Content Centrality Measure for Networks: Introducing Distance-Based Decay Weights

We propose a novel centrality measure that is called Content Centrality for a given network that considers the feature vector of each node generated from its posting activities in social media, its own properties and so forth, in order to extract nodes who have neighbors with similar features. We assume that nodes with similar features are located near each other and unevenly distributed over a network, and the density gradually or rapidly decreases according to the distance from the center of the feature distribution (node). We quantify the degree of the feature concentration around each node by calculating the cosine similarity between the feature vector of each node and the resultant vector of its neighbors with distance-based decay weights, then rank all the nodes according to the value of cosine similarities. In experimental evaluations with three real networks, we confirm the validity of the centrality rankings and discuss the relation between the estimated parameters and the nature of nodes.