Study On Search For Emergencies Endangering National Security In Socail Network

At present,with the rise of social networks,Weibo has rapidly become popular because of its convenience.And now it has become the main channel of network information dissemination.Based on the acquired social network data that endanger national security,this thesis establishes an emergency search system that is endangering national security in social networks.The main work done in the thesis is as follows:(1)We obtain large scale spatiotemporal data related to emergencies in social networks by the use of swarm intelligence and spatio-temporal pattern in mining technology.Classified the emergencies that endanger the national security,built the event keyword database that endangers the national security,obtained the emergency data that endangers the national security in the social network.Then we segmented the data,removed stop words and stored the data in the database.Based on the emergency classification system that endangers the national security,the emergency data is expressed at different levels such as events,scenarios,processes,actions and spatiotemporal relationships.(2)A WMD-based emergency detection algorithm(EDW)is proposed.EDW algorithm is based on the crawling data of the theme of emergency that endangers the national security,word segmentation of the data,stop words and other pretreatment.Get the word vector representation of each document.WMD algorithm is used to calculate the similarity of documents,cluster the emergencies,calculate the keywords of emergencies and get the final emergencies.The experimental results show that compared with the traditional emergency detection algorithm,the EDW algorithm improves the accuracy of emergency detection results and improves the recall and F-means.(3)An emergency search algorithm based on long short term memory network called EventScore is proposed.EventScore algorithm establishes a long-short term memory network,which combines with the time space and user features,and achieves high national security search performance in large-scale social network data.The experimental results show that compared with the traditional emergency search algorithm,the average accuracy of EventScore algorithm for emergency search results is improved by 10%.(4)The design and implementation of the emergency search system in the social network that endangers the national security.The system is divided into three functional modules,which are the national security emergencies data acquisition module,endangering national security emergency detection module,endangering national security emergency search module.This thesis realizes the detection of emergencies that endanger national security in social networks,completes the emergencies search system that endangers national security in social networks,and realizes monitoring public opinion on endangering national security.