Research On Crisis Information Recognition And Source Prediction In Social Networks

The popularity of social network provides convenience for information acquisition and interaction.Especially in recent years,with the development of instant messaging tools such as We Chat,Tencent QQ and the rise of emerging social platforms such as Tik Tok,Kwai,Sina micro-blog,the interaction of information has become more convenient.The positive information in cyberspace plays a positive effect in promoting economic development and social stability and harmony,while the dissemination of crisis information such as violent terrorism information,false political statements and rumors will pose a huge threat to the economy and social security.Therefore,it is of great practical significance to analyze and study crisis information in social networks.This thesis focuses on two aspects of crisis information in social networks.On the one hand,aiming at the identification of crisis information in social networks,a recognition model based on BP neural network is proposed.On the other hand,the crisis source prediction method of the whole network is analyzed and discussed.According to the characteristics of crisis information dissemination,a prediction algorithm of crisis information sources based on K-means method is proposed.The identification of crisis information depends not only on the text content characteristics of crisis information itself,but also on the characteristics of crisis information publishers and the dissemination characteristics of crisis information.According to the content characteristics,user characteristics and dissemination characteristics of crisis information,this thesis proposes a crisis information recognition model based on BP neural network.Firstly,the crawler technology is used to collect crisis information data,and the key characteristics affecting the identification of crisis information are extracted.Meanwhile,the feature values are defined and quantified.Secondly,the crisis information data is labeled manually,and the crisis information feature label library is constructed based on the extracted crisis information features.Then,the BP neural network is used to train the crisis information feature label library to obtain the crisis information evaluation mechanism.When analyzing the suspected crisis information data,the trained network is used to judge and identify the crisis information.Finally,the performance of the model is verified by comparative experiments.The validity of the crisis information recognition model based on BP neural network is proved,and the validity of the selected features is verified.After the crisis,the crisis information is rapidly disseminated through the network platform.In the process,if the sources of crisis information are not found in time and effective intervention controls are not carried out,it may cause immeasurable consequences.Therefore,the study of sources prediction methods has important practical significance.Aiming at the prediction of crisis information sources,this thesis mainly uses the concept of "effective distance" proposed by Brockmann D and Helbing D in science,and uses the conclusion proposed by Jiang J and others to transform the prediction of crisis information sources into the minimization of effective distance,puts forward the prediction algorithm of crisis information sources based on K-means,and calculates the prediction of crisis information sources.The validity of the method is proved by experiment,and the limitation of the algorithm is explained.