Dynamic Portraits,Evolution Law Discovery And Prediction Of The Evolution Of Social Network National Security Emergencies

Social networks,as the main media and platform for recording and disseminating information about the daily life of the public,generate a large amount of data information every day,of which text information is its main form.Social network text information is characterized by short text,noise,and sparsity.At the same time,from the perspective of national security incidents,the role of social networks in national security emergencies is becoming more and more important.Based on the characteristics of social network national security emergencies,this thesis conducts research on the collection,processing and storage of social network national security emergencies,studies on dynamic portraits of social network national security emergencies,and discovers and predicts the evolution of social network national security emergencies the study,and eventually realize the social network national security emergency dynamic portrait,evolution law discovery and prediction system.The specific work is as follows:(1)In terms of social network national security emergency data collection,processing and storage,according to the characteristics of social network national security emergency data,a method for acquiring social network national security emergency multi-attribute data is proposed by the thesis.The Weibo web platform and crawler tools are used to obtain data,while the corresponding social network national security emergency thesaurus was constructed for data noise processing and normalization processing.Finally,Mysql relational database is used for effective data storage.Use the characteristics of Weibo web data to obtain Weibo text data,Weibo user data,and Weibo attention data.The text obtained a total of 241,145 Weibo posts on the four social network national security emergencies of " Wuhan Rainstorm","Beijing Haze","Tianjin Explosion",and "China-US Trade War",as well as user information of Weibo posts with corresponding location and time information.(2)In terms of dynamic portraits of social network national security emergencies,according to the current method of neural network dynamic portraits and text data characteristics of social network national security emergencies,use knowledge maps to construct dynamic portraits,and from the key technologies of constructing the knowledge map,a named entity recognition algorithm that integrates Chinese word segmentation part-of-speech attention mechanism algorithm(BLTAC)and a multi-attention mechanism-based entity relationship algorithm(BLMA)are proposed by the thesis.These two algorithm are used to construct the social network national security emergency knowledge map,and combining the time series attributes to realize the dynamic portrait of social network national security emergency.For the method of named entity recognition,the accuracy of the proposed BLTAC algorithm on the social network national security emergency data set is improved by about 4.5%compared with the currently popular algorithm.At the same time,for the entity relationship extraction technology,the precision of the proposed BLMA algorithm in social network national security emergencies is improved by about 4%compared with the current popular algorithms.(3)In terms of the discovery and prediction of the evolution law of social security national security emergencies,combined with the characteristics of the multi-attributes of social network national security incident data,a method for the evolution of multi-dimensional social network national security emergencies is proposed by the thesis.This method can discover the evolution of social network national security emergencies from multi-dimensional and multi-attributes such as Weibo text data,Weibo user data,and Weibo attention data.At the same time,the social network national security emergency heat trend is predicted,and a heat trend prediction algorithm based on multi-granularity features(MHTP)is proposed by the thesis.Features of multi-attributes of event data combined with deep learning technology to predict heat trends.Compared with the current popular algorithm,the MHTP algorithm proposed in this thesis improves the precision of thermal trend prediction on social network national security emergency data sets by 2.4%,and achieves accurate prediction of social network national security emergency thermal trends.(4)Designed and implemented a social network national security emergency dynamic portrait,evolution law discovery and prediction system.Social network national security emergency data collection,processing and storage modules,social network national security emergency dynamic portrait module and social network national security emergency evolution rule discovery and prediction module,realizing the functions of social network national security emergencies dynamic portrait,evolution law discovery and prediction.At the same time,a user-friendly algorithm interface and a simple and clear interactive interface are designed,and the results of running the proposed algorithm model are fully displayed.