EqRank: A Self-Consistent Equivalence Relation on Graph Vertexes

TL;DR

The paper introduces EqRank, a hierarchical clustering method for graph vertices based on a recursive equivalence relation, applied to citation graphs to produce a meaningful classification scheme.

Contribution

It presents a novel self-consistent equivalence relation for hierarchical clustering of graph vertices, demonstrated on citation networks.

Findings

Effective hierarchical clustering of citation graphs.

Produces a meaningful classification scheme for hep-th papers.

Method shows adequate classification performance.

Abstract

A new method of hierarchical clustering of graph vertexes is suggested. In the method, the graph partition is determined with an equivalence relation satisfying a recursive definition stating that vertexes are equivalent if the vertexes they point to (or vertexes pointing to them) are equivalent. Iterative application of the partitioning yields a hierarchical clustering of graph vertexes. The method is applied to the citation graph of hep-th. The outcome is a two-level classification scheme for the subject field presented in hep-th, and indexing of the papers from hep-th in this scheme. A number of tests show that the classification obtained is adequate.