Blogger-link-topic model for blog mining

Abstract

Blog mining is an important area of behavior informatics because produces effective techniques for analyzing and understanding human behaviors from social media. In this paper, we propose the blogger-link-topic model for blog mining based on the multiple attributes of blog content, bloggers, and links. In addition, we present a unique blog classification framework that computes the normalized document-topic matrix, which is applied our model to retrieve the classification results. After comparing the results for blog classification on real-world blog data, we find that our blogger-link-topic model outperforms the other techniques in terms of overall precision and recall. This demonstrates that additional information contained in blog-specific attributes can help improve blog classification and retrieval results. © 2012 Springer-Verlag.