Analysis and application research on spatiotemporal characteristics of microblog information for rainstorm flood disasters emergency rescue based on text classification

Abstract

The frequent occurrence of rainstorm flood disasters has seriously affected the production and life of the public, as well as the sustainable development of the economy and society, posing significant challenges to disaster emergency management. In the context of big data, social media data has become an emerging data source for disaster response. This study employed text mining techniques, utilizing Microblog texts related to the 2021 Henan rainstorm flood disaster event as research data. It conducted spatiotemporal feature analysis of public opinion information during the disaster and formulated targeted emergency rescue decisions based on the research findings. Firstly, Microblog texts related to the rainstorm were collected and classified using Web Crawler technology and Convolutional Neural Network. The Microblog texts were divided into six themes: “Disaster Rescue Situation, Distress Message, Positive Prayer, Reminding, Rainfall and Weather Condition, and Traffic Information”. Secondly, based on the theme classification, techniques such as keyword matching and DBSCAN spatial clustering were used to extract the spatiotemporal features of each Microblog theme. The spatiotemporal features of public opinion were then analyzed. Finally, according to the results of public opinion analysis, disaster relief was categorized into the “Start Emergency Plan Stage, Emergency Rescue Stage, and Continuous Rescue Stage”, with targeted emergency rescue decisions made at each stage. The results showed that the classification and public opinion analysis of disaster related Microblog texts can provide the basis and new ideas for the emergency management and disaster rescue decision-making of rainstorm flood disasters. At the same time, this study also provides the corresponding support for the emergency response to similar disaster events in the future.