Social Media Based Disaster Event Extraction And Spatiotemporal Analysis

In the current era,with the rapid development of the Internet,more and more users publish their opinions,opinions and events at any time on social media.Social media is used as a major data platform for network information,and it is used to sense the events and developments around them at any time.When a disaster event occurred,a large amount of text and media data containing location information quickly flooded the entire network.This article discusses a method for disaster information emergency information mining and analysis.Using social reptiles method to obtain social media data,based on such data,using the LDA theme model extraction method,a disaster catastrophe event subject feature classification model library is established,and a SVM algorithm is used to quickly extract and classify disaster events from a large amount of social media data,and then The method o f extracting Chinese geographical names combined with the rules of pre-suffix feature words and rules for disaster incidents,combined with the geocoding function of the open source network map API,address extraction and spatial location of disaster events and emergency information;finally,by 2017 Taking Jiuzhaigou earthquake as an example,for the time trend of earthquake disasters,thematic trends,and spatial positioning of different topics,etc.,statistical analysis,time analysis,and spatial analysis methods are used to explore the temporal trends and spatial distribution of emergencies and provide emergency response policy support..The main research contents of the thesis include:(1)Obtain Weibo disaster events through Sina Weibo's API and web crawler method,and build a model library of disaster event topics through a weighted LDA topic model extraction method.Use SVM algorithm to identify text events of micro-blog events.C lassification with topics.(2)Using the statistical method of man-machine integration,put forward the technology of address extraction of place names combined with pre-suffix feature words and rules in short texts of disaster events.Effectively extract disaster events and related emergency information.(3)Combine the geocoding function of the open source network map interface to effectively spatially locate disaster events with fuzzy and bearing attributes.(4)Using the method described in this article,using the “Jiuzhaigou earthquake” event as an example,the classification,positioning,and statistical analysis of the topic,time,and spatial information of typical disaster events from the Weibo data were realized.