Evolutionary Prediction of Nonstationary Event Popularity Dynamics of Weibo Social Network Using Time-Series Characteristics

Figure 3

A popularity prediction framework for information propagation of a hot event on the Weibo social network. First, the explanatory (time) variables of an event were determined by the data preprocessing, and the key variables affecting the event propagation were identified by the correlation analysis. Second, the framework recognizes the propagation type (stationary or nonstationary) of the event. Third, a DLM is established based on the variables and the event propagation type. Consequently, time-series analysis can be conveniently achieved based on the DLM.