Identifying influential nodes on directed networks

Identifying influential nodes on directed networks is a challenging and widely studied task that keeps drawing extensive attention from both academia and industry. The simultaneous consideration of both local and global structural information has demonstrated its effectiveness in identifying influential nodes on un-directed networks. Nevertheless, how to better utilize these two types of information on directed networks remains a challenge. In this paper, we address the influential nodes identification problem for directed networks where a node can directly affect its in -neighbors, like the social network of Twitter. We present a general iterative framework that integrates both local structural information and global influence. The global influence exerted by the target node is determined as the cumulative sum of the global influences originating from its in -neighbors, achieved through an iterative procedure. Meanwhile, the in -degree of the target node is leveraged to capture local structural information, which is consistently reinforced throughout the iterative process to prevent the attenuation of its significance over successive iterations. Our algorithm demonstrates significant performance improvement, averaging 21.61% in Kendall's tau and 23.43% in precision@0.05 over the 8 benchmarks across 15 real networks. Moreover, it outperforms the benchmarks on artificial networks, and can effectively identify fast influencers.