A Label Propagation Algorithm Based on Neighbor Label Influence

The present study introduces a novel label propagation algorithm, namely NLI-LPA, which focuses on the influence of neighborhood labels to enhance the precision of community segmentation in complex networks. This algorithm aims to augment both the stability and accuracy of community detection while minimizing the impact of random strategies on the outcomes. The update sequence of the algorithm is determined based on the descending order of node importance, while incorporating neighbor label influence (NLI) to facilitate label selection and updates. Additionally, a community similarity measure is defined to reduce community dimensionality and output the community structure. Experiments were conducted on 8 artificial networks and 7 real networks to compare the performance of the NLI-LPA algorithm with that of the Fast Newman algorithm (FN), Spectral Clustering, LPA, and TNSLPA algorithms. The findings indicate that NLI-LPA outperforms the other four algorithms in evaluation metrics such as Normalized Mutual Information (NMI), Adjusted Rand Index (ARI), and Modularity (Q). These results indicate that the proposed algorithm exhibits enhanced accuracy and effectiveness in discovering network communities.