Modeling User Interests Based On Multi-features Fusion For Micro-blog

In recent years, micro-blog applications have been popular around the world which lead micro-blog to be a virtual place for online people present their comments freely. However, the special social impact, user behaviour modes and recommendation based on users’ interests have been concerned by scholars.Among these problems, one of the hottest research problems is the personalized recommendation based on users’ interests. Two key research problems are blog users’ interest model and fake users’ identification. In this thesis, we propose a scheme to model micro-blog users’ interests for personalized recommendation services, and we propose a method to identify fake accounts. According to them, we need to finish the following jobs.For question about micro-blog user’s interests model, we proposed three aspcets: the representation way for micro-blog user interest model, micro-blog user feature extraction and the establishment of user interest model. The users’ interests should be shown by some modes, meanwhile, model is a straight way to represent users’ interests. Variours interests models can affect the model ways or choices of model information, and the succeed computation. So appropriate models for users’ interests are very important. With the extraction of micro-blog user feature, analyzing the structure and content of micro-blog; with the result of analysis, select correlated three parts of micro-blog as data source for building user interest model. Then using the topic model, LPA and other methods to extract some keywords which can appropriately reflect user’s interest. For establishment of user interest model, according to the analysis of the structure and content of the micro-blog, we propose a fusion algorithm to fuse three kinds of data. That is basic personal information, micro-blog content and social relationship, we adopt linear fusion algorithm.As for fake user identification, analyze and investigate fake users ’ characteristics, then extract critical features which are different from ordinary users,and propose a method for detecting fake user accounts.With the analysising about the above research content, we prompt micro-blog user interest model validation and fake identification algorithm verification. By analyzing the results of coverage accuracy in the experimental, we verify the validity of the modeling programs; and we verify user identification algorithms to identify fake user’s accuracy, finally we verify the feasibility of the algorithm.