Spreading Rumours without the Network

In this paper we tackle the following question: is it possible to predict the characteristics of the evolution of an epidemic process in a social network on the basis of the degree distribution alone? We answer this question affirmatively for several diffusion processes-- Push-Pull, Broadcast and SIR-- by showing that it is possible to predict with good accuracy their average evolution. We do this by developing a space efficient predictor that makes it possible to handle very large networks with very limited computational resources. Our experiments show that the prediction is surprisingly good for many instances of real-world networks. The class of real-world networks for which this happens can be characterized in terms of their neighbourhood function, which turns out to be similar to that of random networks. Finally, we analyse real instances of rumour spreading in Twitter and observe that our model describes qualitatively well their evolution.