A Research Of The Correlation Analysis Method Of Twitter Event Under Graph Mining

With the development of the Internet,social media plays a more and more important role in people's lives.Twitter,as a leader in social media,has become one of the most popular social media applications in recent years.Barack Obama,the former president of the United States,won US presidential election in 2012 with the help of twitter.The United Kingdom's prospective withdrawal from the European Union is widely known as “Brexit”.Twitter tends to gradually replace the traditional opinion polls in reflecting the political tendencies of the people.At present,the study of political tendencies in social media mainly focuses on the specific information in the text,such as the hashtags(#),the references(@)and so on.Because social media data is not formal,the results of political bias analysis are not accurate enough.At the same time,in the social media,there is no particularly perfect method to predict the election.On the basis of studying the Twitter events,this thesis puts forward and designs the forecasting model for the election of the 2016 US presidential election.On the basis of the detection of Twitter events,the emotional tendency and voting tendency of American users are predicted by emotion analysis method and complex network community detection method.The main work and innovation of this thesis are summarized as follows:(1)Analysis of political tendencies and Political event detection on Twitter data sources.In the emotional analysis,Tweets is short,informal and lack of supplementary information.This thesis applies a dictionary-based emotional analysis method to judge the political emotions of the tweets.At the same time,sarcasm transforms the polarity of the message into its opposite.through the tweets of the expression and the user's history to promote the text to enhance the results of emotional analysis.In the US presidential election,the event detection using a number of clusters,synonyms,keyword weight lifting and other methods to enhance the performance of event detection results.(2)The method of complex network analysis is used to model the Twitter data.Because Twitter data sources have a variety of information,such as users,tweets,pictures,video and so on.And the user's liked,retweeted and commented often shows the user's political tendencies.Therefore,this thesis uses the method of complex network to project different types of social media data into complex network,and then analyzes the complex network of social media according to the actual demand.The main purpose of this thesis is to select the general election,so the focus of the forecasting model is to find the user associations that support the presidential candidates,and to analyze the core characters in the community.This method of analysis is not only applicable to the election situation prediction,but also can be applied to the social media public opinion analysis,user influence and so on.This thesis uses real social media data to experiment with emotional analysis methods,event detection models,and election prediction models.The experiment results show that the forecasting model plays well on the actual data source.Compared with the results of the real presidential election in the United States,the forecast model can predict the final result of the election.