Hybrid-LDA:Incorporating Text And Link Information To Mine Interests

The development of social network brings great convenience to the interaction be-tween people. Over time, the social network accumulates a large number of users’data, which usually covers not only the user’s remarks, photos, and location information, but also includes the social relationships between users. These data afford researchers tremen-dous research value. Among the diverse research projects, mining user’s interest points is very significant, which can provide scientific basis for user’s personalized recommending.Currently, the commonly used model of mining user’s interests is topic model, which relies mainly on the users’comments. The users’comments can be interpreted as having probability distributions on the words, which means that the documents are represented as random mixtures over latent topics, where each topic is characterized by a distribution over words. At present, the LDA is the most efficient and classic topic model.In this paper, we use Twitter data as a background, the aim of this paper is mining the Twitter users’interests. We believe that the analysis of the users’interests not only depend on the contents of the users’own tweets, but also should depend on the other dimensions of users’information. Based on this idea, we propose an improved model called Hybrid-LDA. Through the experiments, being compared to LDA, Hybrid-LDA verifies the validity of the proposed model.