Interaction effect of building construction accident attributes based on complex network

This paper aims to effectively utilize data related to building construction accidents by delving deeply into the correlations between key causes and accident attributes. Firstly, the authors gathered 1134 accident investigation reports and employed the "5W1H" analysis method to extract six types of accident attributes: time, location, cause category, activity, building type, and accident type. Subsequently, a word cloud map was employed to identify the primary direct causes, and the correlation characteristics among the four cause categories were analyzed. Finally, a heterogeneous correlation network of building construction accident attributes was constructed using Gephi software. Topological parameters were introduced to analyze the relationships among the six accident attributes. The results indicate that a complex network can effectively analyze the interplay among various construction accident attributes, thus revealing the correlation laws and accident characteristics between various accident attributes. The set of key nodes is represented as {F1, F2, B2, B17, F3, B4, B16, B5, F4, B15, D20}. Thirteen highly correlated sets of accident attributes were identified, highlighting the need for collaborative accident prevention strategies. These findings have the potential to visually present accident knowledge, offering innovative insights for the analysis of building construction accidents.