Link Predictability Analysis of US Political Blog Network with Structural Perturbation Method

Abstract

Many link prediction methods have been developed in past few decades in network science, and most of them are tailored for specific fields that lose generalities. Luckily, a recent work named Structural Perturbation Method (SPM) proposed a consistency index of network organization without priori knowledge, and it did not need to test those predicting methods first. Since demonstrating whether there is a link between two nodes is usually costly, we want to replace link confirmation with link prediction. In this paper, we use the SPM method to study the structure of US political blog network and analyze the link predictability at different network scales. The experimental results show that we can obtain basic intrinsic features of the network structure and can get ideal prediction results as long as 20% of the network structures.