintendents enhance the pertinence of identification, and then the problem of incomplete hazard identification can be solved. Secondly, with the help of the data visualization technique, superintendents can intuitively understand the potential relationship between hazards and obtain more valuable information to identify and control hazards early, thus improving efficiency. Case studies of standardized management of Chinese enterprise workplaces are presented. The case studies show that up to 47.37% of the hazards can be predicted, and the efficiency is increased by an average of 31.53%.