Characteristic-Based Network Analysis on Construction Hazard Warning: A Novel Approach of Network of Networks

This study aims to realize construction hazard warning by determining critical characteristics of hazards and using network of network (NoN) analysis. First, text mining is adopted to analyze the hazard text records. Word segmentation, part-of-speech tagging, and word frequency statistics are performed to extract the high-frequency and meaningful nouns as hazard characteristics. Then, the co-occurrence matrix between characteristics is calculated as the data basis of characteristic hierarchical clustering to determine characteristic indicators. After that, the NoN is established with hazard network as the bottom layer, hazard characteristic network as the middle layer, and characteristic indicator network as the top layer. Finally, the critical characteristics and hazards are determined by considering the correlation strength and network index. The results show that the correlation between each network layer is close and complex. And the layer of hazard characteristic network has more advantages than hazard network layer with tighter connections and higher quality nodes. The proposed NoN based on hazard characteristic may assist to focus on critical hazards and characteristic for improving the efficiency of hazard identification.