ESTG: Entropy-Weighted Semi-Global Triangle Centrality and Directed Resistance Distance for Identifying Influential Nodes
Abstract
To accurately identify key spreading nodes in complex networks, this paper proposes a local gravity model, ESTG, based on entropy-weighted triangle structure and directed resistance distance. The main innovations of this method include: introducing the k-shell information of nodes into the triangle structure measurement, and using information entropy to measure the diversity of k-shell distribution within the triangles, thereby giving higher weights to triangles located in the core or with significant hierarchical differentiation, and constructing entropy-weighted triangle coreness; calculating resistance distances based on the k-order neighborhood subgraph of each node and using the pseudoinverse, introducing directed resistance distance to simultaneously capture multi-path coupling effects and asymmetry in propagation; extending the semi-global triangle centrality as the node mass, integrating it with directed resistance distance into the local gravity model, and obtaining the ESTG node influence score. The performance of various baseline methods is compared on nine real networks and three LFR synthetic networks. Experimental findings demonstrate that ESTG achieves superior performance over the comparison methods with respect to node discrimination, ranking accuracy, and top-k node identification.