Hybrid Weibo Tags and Topic Mining for User Similarity Model

Abstract:
With the rapid development of personalized services such as collaborative filtering, user similarity calculation has become a key issue for such services. One of the hot research perspectives is to explore the interests of users from the perspective of Weibo content and analyze the similarities between users. Tags can reflect users’ explicit opinions and interests, and Weibo content can reflect users’ hidden interests. We constructed a user similarity model based on users’ tags and topic, and set the model's adaptable adjustment parameters. Its effectiveness is confirmed on the collected real Weibo data set. The experimental results show that compared with the traditional tag-based user similarity model, the new model obtains a better index and Rank Accuracy index.