Original Paper

Journal of the Brazilian Computer Society

, Volume 17, Issue 1, pp 19-29

First online:

Open Access This content is freely available online to anyone, anywhere at any time.

A graph clustering algorithm based on a clustering coefficient for weighted graphs

  • Mariá C. V. NascimentoAffiliated withInstituto de Ciências Matemáticas e de Computação, Universidade de São Paulo Email author 
  • , André C. P. L. F. CarvalhoAffiliated withInstituto de Ciências Matemáticas e de Computação, Universidade de São Paulo

Abstract

Graph clustering is an important issue for several applications associated with data analysis in graphs. However, the discovery of groups of highly connected nodes that can represent clusters is not an easy task. Many assumptions like the number of clusters and if the clusters are or not balanced, may need to be made before the application of a clustering algorithm. Moreover, without previous information regarding data label, there is no guarantee that the partition found by a clustering algorithm automatically extracts the relevant information present in the data. This paper proposes a new graph clustering algorithm that automatically defines the number of clusters based on a clustering tendency connectivity-based validation measure, also proposed in the paper. According to the computational results, the new algorithm is able to efficiently find graph clustering partitions for complete graphs.

Keywords

Clustering coefficient Graph clustering Combinatorial optimization