Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Complex network community discovery using fast local move iterated greedy algorithm

Salaheddine Taibi, Lyazid Toumi, Salim Bouamama

TL;DR

A novel algorithm, fast local move iterated greedy (FLMIG), which enhances the Louvain Prune heuristic using an iterated greedy (IG) framework to maximize modularity in non-overlapping communities.

Abstract

Examining the community structures within intricate networks is crucial for comprehending their intrinsic dynamics and functionality. The paper presents the Fast Local Move Iterated Greedy (FLMIG) algorithm, a novel method designed to effectively identify community structures in intricate networks. The FLMIG algorithm improves the modularity optimization process by including a rapid local move heuristic and an iterated greedy mechanism that switches between destructive and constructive phases to strengthen the community partitions. The main innovation is the integration of random neighbor moves with an enhanced Prune Louvain algorithm, which guarantees fast convergence while maintaining the connection of the identified communities. The results of our comprehensive studies, conducted on both synthetic and and real-world networks, clearly show that FLMIG surpasses existing cutting-edge techniques in terms of both accuracy and computing efficiency. This algorithm not only provides a strong tool for identifying communities, but also makes a valuable contribution to the broader field of network analysis by offering a method that can effectively handle large-scale and dynamically evolving networks.

Complex network community discovery using fast local move iterated greedy algorithm

TL;DR

A novel algorithm, fast local move iterated greedy (FLMIG), which enhances the Louvain Prune heuristic using an iterated greedy (IG) framework to maximize modularity in non-overlapping communities.

Abstract

Examining the community structures within intricate networks is crucial for comprehending their intrinsic dynamics and functionality. The paper presents the Fast Local Move Iterated Greedy (FLMIG) algorithm, a novel method designed to effectively identify community structures in intricate networks. The FLMIG algorithm improves the modularity optimization process by including a rapid local move heuristic and an iterated greedy mechanism that switches between destructive and constructive phases to strengthen the community partitions. The main innovation is the integration of random neighbor moves with an enhanced Prune Louvain algorithm, which guarantees fast convergence while maintaining the connection of the identified communities. The results of our comprehensive studies, conducted on both synthetic and and real-world networks, clearly show that FLMIG surpasses existing cutting-edge techniques in terms of both accuracy and computing efficiency. This algorithm not only provides a strong tool for identifying communities, but also makes a valuable contribution to the broader field of network analysis by offering a method that can effectively handle large-scale and dynamically evolving networks.
Paper Structure (25 sections, 5 equations, 10 figures, 6 tables, 6 algorithms)

This paper contains 25 sections, 5 equations, 10 figures, 6 tables, 6 algorithms.

Table of Contents

  1. Introduction
  2. Related work
  3. The problem definition
  4. Modularity optimization-based methods
  5. Greedy methods
  6. Metaheuristic methods
  7. Proposed approach
  8. Generate Initial Solution Algorithm
  9. FastLocalMove Algorithm
  10. Destruction solution Algorithm
  11. Reconstruction Heuristic Algorithm
  12. Select Next Solution Algorithm
  13. Complexity analysis
  14. Experimental Evaluation
  15. Problem instances
  16. ...and 10 more sections

Figures (10)

  • Figure 1: Examples of FLMIG procedures
  • Figure 2: The impact of $\beta$ values on the modularity value.
  • Figure 3: The impact of $\beta$ values on the computational time.
  • Figure 4: Convergence analyse of FLMIG, IG, ICG, CC-GA and DBAT-M algorithms
  • Figure 5: Experimental results of FLMIG and the compered algorithms on smaller and medium size real-world networks (1)
  • ...and 5 more figures