Discovery and Prediction with Transactional Network Data

• Activity Number: 485
• Type: Invited
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: SSC
• Author(s): Hugh Chipman*+ and Mahdi Shafiei
• Companies: Acadia University and Dalhousie University
• Keywords: Social network ; complex data ; transactions

Abstract

Network data arise in a wide variety of contexts including biology, computers, social interactions and email communication. Models often focus on static network data: a collection of objects (e.g. people) and relations (e.g. friendship) between them. Community discovery in social networks is a challenging problem given the sparse and dynamic nature of these networks. Link prediction (prediction of an edge) is another fundamental problem. Recently, a mixed membership stochastic block model (Airoldi et. al. 2008) has been proposed to simultaneously identify communities and predict edges for binary relational data. The method is limited to data with single pairwise relations between objects. We consider an extension of this model to transactional networks (eg email) with multiple transactions (e.g. multiple emails) and one-to-many relations (e.g. one sender and multiple recipients). Several examples will be used to illustrate the method.