AutoPart: Parameter-free graph partitioning and outlier detection

Abstract

Graphs arise in numerous applications, such as the analysis of the Web, router networks, social networks, co-citation graphs, etc. Virtually all the popular methods for analyzing such graphs, for example, k-means clustering, METIS graph partitioning and SVD/PCA, require the user to specify various parameters such as the number of clusters, number of partitions and number of principal components. We propose a novel way to group nodes, using information-theoretic principles to choose both the number of such groups and the mapping from nodes to groups. Our algorithm is completely parameter-free, and also scales practically linearly with the problem size. Further, we propose novel algorithms which use this node group structure to get further insights into the data, by finding outliers and computing distances between groups. Finally, we present experiments on multiple synthetic and real-life datasets, where our methods give excellent, intuitive results. © Springer-Verlag Berlin Heidelberg 2004.