ECRModel: An Elastic Collision-Based Rumor-Propagation Model in Online Social Networks

Abstract

With the rapid development of online social networks (OSN), the influence of rumor propagation on social life raises great concern. Traditional rumor-propagation models, which do not fully consider the features of OSN, are not suitable for use in OSN. In this paper, we focus on discovering a pattern of rumor-propagation phenomena in OSN, and propose a novel rumor-propagation model, inspired by a ball elastic-collision model, called the elastic collision-based rumor-propagation model (ECRModel). We investigate the dynamics of ball elastic collisions, which is similar to the dynamics of rumor propagation between nodes in OSN. We adopt the parameter relationships of the elastic collision model and apply them to rumor propagation in social networks. In the ECRModel, we do not directly adopt the node classification categories of “Ignorants, Spreaders, and Stiflers”, but divide the user nodes into three types: 1) inactive and never spread rumors; 2) active and spread rumors forward; and 3) inactive but have previously spread rumors. We mathematically model node interaction attributes, and analyze the spreading probabilities and the steady state, considering both individual perspectives with detailed attributes and integral perspectives with node-state densities. At last, we conduct a series of simulations, and the results verify the correctness of the analytical results. We further investigate the effects of detailed properties on rumor propagation, such as average out-degree of OSN, rumor confusingness degree, and each node’s comprehensive influence.