Research on Key Node Cluster Identification Algorithm Based on Louvain and Cycle Ratio

The identification of important nodes in complex networks is crucial in various real-world domains. Recognizing key node clusters has become a hot topic in the study of complex networks. While several algorithms have been proposed to identify key node clusters in static networks, many of them suffer from issues such as overlapping node influence and poor algorithm integration, leading to suboptimal propagation of identified key node clusters. To address these shortcomings, this paper proposes an algorithm, LC, for identifying key node clusters based on the Louvain method and CycleRatio (LC). The algorithm utilizes the Louvain community detection algorithm to partition the local community structure of the network, and within each community, it integrates the IT, IKS and \(E_i+\) metrics to identify key nodes. Furthermore, it optimizes the clusters using the CycleRatio (LC) to ensure that the key node clusters are distributed across the entire network while maintaining maximum influence within each local community. This approach reduces overlapping influences between different communities, optimizes information propagation and influence transfer within each community, and enhances the overall performance of the algorithm. Through simulation experiments using the SIR propagation model on 8 real static network datasets, the results demonstrate that the LC algorithm exhibits higher infection rates in most experiments compared to 7 benchmark algorithms, and it shows stronger adaptability to different networks. The LC algorithm effectively addresses the issue of overlapping node cluster influences, resulting in key node clusters with greater comprehensive influence. Compared to other algorithms, it can more accurately and stably identify key node clusters in static networks.