Group detection in mobility traces

—Analysis and modeling of wireless networks greatly depend on understanding the structure of underlying mobile nodes. In this paper we present two clustering algorithms to determine the number of groups and their identities: k-means chain and spectral clustering. Different from traditional k-means clustering, k-means chain can identify the number of groups in dynamic graphs, and the chaining process can also keep track of group trajectories over the entire trace. The second approach uses spectral clustering, which measures the similarities between each node pair to group nodes of similar behaviors. We show that critical information of a mobility trace, such as the number of groups and group members, can be precisely extracted with little or no prior knowledge of the properties of a trace.

• Type: Conference
• Year: 2010
• Where: IWCMC
• Authors: Yung-Chih Chen, Elisha J. Rosensweig, Jim Kurose, Donald F. Towsley
• Keywords: IWCMC 2010, K-means Chain, K-means Clustering, Spectral Clustering