HN-Sim: A Structural Similarity Measure over Object-Behavior Networks

Measurement of similarity is a critical work for many applications such as text analysis, link prediction and recommendation. However, existing work stresses on content and rarely involves structural features. Even fewer methods are applicable for heterogeneous network, which is prevalent in the real world, such as bibliographic information network. To address this problem, we propose a new measurement of similarity from the perspective of the heterogeneous structure. Heterogeneous neighborhood is utilized to instantiate the topological features and categorize the related nodes in graph model. We make a comparison between our measurement and some traditional ones with the real data in DBLP and Flickr. Manual evaluation shows that our method outperforms the traditional ones.