Analysis Of User's Behavior In Social Networks Based On Advanced LDA Model

With the development of Internet,social network in particular micro blog has become a significant way for users to propagate and retrieve information.On the one hand,the explosive growth of information in Internet brings people more source,on the other hand it also makes it harder to obtain useful information.The discovery of latent topics in social networks not only carries high value in public option monitoring and commercial promotion also contributes to the management of internet.Based on sina weibo,the most popular one in China,this thesis do research on sina weibo data collection and user behavior.The main contribution and innovation are as follows:1.Weibo information collecting system.This thesis designs and realizes a sina weibo information crawler system based on improved bloom filter,which mainly decreases the error rate with providing a highly efficient to make URL distinct.This system solves the sina weibo user data acquisition and storage problems.2.Abnormal user behavior analysis.According to the data collecting in weibo,this thesis analyzed the behavior of abnormal user.First we transfer user's attribute into words and generate it into attribute document,then we use UR-LDA topic model for analysis and clustering.3.User concern topic analysis.This thesis mine the user interest topic.First we participle user's message with ICTCLAS and extract the keywords,the we use the UR-LAD topic model to dig and cluster user's interested topic.In this thesis,we compare the data collecting system based on improved bloom filter with the one based on bloom filter.The results show that the former get weibo data faster with a higher efficient to make URL distinct.At the same time,experimental results on actual dataset show that UR-LDA can offer an effective solution to topic mining for weibo.