Topic based automatic news recommendation using topic model and affinity propagation

Yonghui Wu, Yuxin Ding, Xiaolong Wang, Jun Xu

Abstract: This paper presents a topic based web news recommendation method combining Affinity Propagation (AP) and Latent Dirichlet Allocation (LDA), which could automatically find the topics exist in the web pages and recommend the topic based news to Internet users. The topic distance is defined using LDA, which is used to generate the topic distance matrix. AP clustering is used to cluster the web page collections into different topic clusters. In order to prove the effect of combining AP and LDA, we sampled web page collections with different topics and web page collections. A series of experiments are implemented in these web pages. The comparison of clustering result of AP with information distance and AP with LDA are presented. The experiments show that our method combining AP and LDA is effective in topic based news recommendation system.