A Jungle Community Detection Algorithm Based on New Weighted Similarity

This article investigates fields of discovering community structures, which is an essential key to understanding and uncovering organizational principles in complex networks. This tool is of theoretical interest and practical significance in modern science. Commonly, the community can be defined as a set of similar nodes led by a node core. Many algorithms use the above definition to find the community structure. However, most of them cannot reveal the node cores and failed to give excellent similarity precision. To overcome the above limitations, we propose a new weighted similarity WS that computes with high precision the closeness between the nodes and clusters based on all local information, including the weight information. A new metric called node strength NS is also proposed to extract the node core of each community. Furthermore, we propose an original method to detect the community structure based on jungle law. Our jungle algorithm (JA) detects the communities with a three-phase strategy: first, the network will be updated by re-weighting and removing the noisy edges. Second, considering the network as a jungle, the strongest or most influential nodes will attract by force just the neighboring weak nodes. Finally, we merge the obtained groups according to the proposed metrics until the community criterion is steady. The experiments on real-world networks are carried out, and the obtained results demonstrate that our JA and WS outperform several community detection methods and similarities.