FP7 Logo

EC Project 257859

European Union

Co-funded by the 
European Union
 

ROBUST Investigates Structural Changes in Networks

Researchers from the University of Koblenz-Landau will present results of ROBUST research at this year's ICWSM conference. This work is concerned with the link dynamics in evolving knowledge networks and has the goal to detect links that will be removed, links that will appear or that remain stable. Preserving and developing the knowledge contained in business communities is crucial for the success of these communities. Existing knowledge in the system should be preserved whereas unconnected but related knowledge items should be connected.

Additions and removals are inherent in many networks but how can the interplay of knowledge growth and decay processes be predicted? Whereas traditional link prediction methods evaluate methods by how well they can solve the link prediction problem, we conjecture that suitable link-prediction functions should also distinguish between links to be added and instable links that may appear in the next step, but then may disappear and even appear again etc. Analogously, a good link-removal prediction method should distinguish between links that are instable, i.e. will disappear and then appear again and so on, and links that will be removed for the long run. Thus, structural changes were classified into four states: growth, decay, stability and instability of network links. 

Using the largest knowledge network, the Wikipedia hyperlink network, structural characteristics were then evaluated as to how far they are indicative of one of these four states. 

Links

Structural Dynamics of Knowledge Networks, Julia Preusse, Jérôme Kunegis, Matthias Thimm, Thomas Gottron, and Steffen Staab. ICWSM'13: Proceedings of the 7th International AAAI Conference on Weblogs and Social Media, (2013)

 

 

Comments (0)