Finding influential nodes via isolated local average shortest path with an extended neighborhood

Abstract

Effectively determining the most prominent seed nodes is becoming more important as networks rapidly emerge and play a crucial role in spreading information or disease across various complex network applications. Many centrality measures exist for identifying these highest influence nodes; however, they have drawbacks, such as considering only local or global information and ignoring neighboring node interactions, despite dense networks. To mitigate these issues, we propose a new centrality metric known as Local Average Isolating Centrality (LAISC). It integrates isolating centrality, which finds nodes whose removal severely impairs network connectedness by increasing the number of unconnected components, along with the neighbor’s isolating centrality influence. Additionally, it incorporates the node’s local relative change in the average shortest path along with it’s neighbor nodes, by applying average shortest path analysis to the subgraph induced by the node. We incorporates the extended neighborhood concept for finding the nearest neighbor nodes for all nodes in a graph. The efficacy of LAISC is tested using the epidemic model like susceptible-infected-recovered (SIR) model on multiple real-world datasets against both recent and traditional centrality measures. The evaluation of LAISC is calculated using Kendall’s tau correlation coefficient. Our method LAISC performs well in terms of information spread compared to other existing centrality measures. Experimental findings show that LAISC surpasses traditional centrality measures in effectively spreading information across complex networks. In comparison to traditional measurements, the suggested LAISC achieves improvements ranging from 0.26% to 5.76% across several datasets, demonstrating superior dissemination of information.