An Improved Label Initialization based Label Propagation Method for Detecting Graph Clusters in Complex Networks

Community structure is one of the fundamental characteristics of complex networks. Detection of community structure can provide insight into the structural and functional or-ganization that helps to understand various dynamical processes such as epidemics and information spreading. Label propagation algorithm (LPA) is a well-known method for community struc-ture identification due to linear time complexity. However, the communities extracted by the LPA is unstable since it produces different combinations of communities at each run on the same network. In this paper, a novel label initialization method for label propagation algorithm (ILI-LPA) is proposed to detect stable and accurate community structures. The proposed ILI-LPA focuses on more accurate label initialization rather than assigning unique labels thereby reduce the effect of randomness in LPA. The experiments on several real-world and synthetic networks show that the ILI-LPA improves the quality and stability of communities compared to existing algorithms. The results also demonstrate that appropriate label initialization can significantly improve the performance of label propagation algorithms, and the stability has been improved up to 50-78% relative to the standard LPA.

@article{Chandran2022,
  title     = {An Improved Label Initialization based Label Propagation Method for Detecting Graph Clusters in Complex Networks},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {13},
  number    = {5},
  year      = {2022},
  publisher = {The Science and Information Organization},
  author    = {Jyothimon Chandran and V Madhu Viswanatham},
  doi       = {10.14569/IJACSA.2022.0130588},
  url       = {https://doi.org/10.14569/IJACSA.2022.0130588}
}