Two Models of Random Intersection Graphs for Classification

Abstract

Graph concepts generally are usefulfor defining and detecting clusters. We consider basic properties of random intersection graphs generated by a random bipartite graph BGn, mon n+m vertices. In particular, we focus on the distr ibution of the number of isolated vertices, and on the distribution of the vertex degrees. These results are applied to study the asymptotic properties of such random intersection graphs for the special case that the distribution P(m) is degenerated. The application of this model to find clusters and to test their randomness especially for non-metric data is discussed.