Affordable Access

Access to the full text

Correspondence analysis-based network clustering and importance of degenerate solutions unification of spectral clustering and modularity maximization

Authors
  • Kimura, Masaomi1
  • 1 Shibaura Institute of Technology, 3-7-5, Toyosu, Koto City, Tokyo, 135-8548, Japan , Koto City (Japan)
Type
Published Article
Journal
Social Network Analysis and Mining
Publisher
Springer Vienna
Publication Date
Aug 25, 2020
Volume
10
Issue
1
Identifiers
DOI: 10.1007/s13278-020-00686-z
Source
Springer Nature
Keywords
License
Yellow

Abstract

Methods to find clusters in a network have been studied extensively because clustering has practical importance in many applications. Commonly used methods include spectral clustering and Newman’s modularity maximization. However, there has been no unified view of the two methods. In this study, we introduce an innovative guiding principle based on correspondence analysis to obtain node coordinates and discuss its equivalence to spectral clustering and Newman’s modularity. Besides, we discuss a degeneration case and its significance.

Report this publication

Statistics

Seen <100 times