Data Mining Research Lab DMRL graduate students Sitaram Asur and Duygu Ucar along with their advisor Professor Srinivasan Parthasarathy have received a Best Paper (in the applications category) award for their work on An Event-based Framework for Characterizing the Evolutionary Behavior of Interaction Graphs at the annual ACM Knowledge Discovery and Data Mining (SIGKDD) conference in 2007. A video recording of the presentation made by Sitaram Asur at the conferenceis available here. The best paper selections were revealed by the awards committee at the opening ceremony of the conference in San Jose.
The work presented in this paper presents a novel approach for modeling and mining evolving interaction networks that are becoming increasingly ubiquitous in social, behavioral, biological and scientific settings. The key ideas brought forth by this work is a structured way to reason about how communities and individual elements within such networks evolve over time and what are the critical events that characterize their behavior. The authors demonstrate how behavioral indices such as stability and influence as well as a diffusion model can be efficiently composed from the events detected by their framework and can be used to effectively analyze real-life evolving networks in an incremental fashion.
This represents the seventh award nomination and fourth best paper award for the DMRL group over the last 5 years -- a truly remarkable streak. Previous best paper awards received by the group include ones at the IEEE International Conference on Data Mining (ICDM) in 2002, the SIAM International Conference on Data Mining (SDM) in 2003 and at the Very Large Databases Conference (VLDB) in 2005. Previous nominations for an award, including "best-of conference" selections, were received at SDM in 2005, at SIGKDD in 2006, and at ICDM in 2006. SIGKDD, ICDM and SDM are the top conferences in the field of knowledge discovery and data mining and VLDB is one of the top conferences in the field of database systems.