Mining graphs and networks, a theory-based approach.

the availability of large amounts of increasingly complex data has created a need for suitable data mining algorithms. Graph mining is a recent research stream which applies data mining techniques to data represented with graphs. A significant amount of explorative studies has been performed, e.g. involving the direct extension of existing methods to graphs and the use of heuristics. However, in order to ensure a continuing progress, it is important to construct a theoretical foundation for this new domain, allowing a better understanding of the complexity and potential of the field of graph mining. This project aims at contributing to this by constructing a framework and by answering some important theoretical questions. The research will be inspired by concrete problems in real-world applications (e.g. chemistry and biology). Questions which will be addressed include: (1) In which situation frequent pattern mining and related tasks such as frequent closed patterns mining can be solved efficiently. (2) In which cases properties of graphs can be learned and predicted efficiently. (3) Which graph classes ocuur in applications and how can their properties be exploited and (4) how can one define pattern mining and predictive tasks in a good way for databases consisting of one large network?

Keywords:Graph mining, Learning theory, Graph theory, Complexity, Bio-informatics

Disciplines:Applied mathematics in specific fields, Computer architecture and networks, Distributed computing, Information sciences, Information systems, Programming languages, Scientific computing, Theoretical computer science, Visual computing, Other information and computing sciences