Topic Detection And Tweet’s Trends Warning For Chinese Microblog

With the development of mobile network technology, microblog hasbecome an emerging new kind of media. Information spread rapidlythrough microblog. Therefore, the study of topic spreading, detection andwarning technology for microblog has become an important issue. Thispaper studied the user relationship model, topic propagation model andtopic propagation characteristics for microblog firstly, and then proposed atopic detection model for Chinese microblog and a tweet warning model.For topic detection algorithm for Chinese microblog, this paperoptimized the data pretreatment, feature selection, text expression andweight computing part, then proposed a new scoring method for tweets.We consider a tweet belongs to noises when its score is lower than a presetthreshold. After that, this paper proposed an improved topic clusteringalgorithm for Chinese microblog based on Single-Pass incrementalclustering algorithm. This algorithm used a new distance computingmethod and center vector update method. Experiment results show that thetopic detection method proposed can filter most of the noises effectivelyand find out the hot topic from large amount of tweets accurately;meanwhile, it can also classify each tweet into correct corresponding topicclusters.For tweet’s warning model, this paper proposed a prediction algorithmfor key users in tweets’s forwarding list and for user’s retweet behavior.Then we combine the above two prediction algorithm to propose a warningalgorithm for a single tweet. We predict a tweet’s retweet number by predicting the key users in its forwarding list and the retweet behavior ofpredicted key users. Then make a warning for those tweets whosepredicted retweet number is bigger than a preset threshold. Experimentresults show that the tweet warning algorithm can predict tweets with largenumber of retweets effectively, and then we can take some precautionarymeasures in advance to control the transmission and spread of informationon microblog.The topic detection method for microblog can help users to find out hottopics easily; it can also help government to understand the socialdynamics and thoughts of people. While after detected hot topic inmicroblog, we can also make a warning for those tweets that may cause alarge amount of retweets, which can help governments intervene thedissemination of information on microblog and to increase or reduce theinformation dissemination.