Data Mining Work Receives Best Student Paper Award at KDD 2012

University of Illinois computer science graduate student Yizhou Sun has received the Best Student Paper Award at the 2012 ACM Conference on Knowledge Discovery and Data Mining (KDD 2012) for her paper, “Integrating Meta-Path Selection with User-Guided Object Clustering in Heterogeneous Information Networks.”  The research focuses on how to better discover the data that is present in information networks.

Link-based clustering, where data objects are grouped by relationships rather than attributes, is increasingly important in order to discover new knowledge in large social or information networks.  While most research has focused on homogeneous networks, where links all carry the same semantic meaning and only differ in strengths, most real-world networks are heterogeneous – that is, there may be multiple types of objects that are linked together by different types of relationships.  The difficulty in mining these networks is in selecting the proper relationships to explore in order to define and detect the clusters that are relevant to the user.

Illinois graduate student Yizhou Sun

In her thesis work on mining heterogeneous information networks, Sun has created a new methodology for social and information network analysis. She investigates networks with semi-structures, where multiple types of objects and links are involved, and she mines such networks by exploring the semantics of different types of links systematically. In the KDD 2012 paper, by asking a user to provide limited information as guidance, her algorithm can automatically determine which type of linkage needs to be considered in the processing of clustering, and finally output the user-desired clustering results.

Sun conducts her research under the supervision of computer science professor Jiawei Han, and will join Northeastern University as an assistant professor in January 2013.  Her paper was co-authored with other members in Han’s group, Brandon Norick and Xiao Yu, as well as UCSB professor Xifeng Yan and UIC professor Philip S. Yu.  KDD is the largest and most prestigious data mining conference.  This year, Han’s research group presented an impressive five full research papers and one demo paper at the conference, including the winners of the Best Student Paper and Best Poster Awards.