Identification of Key Propagation Nodes in Complex Networks Based on Weighted Multi-Feature Fusion and Approximate Influence Radius

Abstract

Identifying key propagation nodes in complex networks is an important research topic. We propose a new gravity model based on weighted multi-feature fusion and approximate influence radius (WMGM) to identify key propagation nodes. The core of this method is to first determine the approximate influence radius of nodes based on node similarity and network structure. Secondly, the normalized maximum eigenvector was introduced, and the element value of the eigenvector was regarded as the node weight value. Then, the K-shell value, degree value, and PageRank centrality of the node are fused, and the fused value is used as the mass of the node. Finally, based on the multi-feature fusion gravity model with weight attribute, the interaction force between nodes was calculated, and the importance score of nodes was determined by accumulating the interaction force of all nodes within the approximate influence radius. The WMGM method is compared with the classical centrality methods, the similar methods, and the state-of-the-art methods on 10 different real datasets. The experimental results show that the WMGM method can effectively identify the top 10 critical nodes in different networks, and the top 200 identified nodes are highly similar to the standard ranking results. In addition, the WMGM achieves high node ranking accuracy across all 10 datasets, attaining the best overall performance on 80% of them.