Edge Clustering Coefficient based Label Initialization for Label Propagation Algorithm to Detect Community Structures in Complex Networks

Abstract

Identifying community structure in complex networks is a critical process that divides the entities according to their similarities in characteristic or behavior that define and control the function and organization of networks. One of the fastest and simplest community detection algorithms is the label propagation algorithm (LPA). However, the LPA produces different results in each run due to the randomness in label propagation, leading to uncertainty and instability to the detected communities. To address this problem, several algorithms have been proposed which mainly concentrates on eliminating randomness. In this paper, an improved label propagation method (ECLI-LPA) based on edge clustering coefficient-based label initialization has been proposed. In ECLI-LPA, instead of assigning unique labels to every node, the same labels are assigned to nodes whose edge clustering coefficient is above a threshold value to detect communities. The experimental results on real-world networks and synthetic networks show that the proposed method improves stability and performs better than the compared algorithms.