基于广度-深度采样和图卷积网络的谣言检测方法
|
王友卫,王炜琦,凤丽洲,朱建明,李洋
|
Rumor detection method based on breadth-depth sampling and graph convolutional networks
|
Youwei WANG,Weiqi WANG,Lizhou FENG,Jianming ZHU,Yang LI
|
|
表 5 不同谣言检测方法在3种数据集上的实验结果 |
Tab.5 Experimental results of different rumor detection methods in three datasets |
|
方法类别 | 方法名称 | Weibo2016 | | CED | | Twitter-COVID19 | Acc | F1 | | Acc | F1 | | Acc | F1 | 基于传统机器学习 | DTR | 0.732 | 0.733 | | 0.672 | 0.668 | | 0.377 | 0.329 | DTC | 0.831 | 0.825 | | 0.740 | 0.741 | | 0.492 | 0.426 | SVM-TS | 0.857 | 0.859 | | 0.746 | 0.756 | | 0.510 | 0.498 | 基于事件传播序列 | GRU | 0.898 | 0.899 | | 0.861 | 0.864 | | 0.498 | 0.401 | RvNN | 0.908 | 0.908 | | 0.892 | 0.891 | | 0.540 | 0.391 | PLAN | 0.932 | 0.936 | | 0.916 | 0.913 | | 0.573 | 0.432 | 基于事件传播结构 | Bi-GCN | 0.927 | 0.928 | | 0.894 | 0.898 | | 0.616 | 0.415 | RDEA | 0.921 | 0.921 | | 0.910 | 0.916 | | 0.638 | 0.504 | EBGCN | 0.937 | 0.935 | | 0.880 | 0.879 | | 0.589 | 0.563 | ACLR-BiGCN | 0.924 | 0.922 | | 0.898 | 0.903 | | 0.765 | 0.686 | UPSR | 0.934 | 0.928 | | 0.896 | 0.895 | | 0.602 | 0.587 | BDS-GCN | 0.944 | 0.943 | | 0.928 | 0.928 | | 0.682 | 0.674 |
|
|
|