Accident Analysis: A Text Mining and Complex Networks Framework for Safety Incident Prevention

Safety is a key monitoring point for the construction project by engineers. Accidents in the project are often accompanied by casualties, property losses and lead to negative social impact. It is vital to learn from what is going wrong in previous safety accidents to prevent the reoccurrence of similar accidents in the future. Large quantities of accident investigation report texts provide a solid foundation for accident analysis. However, accident reports are typically documented and stored as unstructured or semi-structured text data, which makes analytical work difficult. This paper proposes an analytical framework to extract accident causes from accident narratives using text mining technology. An evaluation function TF-K* based on term frequency and complex networks is proposed to identify the importance of safety risk factors. The proposed framework is validated by analysing accident reports of housing and municipal construction in China for ten years.