UserRBPM: User Retweet Behavior Prediction with Graph Representation Learning

Table 6

Prediction performance of different methods for retweeting behavior (%).

Methods	Precision	Recall	F1-score	AUC
Spatial- & Temporal-level & Handcrafted features + LR (ST&HC+LR)	69.74	71.58	70.65	77.27
Spatial- & Temporal-level & Handcrafted features + SVM (ST&HC+SVM)	68.38	69.15	68.76	78.01
DeepWalk + ST + GAT	78.21	78.46	78.28	82.81
DeepWalk + ST&HC + GAT	79.68	80.24	79.96	82.75
Node2vec + ST + GAT	78.54	81.50	79.99	82.53
Node2vec + ST&HC + GAT	79.88	81.25	80.55	82.96
Our method (UserRBPM)	81.97	82.58	82.27	83.21