A Research On Weibo (Micro-blog) Data And The Construction Of A Blogger Analysis System

As the carrier of information transmission,the Internet has a large number of users to produce massive data,then how to effectively analyze these data has become the attention of an increasing number of researchers.Among them,there exists a significant and difficult part of data mining is that digging out characteristics of those Internet users from a large amount of data,that means also digging out which sort of information those users are interested in.Furthermore,the users are divided into groups in a reasonable way.In this paper,we use a considerable amount of the bloggers' published information obtained from the "Weibo"(microblog)platform as the data source,these data in most of the contents reflect the willingness to bloggers.Research works operated on this data set will have extremely high value both on scientific and commercial aspects,especially have a crucial meaning of the prediction tasks about social media.In a practical point of view,topic model is utilized to extract the text theme of one single Weibo user through text processing from the blogger,and then calculate the similarity between single user text topic and topics trained from corpus,according to the similarity results by order,artificially add multi-tag description,combined with Weibo text categorization using topic model,bloggers hold the similar tags of interest topics will be distributed into the same group,the interest analysis will be implemented by groups.Consulting the results of the final division,we elected topics that bloggers interested,and their interest tendency.Therefore,the bloggers' potential characteristics will be found out and categorized by interest,meanwhile user group division is completed.On the other hand,from the enterprise promotion point of view,time slice partition model is established on users within the group.In a fixed length of time fragment,we consider influence factors and active degrees of each bloggers,then determine the optimum promotion time and target user groups on the reference of the cluster analysis also.