Joint Modeling of Communities and Node Features in Networks

• Activity Number: 439
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Yuan Zhang*+ and Liza Levina and Ji Zhu
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: Community detection ; Overlapping community membership ; Node features ; Joint modeling

Abstract

Community detection has been a fundamental problem in statistical network analysis. A lot of work has been done in the field of network community detection. Most past work focused on studying the network structure. We propose a joint estimation model that incorporates information from both the network structure and the features on nodes. We tested our model under both simulated and real world networks. Simulation studies show that our model is robust and enjoys a higher detection accuracy over benchmark models. Data applications show that our model aligns with the community structure in real world networks. We also describe how our model identify node features that are significantly related to the network structure.