Using Data Depth vs. Depth Classifier for Detecting Communities in Networks

• Activity Number: 548
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Yahui Tian*
• Companies: The University of Texas at Dallas
• Keywords: spectral clustering ; DD-classifier ; data depth ; K-means ; misclassification rate ; nonparametric

Abstract

We propose a new nonparametric spectral clustering algorithm for detecting communities in large networks using the Depth vs. Depth (DD) classifier. Under the stochastic block model (SBM) and pre-defined number of clusters, we study the performance of four data depth functions, i.e. simplicial depth, half-space depth, Mahalanobis depth, and projection depth. We investigate the effect of regularization on a classification error and compare the new DD classifier on networks with the regularized clustering algorithm based on the K-means approach.