Topic reading dynamics of the Chinese Sina-Microblog

Abstract

A topic in the Sina-Microblog consists of multiple Weibos sharing a specific set of key words. Therefore, a reader of a Weibo may be susceptible to other Weibos of the same topic, and hence may undergo multiple transitions between the susceptible and the exposed states before eventually becoming infectious. This complicates the reading dynamics that traditional epidemic susceptible-exposed-infectious compartmental models cannot capture, and requires a different setting than a single Weibo forwarding dynamics model. Here we formulate a topic reading dynamics model; introduce some summative indices characterizing the reading “outbreak” potential; derive analytic formulae to calculate these indices; and examine the sensitivity of these indices and the ability of prediction on model parameters and on sampling frequencies. We conduct some numerical experiments based on historical data of popular reading topics in the Chinese Sina-Microblog. In our experiments, different sampling frequencies (4 h, 8 h or 12 h) nearcasting the turning points are feasible, in particular, a good nearcasting prediction for the accumulated R-users are 72 h with the sampling frequencies of every 4 h and every 12 h and 78 h with the sampling frequency of every 8 h.