Identifying Traffic Incident Patterns in Motorway Work Zones Under Live-Traffic Construction Conditions Using Text Mining

Emergency traffic incidents during motorway reconstruction and expansion projects exhibit characteristics distinct from those under normal operating conditions. To identify incident types and examine their temporal, seasonal, and vehicle-related patterns during the construction period, this study analyzed 372 emergency traffic incident reports from the Guilin–Liuzhou Expressway reconstruction and expansion project in 2023. Text mining was applied to extract incident-related keywords, followed by term frequency–inverse document frequency (TF-IDF) vectorization. Unsupervised learning methods were then used to identify strong word–word association rules. A text network was constructed based on these associations, and the Leiden algorithm was employed for community detection to derive thematic clusters and further examine incident characteristics and response differences. The results indicate that emergency traffic incidents during motorway reconstruction and expansion were grouped into six categories: night-time expressway incidents, summer weekday afternoon tyre blowouts, morning incidents during spring holidays, congestion caused by immobile vehicles, winter lorry incidents, and incidents during the National Day holiday. These categories display pronounced variation across seasons, time periods, and vehicle types, suggesting that traffic safety management should pay particular attention to holidays and peak travel periods. Incident duration analysis further shows that congestion-related incidents caused by immobile vehicles exhibit the widest incident-duration range and the longest duration, indicating substantial operational challenges for emergency handling. Overall, the findings demonstrate that unstructured traffic incident reports contain rich operational information that can be effectively extracted through text-mining techniques. The proposed analytical framework offers a practical approach for identifying risk patterns during motorway reconstruction and expansion projects and for supporting targeted traffic safety management, emergency resource allocation, and incident response optimization.