Fast parallel community detection algorithm based on modularity

Ehsan Moradi, Hadi Tabatabaee Malazi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

9 Citations (Scopus)

Abstract

In recent years, detecting dense sub-graphs that are known as communities in massive graphs has been a common issue in different fields of science. It provides the facility of studying complex graphs by simplifying them through utilizing communities. Due to ceaseless increases in graph size that are used in social networks (with billions of nodes and edges), algorithm execution time is an important factor for detecting communities. To cope with this problem, a new parallel community detection algorithm is presented in this paper. The main idea behind the proposed method is to assign parallel threads for the calculation of adding qualified neighbor nodes to the community. Proposed algorithm is tested using a general PC (IntelCorei7, 4 GByte). It leads to abating the algorithm execution time from 25% to 78% compared to the fastest previous parallel algorithms.

Original languageEnglish
Title of host publication18th CSI International Symposium on Computer Architecture and Digital Systems, CADS 2015
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
ISBN (Electronic)9781467380232
DOIs
Publication statusPublished - 8 Jan 2016
Event18th CSI International Symposium on Computer Architecture and Digital Systems, CADS 2015 - Tehran, Iran, Islamic Republic of
Duration: 7 Oct 20158 Oct 2015

Publication series

Name18th CSI International Symposium on Computer Architecture and Digital Systems, CADS 2015

Conference

Conference18th CSI International Symposium on Computer Architecture and Digital Systems, CADS 2015
Country/TerritoryIran, Islamic Republic of
CityTehran
Period7/10/158/10/15

Keywords

  • Community detection
  • Massive graphs
  • Parallel algorithm

Fingerprint

Dive into the research topics of 'Fast parallel community detection algorithm based on modularity'. Together they form a unique fingerprint.

Cite this