Efficient Parallel Algorithm for Approximating Betweenness Centrality Values of Top k Nodes in Large Graphs

TOROSLU İ. H., Suleymanli G.

Concurrency and Computation: Practice and Experience, vol.37, no.4-5, 2025 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 37 Issue: 4-5
  • Publication Date: 2025
  • Doi Number: 10.1002/cpe.70022
  • Journal Name: Concurrency and Computation: Practice and Experience
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Applied Science & Technology Source, Communication Abstracts, Compendex, Computer & Applied Sciences, INSPEC, Metadex, zbMATH, Civil Engineering Abstracts
  • Keywords: betweenness centrality, community detection algorithms, heuristics, Louvain clustering, multiprocessing, parallelization
  • Middle East Technical University Affiliated: Yes

Abstract

Computing betweenness centrality (BC) in large graphs is crucial for various applications, including telecommunications, social, and biological networks. However, the huge size of the data presents significant challenges. In this paper, we introduce a novel approximate approach for efficiently extracting top k BC nodes by combining the Louvain community detection algorithm with Brandes' algorithm. Our method significantly enhances the runtime efficiency of the traditional Brandes' algorithm while preserving accuracy across both synthetic and real-world datasets. Additionally, our approach is suitable for parallelization, further improving its efficiency. Experimental results confirm the effectiveness of our method for large and sparse graphs.