Effects of Network Structure on Information Diffusion Reconstruction

Abstract

This paper considers the effect of network structure on the reconstruction of information diffusion in a network. We employ the independent cascade model and a generalized independent cascade model to describe the network diffusing process with a single influence attempt and multiple influence attempts occurred between a pair of nodes, respectively. The diffusion reconstruction is formulated as a maximization likelihood problem. Based on this, we investigate the effect of the node number and the edge density of a network on the performance of diffusion reconstruction with numerical experiments on synthetic and real networks. The results show that reconstruction accuracies are inversely related to the node number and nonlinearly depends on the edge density. We also discuss the effect of the number of influence attempts in diffusion on the reconstruction accuracy.