Publication
International Conference on Natural Computation 2007
Conference paper

Online detecting and tracking of the evolution of user communities

View publication

Abstract

Clustering, known as divide a set of static data points into densely distributed groups, has long been a well-known research area. However, many real-life problems require a novel and generalized form of clustering, the evolutionary clustering. Given a dynamic set of data points that may move, disappear and emerge, the evolutionary clustering is to track the move, disappear and emerge of the corresponding clusters. In this paper, we propose converting this novel problem into an iterative form of learning a mixture model, and present a structural-EM algorithm as the solution. © 2007 IEEE.

Date

Publication

International Conference on Natural Computation 2007

Authors

Topics

Share