When Will a Repost Cascade Settle Down?

Repost cascades play a critical role in information diffusion on social media sites. They are developed by series of reposts and stop eventually. Substantial previous work has studied and predicted various aspects of repost cascades such as growth, burst and recur. However, how or even whether it is possible to predict when a repost cascade will settle down remains to be an open problem. Existing models cannot be directly applied to solve the problem as the feature based models are sensitive to features, while the point process based models assume that the followers of all reposters are disjoint. In this paper, we propose a novel definition settling time to model this problem. We develop a point process based model to get rid of the restriction in previous studies and make an accurate prediction of the settling time. We conduct an extensive set of experiments on Sina Weibo dataset. The results show that our model achieves over 10% performance gain than the state-of-the-art approaches after observing the cascades for 24 h.