188/4 E-Commerce Group
Institute of Software Technology and Interactive Systems
Vienna University of Technology
Favoritenstrasse 9-11/188, A-1040 Vienna, Austria

Exploring supervised methods for temporal link prediction in heterogeneous social networks

Authors: 
N. Rümmele, R. Ichise, H. Werthner
Publisher: 
Proceedings: 
WWW 2015 Companion
Pages: 
1363 - 1368
Year: 
2015
Type: 
Speech with proceedings
Hidden Keywords: 
Department Focus: 
Business Informatics
TU Focus: 
ISBN: 
Abstract: 
In the link prediction problem, formulated as a binary classification problem, we want to classify each pair of disconnected nodes in the network whether they will be connected by a link in the future.<br> We study link formation in social networks with two types of links over several time periods. To solve the link prediction problem, we follow the approach of counting 3-node graphlets and suggest three extensions to the original method. By performing experiments on two real-world social networks we show that the new methods have a predictive power, however, network evolution cannot be explained by one specific feature at all time points. We also observe that some network properties can point at features which are more effective for temporal link prediction.
Abstract German: