Research And Implementation Of Situational Awareness For Emergencies In Social Networks

Social networks,such as Twitter.Sina etc.,have been an important part of people's daily life.All the time,there is a large amount of data that is been published and disseminated in social networks.And some of these data may be generated by a public opinion event or a disaster event happening in the reality.The text data of social networks contains much information of emergencies.The dissemination of these information of emergencies may have large influence on the public opinion of social networks.Therefore,it is very important to extract information of emergencies and predict their propagation by using the text data of social network.The traditional detection of events and some information diffusion models cannot satisfy the requirements of social networks.Therefore,this paper improves algorithms of situational awareness and achieves the detection and prediction of emergencies in social networks.Situational awareness consists of three parts: perception,comprehension and projection.(1)In the part of perception,this paper proposes an algorithm of weight calculation which is calculated by using influence of social information.And then by combining some algorithms such as time window,document frequency,we can get burst words which we need.This paper also designs an experiment that verifies the new algorithm improving the accuracy of burst detection.(2)In the part of comprehension,this paper implements text classification and text clustering by using Naive Bayesian classifiers,K-Means.agglomerative hierarchy clustering.The text classifier bases on word vector,TF-IDF,Naive Bayes algorithm and make it possible to find out disaster events in social networks.This paper also improves the accuracy of text clustering algorithm by adding information of time in the calculation of text similarity.(3)In the part of projection,based on the traditional information diffusion model, this paper constructs a situation prediction model which include influence of users and popularity of events.This model consists of two parts,the historical information model and the prediction model of the event.The historical information of the emergency is calculated by Newton interpolation.And the prediction model is constructed by some conception of SIR model,but the calculation of the model and the setting of the node both depend on users' and events' own properties.Experiments in the paper verify that thesituation prediction model can predict the propagation of emergencies correctly.Based on the above algorithms,this paper achieves a model of situational awarenessfor emergencies using in social networks.