| Social network is a kind of web service that spring up rapidly in recent years. Twitter and Sina Weibo are the representatives of the second generation of social network services, i.e. the microblogging services. With the advancing of mobile Internet, microblogging has become the most popular platform of great influence for information propagation with an enormous amount of user social data.At present, the analysis and data mining to social network services are becoming the focus. However, the existing methods to handle the recommendation mostly consider the relationships between users. They usually take the structure of social networks into account, to analyze the interest of users based on the relationships thus to recommend. They merely consider about the contents users post.To deal with the problems raised above, this paper proposes an algorithm named Microblogging Content Recommendation Algorithm,MCRA for short. MCRA is a kind of strategy by considering to model users according to the contents they post. The main strategies are as follows:(1) The User's Topic Model would be trained by LDA algorithm. The computing method for similarity between candidate microblog and User's Topic Model would be introduced. The User's Keywords Vector Model would be built based on TF-IDF algorithm. The similarity between candidate microblog and User's Keywords Vector Model would be calculated.(2) The recommendation method for microblogs based on the scores of candidate microblogs would be introduced. The scores of candidate microblogs are computed by a weighted method combining the matching degree between candidate microblogs and User's Topic Models as well as the matching degree between candidate microblogs and User's Keywords Vector Model.The experiment results show that the algorithm introduced by this paper can perform better than the baseline algorithms by average precision(AP). Thus it can solve the content based recommendation problems targeting users. |
PDF Full Text Request