Community discovery is a popular topic in social network analysis which facilitates several real-world applications. Most commonly, communities are discovered according to their structural similarities resulting in tightly connected user subgroups. Distinct from existing works, we focus on mining a novel form of community, called functional, where users with similar interests are grouped. The process of identifying the structurally-cohesive user groups with similar interests faces the challenges of dealing with the sparsed user interaction networks as well as the limited access to the user-generated content. To overcome this, we propose to use the user hierarchy inherent in interaction networks to identify connected user groups with similar interests. A novel hierarchy-guided functional community mining method, FCMiner, is developed to identify functional communities without utilizing content information. We empirically evaluate the effectiveness of FCMiner using several real-world datasets with different characteristics benchmarking the state-of-the-art community and hierarchy discovery methods. FCMiner is found to be more effective than the benchmarked methods for the interaction networks which demonstrate less reciprocity. Moreover, we propose to use frequent interactions to discover prominent functional communities. The empirical analysis validates the usefulness of identifying prominent functional communities to recognize influential/impactful users and roles.
展开▼
