A novel semi local measure of identifying influential nodes in complex networks

Abstract

How to identify influencers is very significance in mastering the nature of node, controlling spreading process in complex networks. In this research field, each method has its own advantages and limitations. For example, local metrics are relatively simple, global metrics can give better results, but the computational complexity is also relatively high. A semi-local approach on basis of node dimension is proposed to identify influencers. The node dimension can detect regions with different dimensional structures by scaling the local dimension on the scale. The saturation effect is discovered in the process of identifying influencers by the node dimension. When the maximum dimension radius is close to the mean shortest path length of networks, the method has better performance. Through the saturation effect, our approach can be a tradeoff between local and global metrics. In addition, we show the correlation between different measures and node dimension with different maximum dimension radii. We employ Susceptible-Infected-Recovered (SIR) model to verify the effectiveness of our designed approach. Simulation results indicate the superiority of our algorithm.