Spreader node detection based on the Perron-Frobenius theorem in complex networks

Spreader node detection is a fundamental research topic in complex networks aiming to detect the most influential nodes of the network. Feedback-based centrality measures a well-known approach assumes that the importance of a node depends on how important are the neighbor nodes. Eigenvector as a basic feedback-based approach, considers the power of neighborhood, but the information of neighborhood and the connection between neighbors are neglected. Regarding this issue, in this paper, we propose a novel node centrality measure based on the Perron-Frobenius theorem. A novel weighting technique is introduced based on the local importance of the graph edges. Then, the problem of central node detection is transferred to an eigenvector problem based on the newly defined weighted matrix. Experimental results indicate the superior performance of the approach in all the real-world datasets.