Abstract: Two important data sources in social networks, i. e. the network structure and the user gene-rated contents, were combined to dynamically identify user interest. When building topic models, the topic distributions of contents for each user at each time are obtained. And features used for prediction are extracted by summarizing the topical information based on the social network structure. Finally, these prediction features are exploited to dynamically predict user interest via several classification methods, such as logistic regression and support vector machine. The effectiveness of the proposed method is illustrated based on the Sina Weibo dataset.

Key words: network structure, topic model, user interest, dynamic identification