Postmortem Analysis of Decayed Online Social Communities: Cascade Pattern Analysis and Prediction

Figure 11

The figure shows feature ranking such that , where is the feature rank for the feature . The method used for generating the importance is random forests where the importance of a feature increases whenever a split in the tree using that feature minimizes the prediction error [43].

(a) Feature ranking for predicting cascade size

(b) Feature ranking for predicting cascade virality