Feature representation for microblog followee recommendation in classification framework

In recent years, microblog has experienced an explosive growth, which makes it difficult to find useful information, especially for new users. Followee recommendation can help users exploring new friends and build more social connections, in order to filter out the useless information, and promote the development of the social network. However, the large amount of online users and the diverse and dynamic social data possess great challenges to support the function. In this paper, by proposing a novel feature representation of microblog users, we design a novel friend recommendation framework, which transform the recommendation problem to a binary classification problem. This system focuses on analyzing and extracting multi-dimensional features of microblog users on followee recommendation, and modeling classifiers to distinct recommended users from non-recommended users. Experimental results show that the classification framework with the efficient feature representation can overcome data sparsity shortcomings in single collaborative filtering and get better recommendation effects.