ABSTRACT

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

Graphs provide a general representation or data model for many types of data, where pairwise relationships are known or thought to be particularly important.∗ Thus, it should not be surprising that interest in graph mining has grown with the recent interest in big data. Much of the big data generated and analyzed involves pair-wise relationships among a set of entities. For example, in e-commerce applications such as with Amazon’s product database, customers are related to products through their purchasing activities; on the web, web pages are related through hypertext linking relationships; on social networks such as Facebook, individuals are related through their friendships; and so on. Similarly, in scientific applications, research articles are related through citations; proteins are related through metabolic pathways, co-expression, and regulatory network effects within a cell; materials are related through models of their crystalline structure; and so on.