An Improved Algorithm for Label Propagation Based on Rough Core

Abstract

Complex networks are a hot topic in social science research today. Analyzing the structure of the relationships embedded in them through community discovery techniques is an important tool for the study of complex networks. To address the randomness problem of label propagation in label propagation algorithms, this paper proposes an improved algorithm for label propagation based on rough kernels. The algorithm first performs the extraction of rough kernels and assigns unique labels to each rough kernel to achieve fast label preprocessing; then intervenes in the label propagation process through the ranking of node influence to reduce its randomness and make the division results more stable; finally, the community division is performed by the distribution of network labels. Experiments are conducted in real networks, and the results show that the combined performance of the method in this paper is better than the traditional labeling algorithm in terms of modularity and NMI metrics, and the proposed improved algorithm leads to improved community division results.